| Author |
Message |
|---|
John  Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
                 
|
|
PPS (Sieve) Cobblestone Recalibration
PPS (Sieve) will be transitioning from an sr2sieve (original sieve) baseline to a tpsieve (current sieve) baseline. The project will be entering a new phase soon so this is a good time to begin the gradual transition.
What is happening
PPS (Sieve) cobblestone per WU is being decreased to reflect the majority use of the tpsieve application. This will be completed in two phases based upon historical transitions: 1) credit will drop from 4523 to 3607 to reflect the transition of sr2sieve to ppsieve; 2) credit will then drop from 3607 to 2251 to reflect the transition of ppsieve to tpsieve.
Credit will be decreased by ~200 cobblestones per week until those numbers are achieved. Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks.
Status
30 October - 2251
23 October - 2411
16 October - 2609
09 October - 2803
02 October - 3011
25 September - 3209
18 September - 3407
Paused
31 July - 3607 (typo of 3507)
24 July - 3719
17 July - 3923
10 July - 4127
3 July - 4327
2 July - 4523 Summary explanation
The past 14 months have seen extraordinary advances in how the Proth Prime Search is sieved. It has evolved from an sr2sieve dominated approach to a tpsieve dominated approach. 14 months ago, sr2sieve contributed to almost 100% of the sieve. However, since then that contribution has continued to decline to less than 2% as of today. tpsieve now contributes to over 98% of the sieve processing.
Additionally, we are approaching a transition to the 3M-6M sieve file in which sr2sieve was not used. Therefore, it is now appropriate to adopt tpsieve as the standard application instead of the "optimized" application.
Detailed explanation
The general method for coordinating a sieve at PrimeGrid is a manual start, manual distributed continuation, and possibly an automated continuation in BOINC.
For the prime searches at PrimeGrid, many different sieving programs are used:
- srsieve
- gcwsieve
- sr2sieve
- ppsieve
- tpsieve
- fpsieve
- AthGFNSv
For the Proth Prime Search, originally, the applications used were srsieve and sr2sieve. Now the applications are srsieve, ppsieve, and tpsieve.
srsieve is used to create the sieve. ppsieve is used to increase the sieve depth until the factor files produced by tpsieve become manageable. Afterwards, tpsieve takes over, first manually, and then automated in BOINC.
tpsieve quickly becomes the dominate contributing application to the sieve. For example, the next phase of 3M-6M was sieved to 1G using srsieve and then to 5T using ppsieve. tpsieve was then used in a manual coordinated effort to bring it to its current depth of ~650T. It will enter BOINC next. By then, tpsieve will have completed over 99% of the sieve.
Unlike most other BOINC projects, it has been PrimeGrid's policy to adopt the optimized applications as the standard application. This has happened to all of PrimeGrid's applications (LLR, AP26, sr2sieve, gcwsieve). Each significant update to an optimized application is followed by a recalibration of credit. While this has never made PrimeGrid popular, it has provided participants easy access to the most efficient applications without having to fumble around with app_info files.
What made the update from sr2sieve to tpsieve so strikingly visible is the level of optimization that tpsieve provides over sr2sieve. Also, the addition of a highly optimized GPU tpsieve application created an exceptional disparity between 32 bit sr2sieve and tpsieveCUDA especially when a GPU application did not exist for sr2sieve.
sr2sieve is still an excellent application and will continue to be used in the TRP Sieve. In fact, for that sieve, sr2sieve is actually more efficient than tpsieve. tpsieve needs bout 1700 k's per 1M range in order to be competitive.
sr2sieve could continue to inefficiently work on subsets of the current Proth Prime Search sieve and accomplish in decades what tpsieve can do in years.
sr2sieve was originally working on a 598 sequence 5M sieve file. The current sieve now is a 9996 sequence 3M file. As you can see, the search space is exceptionally larger than the previous space. In fact, the sieve file is roughly 3.2G, a size that sr2sieve may spend several hours just initializing…which brings up another advantage tpsieve brought to the Proth Prime Search - the ability to sieve without a data file; thus sieve initialization is just seconds instead of hours.
In essence, it would be supremely impractical to use sr2sieve with a 3.2G sieve file in BOINC. In fact, most host would run into memory issues attempting to sieve a file that large with sr2sieve.
____________
|
|
|
|
|
|
this means, I presume, that the 10% reduction in credit will take place at this time...
will there be a corresponding 10% reduction in processing time?
and does this also include CW Sieve?
Thank you. |
|
|
|
|
|
Project admins never cease to amaze me, you have a dedicated following who use pps sieve more than likely on their nvdia cards and are happy with the credit they are producing. You then doc the credit rating by 10% tell no one, then when the folks show up who are you DOING YOUR WORK and are not to happy, you then reinstate the old credit rating.
Communication makes a happy cruncher |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Communication makes a happy cruncher
Please see this post to understand what happened.
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
this means, I presume, that the 10% reduction in credit will take place at this time...
will there be a corresponding 10% reduction in processing time?
No and no, see updated first post.
and does this also include CW Sieve?
No, CW Sieve will remain untouched at this time.
____________
|
|
|
|
|
Communication makes a happy cruncher
Please see this post to understand what happened.
I already had read thread. The thread explains my OP, decrease by 10% without telling anybody, then reinstate when questions arise. There was no reason given until this thread.
I am also unclear on the future, I crunch only 2x260gtx what will be the outcome for gpu crunchers only in this transition
regards
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Communication makes a happy cruncher
Please see this post to understand what happened.
I already had read thread. The thread explains my OP, decrease by 10% without telling anybody, then reinstate when questions arise. There was no reason given until this thread.
This post clearly states about the miscommunication.
I am also unclear on the future, I crunch only 2x260gtx what will be the outcome for gpu crunchers only in this transition
The tpsieve GPU application will continue to be offered as is normal now. The same PPS (Sieve) WU's go to both CPU's and GPU's alike.
____________
|
|
|
|
|
|
On current credit rating I average 72x4523 around 325k/330k ppd
The new lowest credit rating 72x3607 so around 255k/265k ppd
That is a massive drop 70k or so ppd
My cards will be working the same hours, my electric will be costing the same, but I will be 70k down everyday.
This is what will be the final credit will be or will a new credit scheme come into force when the final move over has occurred ?
|
|
|
 rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
                 
|
|
CPUs and GPUs are much faster than in the past and a credit-recalibration is normal in all projects of the boinc community.
You have only two choices:
1st: live with that like others too
2nd: upgrade your hardware
A 3rd would be useless because all projects will somewhen recalibrate their credits.
____________
Best wishes. Knowledge is power. by jjwhalen
 |
|
|
|
|
|
I run Primegrid because of the credit i get - & when that credit is decreased i will reevaluate if i want to continue crunching for primegrid or can get better credit somewhere else.
____________
 |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
                         
|
CPUs and GPUs are much faster than in the past and a credit-recalibration is normal in all projects of the boinc community.
You have only two choices:
1st: live with that like others too
2nd: upgrade your hardware
A 3rd would be useless because all projects will somewhen recalibrate their credits.
Why bother to Upgrade if their just going to cut the Credits. You spend your Money to get the latest Hardware so you can run the Wu's Faster for the Project just to get less Credits ... Duh Looooosing ... The only one that comes out ahead in this is the Project ...
____________
 |
|
|
|
|
On current credit rating I average 72x4523 around 325k/330k ppd
The new lowest credit rating 72x3607 so around 255k/265k ppd
That is a massive drop 70k or so ppd
My cards will be working the same hours, my electric will be costing the same, but I will be 70k down everyday.
This is what will be the final credit will be or will a new credit scheme come into force when the final move over has occurred ?
I'm afraid you didn't read the 1st post close enough. Over the next 12 weeks the credits will be CUT IN HALF !
____________
|
|
|
|
|
The only one that comes out ahead in this is the Project ...
Actually Steve I don't think the project ultimately wins in this case. 14 months ago ATIs were the way to go if you wanted the bigger credits. With the advent of the PG app coming online there has been a tremendous shift from ATI to Nvidia. Following one of the large team's forums it is quite obvious that almost no one buys a new ATI card.
So by cutting the credits in half it makes the ATIs look more attractive because a 6970 will produce the same credits (or more) on Moo and additionally there are MW, Collatz, and DNETC that will still pay well.
So the bottom line, less reason to buy a Cuda means less new cudas for the project. Why should I spend $350 for a cuda card that will pay well only on 1 project when I can get an ATI that will run more projects that pay very well. Oh BTW, the ATI will run this one too if I get desperate. |
|
|
|
|
|
Personally I believe the squeeze is being employed on Primegrid to
to pressure a devaluation in credits.
What squeeze?
Well at the current rate today (7-2), it would take 23 days to surpass
No. 2 Milkway@Home then onto the granddaddy of Seti@Home'
Even at reduced rates Seti will fall around Thanksgiving day
Seti, your days at No.1 are coming to an end.
Primegrid is going to have Seti turkey for dinner.
Maybe, maybe not!
At least it's one explanation to save Seti from a Labor Day dinner..
____________
 |
|
|
|
|
|
50% reduction in credits.... not a good thing. I would take careful consideration of the historical markers of when, how, and why you had a surge of volunteers flooding to this project and where they came from.
Without the incentive the tide will turn and they will go back from whence they came. I would hopefully assume that the work was useful, desired and you would like it to continue. This is not going to have a positive impact on this project
____________
 |
|
|
|
|
|
On the positive side, I should be able to get a cheap nvidia card when everyone switches back to ATI apps. |
|
|
|
|
|
If I read this right, we could come close to parity in running cost per credit between PG and GPUGRID on CUDA and on the other hand MW/Collatz on ATI. Probably with some advantages for Moo on ATI. That certainly would be interesting - except for all those who invested big.
I am however glad I did not yet get another 570. And to be honest, I don't think Primegrid will be a winner in this contest. |
|
|
|
|
|
I was thinking of buying another GTX 580, guess I'll find a 5970 to go with my two 5870s now.
Also looks like with a 50% reduction, I'll be putting my GTX 580 on other tasks. It's been running 24x7 since I joined Prime Grid.
Admin got a clue yet? Others will think the same, it's will result in people leaving PG. You really should rethink this "strategy". |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Mr. Hankey wrote: I would hopefully assume that the work was useful, desired and you would like it to continue. This is not going to have a positive impact on this project
Yes, the work is extraordinarily useful and productive. Yes, it remains desired. However, credit was never based on desire. It followed from the optimization of tpsieve vs. sr2sieve, the latter of which will no longer be used in the Proth Prime Search. In the next phase (3M-6M), the sr2sieve application is not used.
Michael Herman wrote: Personally I believe the squeeze is being employed on Primegrid to to pressure a devaluation in credits.
An entertaining theory. ;) However, there were rumblings soon after the optimized tpsieve application was released. At that time, sr2sieve was still being used and represented a large part of the sieve so credit justification was valid. Since that time, the sr2sieve application has been discontinued for the Proth Prime Search and now represents just a small fraction of the overall sieve. In fact, in the next phase, it's not used at all.
The credit justification no longer exists based on the sr2sieve application. The natural progression is to adopt tpsieve as the base application.
Dan wrote: Admin got a clue yet? Others will think the same, it's will result in people leaving PG. You really should rethink this "strategy".
This is not a "strategy". Simply put, we will no longer be using sr2sieve in the Proth Prime Search for future sieves. i.e. next phase of 3M-6M.
____________
|
|
|
|
|
CPUs and GPUs are much faster than in the past and a credit-recalibration is normal in all projects of the boinc community.
You have only two choices:
1st: live with that like others too
2nd: upgrade your hardware
A 3rd would be useless because all projects will somewhen recalibrate their credits.
This "1st: live with that like others too" my OP touched on the point, admins/project heads loose touch with their VOLUNTEER base, whether it be ego/politics and totally disregard the contribution they make, it beggars belief that a now 50% reduction would somehow be acceptable, all it will do is alienate. Like quite a few others, I will go for the 4th, off. |
|
|
|
|
|
One of the reasons, if not the main one, I chose to participate in PG was the fact it had the option to use my ATI card, the same card I got in my new system.
No, I don't have an i3, i5, i7, or the Gulftown, rather, I have a meek Core2 Quad Q8400 system. I'm not a gamer, but I do like to do things fast when I need to, and I like the simple multitasking it offers. Seldom did I ever have 4 programs open and running at the same time. I'll save the 'waxing poetic' for a later time....
So now I'm seeing that the one subproject that utilizes my ATI HD4800 series GPU is going to cut credits by something close to 50%, after it's all said and done. My first thought was "Why?"
I looked over all my subprojects again to make sure that PPS Sieve was the only GPU I could run, and regrettably, it is.
My creds are going to 'eventually' be decreased by nearly 50%, all in the name of an app - a single application that, when discontinued, may lead to volunteers leaving the project.
In this thread, there were "three" options mentioned, but only "two" were explained: 1 - Live with it and like it; 2 - Upgrade hardware; 3 - Usless because all projects recalibrate their creds. Well, the third option is to walk away from the project.
Well, I'm going to decide if Option 1 is worth the choice or not. Option 2 is out of the question - Why should I spend more money to upgrade to better hardware when it's going to cost me more money in electricity for usage by the rig and the AC unit to keep my room nice and cool. So I guess the overall answer to Option 1 will determine if Option 3 is viable.
It seems that PG wanted to go away from the "sr2sieve" to the "tpsieve," by way of the "ppsieve." Two things crossed my mind when I scoured the PG subprojects to ensure PPS Sieve was the ONLY ATI availability: 1 - If PG is so interested in making sure their applications are up to date, why is my 64-bit machine still operating on 32-bit programs AND why hasn't PG taken advantage of multi-core machines; and, I'll admit to being the class joker now, but did PG realize that they gave lots of fodder to those like myself when they named their "replacement apps?" The "sr2sieve" is an ok name, but "pp"sieve and "tp"sieve????
Please tell me that I wasn't the only one who chuckled at the "potty" humor in the names?
In any event, I'll climb down of my soapbox, and I'll begin to mull this over in my mind.
____________
 |
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
@Steve*, vandiesel, Rainmaker65
Ups...
I wasn't aware, that John made an edit of his first post after ~10hrs because i changed the sort order to "newest first".
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
"Rainmaker65" wrote: It seems that PG wanted to go away from the "sr2sieve" to the "tpsieve," by way of the "ppsieve." Two things crossed my mind when I scoured the PG subprojects to ensure PPS Sieve was the ONLY ATI availability: 1 - If PG is so interested in making sure their applications are up to date, why is my 64-bit machine still operating on 32-bit programs AND why hasn't PG taken advantage of multi-core machines; and, I'll admit to being the class joker now, but did PG realize that they gave lots of fodder to those like myself when they named their "replacement apps?" The "sr2sieve" is an ok name, but "pp"sieve and "tp"sieve????
sr2sieve is the original app-name of geoffrey reynolds sieving-app on cpu. PPsieve (Proth Prime sieve) was written by Ken Brazzer with help from geoffrey and is a port of the sr2sieve-app to Cuda. In a second step the PPsieve-app should be ready to calculate PPSE (20070000475957 | 4995*2^1822738+1) and Riesel-Sieve (20070000541441 | 3243*2^1584966-1) with only one app and TPsieve (Twin Prime sieve) was born.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
If the credit will decrease, then I think about going into a different project.
I pay the energy for the same tasks, but 50 % less credits.
example
if you buy something in a shop and every 5 minutes, the price changes upward, then you go back to another shop. This is normal! I`m not ill.
____________
|
|
|
|
|
|
I have two weaknesses in what I am about to say on this, first I'm still a "new kid on the block" having crunched limited different PG types in the three short months I've been here; second my 2x5970s are usually parked at Moo!Wrapper, and shift to PPS Sieve only to support Team efforts.
For the Project to continue down the line of Sieve History and actual useage of applications to assuage what must have been anticipated as a Right Royal Mini-Riot in the making, is pretty close to the old saying "when in a hole stop digging".
I cant believe for a moment that the mass exodus that is about to occur if processing times of GPU WUs dont also match (to a substantial degree if not totally) the credit reduction, was not anticipated or realised. I can only conclude that therefore the reduction in the Sieve effort is in fact benificial - maybe in terms of preventing the LLR getting way too long in crunch time, I dont know, but there has to be a substantial reason for not at least moving part way in setting the tpsieve credit level instead of slashing by 50%. After all the PPS Sieve app was way way over the top in the first place, on credit levels, when set against the "standard" BOINC reference model - accepting the latter's distinct creeky foundation and relevance, it is nontheless at least a pointer to BOINC apps.
Maybe there is a "Culture Clash" re traditional PG CPU based activity and GPU crunching, bit wacky to say the least, I have no idea - I am still rattling the brain for a sensible justification in all this. I conceed it does not materially affect me being only an occasional PG GPU cruncher, but it does affect many on my Team. No doubt many others, who made in the past - or have just made, substantial investment in NVidia cards - some $hundreds, some $Thousands. To sweep the feet out from under them is - to put it mildly - somewhat dubious in terms of appreciation for what they have undoubtedly - irrefutably - achieved for PG.
I have been - genuinely - impressed with PG in my three short months. The Overwhelming proportion of Admins/Server Admins do a great job from where I sit, and its a nice calm orderly Project, real nice to see set against some BOINC malestroms that are around. The latter makes it even more surprising - astonishing frankly - its being handled this way. A rethink is needed to minimise damage IMHO.
I'll get back in my box - probably should have stayed there as a PG newb - but this really rankled, and I had to get it of my cheast, warts 'n all :)
Regards
Zy |
|
|
|
|
|
I do not have alot of hardware inventment in any of the DC projects, but it would appear that since the output of my single card (an ATI v5800), which is already lower than it gaming counterparts, is realy not all that usefull anymore. I will likly be ending my contribution after my current buffer of work units is complete. The risk to my hardware that I use for professional applications is not worth its output. |
|
|
|
|
|
If all you are worried about are credits, then this sucks for you, but oh well, life moves on, and so will the project, the project was doing just fine with its efforts before the GPU people came and overtook everyone in the rankings (Note that even with my laptop crunching full time, even with its GPU crunching, my ranking only went down).
____________
 |
|
|
mackerel Volunteer tester
Send message
Joined: 2 Oct 08
Posts: 2959
ID: 29980
Credit: 781,069,441
RAC: 114,280
                         
|
If PG is so interested in making sure their applications are up to date, why is my 64-bit machine still operating on 32-bit programs AND why hasn't PG taken advantage of multi-core machines;
The LLR apps are 32 bit as there is no performance benefit from going 64 bit on them. If you're getting 32 bit CPU sieves on a 64 bit OS that could be a real problem and you might want a separate thread for help.
For taking advantage of multi-core, that's what BOINC does for you. You have n cores, you have n tasks running. For the type of work done at PrimeGrid that is the most efficient way. One task running over multiple cores is generally less efficient in overall throughput, and might only be chosen for other optimisation goals such as rapid completion time. |
|
|
|
|
|
Oh...just bought two 570s for PG......damn..... |
|
|
|
|
|
I understand that 98% of the cruncher are working in PPS-sieve, and other tasks are to remain at PG, because nobody likes to work with the CPU.
Better it would be the remaining 2% Credite higher, and not 98% Credits lower.
Thats what I mean.
Armstrong
____________
|
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
If all you are worried about are credits, then this sucks for you, but oh well, life moves on, and so will the project, the project was doing just fine with its efforts before the GPU people came and overtook everyone in the rankings (Note that even with my laptop crunching full time, even with its GPU crunching, my ranking only went down).
I'm sure it was doing fine before the GPU People came in and raised the Credit Bar but how far has the Project Advanced because we came in ??? and now that they have no use for us anymore they can afford to cut the legs out from under us, that's what sucks.
The Credits still won't be bad here but nothing like they were, in my heyday at DNET I was getting 8-9 Million a day, even at 1/2 Credits here I could probably still get 8 Mill + or - a few hundred thousand so all's not lost. It'll give me time to sell off a few NVIDIA's and get more ATI's to Balance the load ... ;)
____________
|
|
|
HAmsty Volunteer tester
Send message
Joined: 26 Dec 08
Posts: 132
ID: 33421
Credit: 12,510,712
RAC: 0
      
|
|
It is funny how people are arguing about getting lesser pointless credits. the work is the same, so whats wrong with you guys?
____________
 |
|
|
|
|
|
Points are the only one thing we have from Boinc.
I understand recalibration process, it's inevitable. You guys didn't thought that high credit level will last forever, did you? :P |
|
|
|
|
|
Dan wrote: Admin got a clue yet? Others will think the same, it's will result in people leaving PG. You really should rethink this "strategy".
This is not a "strategy". Simply put, we will no longer be using sr2sieve in the Proth Prime Search for future sieves. i.e. next phase of 3M-6M.[/quote]
So you don't get it! You can talk about sieves and next phase all you want. It's pretty meaningless. Our GPUs are here for the credits. Credits drop we leave. Still wondering if you get that point? |
|
|
Vato Volunteer tester
Send message
Joined: 2 Feb 08
Posts: 982
ID: 18447
Credit: 2,340,591,191
RAC: 819,083
                           
|
|
The credits issued were higher than they would have been due to the benchmark being 32bit sr2sieve (which was done to be fair in the first place). The reduction is getting rid of the artificially high score. So, we've been getting a bonus all this time, and now it's going to reduce. If that makes anyone so unhappy that they leave, that's sad, but is their choice. Personally, I'm here for the project purposes, rather than credit.
____________
 |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Steve* wrote: ...and now that they have no use for us anymore they can afford to cut the legs out from under us, ...
This is not true. The need to sieve in the Proth Prime Search will remain the same under tpsieve as the standard application as it did when sr2sieve was the standard application.
It is true that the Proth Prime Search has made significant advances since the introduction of tpsieve. And yes, it is true that later porting it to the GPU made use of its incredible processing power. We continue to be humbled by the advances Ken was able to accomplish.
Note that the Proth Prime Search is only 1 of 17 active prime searches (BOINC and PRPNet included) at PrimeGrid.
Dan wrote: So you don't get it! You can talk about sieves and next phase all you want. It's pretty meaningless. Our GPUs are here for the credits. Credits drop we leave. Still wondering if you get that point?
Yes, I understand: credit drop = GPU exodus. However, switching from sr2sieve to tpsieve is not meaningless...it's quite significant.
____________
|
|
|
|
|
|
Hey................................
Why this war ??
The credit is based on CPU work ! Not GPU.
Some of you say you get more credit on other proect, but i do not see any example on that.
So is it true that we give a very low credit ?
I know our GPU app is much faster on CUDA but I am not the man to make the ATI/AMD faster!
And as you say you do work here because of the high credit ??
Is it not good if we try to keep it inline with other project ?
Lennart |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
Yes, I understand: credit drop = GPU exodus. However, switching from sr2sieve to tpsieve is not meaningless...it's quite significant.
Even @ 1/2 Credit my NVIDIA GPU's will mostly stay here, it'll still be the best Paying NVIDIA Project and Pay as good if not better than any ATI GPU Project.
Once participants see/understand that I don't see that big an exodus, a few may leave because their not getting what they were getting but your always going to have that with a Credit Drop ...
____________
|
|
|
|
|
|
And BTW, even at 50%, PG is still No.1 PPD for NV cards. And it will make the competition of ATI/NV more interesting...
GTX 570 Primegrid 650k PPD
Primegrid(later) 325k PPD
DNETC/Moo 190k PPD
GPUGrid 180k PPD
Collatz 140k PPD
5870 Primegrid 200k PPD
Primegrid(later) 100k PPD
DNETC/Moo 300k PPD
Collatz 200k PPD
Milkyway 200k PPD |
|
|
|
|
It is funny how people are arguing about getting lesser pointless credits. the work is the same, so whats wrong with you guys?
If it is pointless as you say then why not just stop giving any credit? Why are there teams/leagues/challenges/mini challenges/multitude of stats site/active recruitment to teams. The points are there for a very good reason, if I was not interested in BOINC points, I would just keep a couple of rigs on wcg/hcc, where it takes my 24/7 4xAMDx6@4ghz/1xAMDx4@3.5, 2 other rigs tri/dual core part time, that is 33 cores which give a daily ppd of between 15k/17k, a single 4850 on mw/colllatz/dnetc will give you around the 70k/90k ppd. Personally I am in it for both the points and the science |
|
|
|
|
|
I am one not leaving PG. I've heard all these arguments before. When BOINC first started and was just on a few projects, people argued how credits were always changing.
They change for a reason. The work that we are doing is significantly different than the old work. The credits are to match that new work. Please look at it from the projects point of view. They are fixing things that should have been fixed when the changes were going into place instead of a little later.
I know some people may leave, but as some people pointed out it will still give probably the best NVidia credit than any other project out there.
It's also just sieving. Even though important to the final product, it's not finding primes, which is what this project really is here for. Sieving will eventually get finished when it hits the optimal zone, so at that there will be no credit. Then where do you go?
This project is one of the most stable, has a good relevance, and some very awesome project administrators. They listen to their crunchers, but they also need to make sure they follow the formulas they need to follow.
I wish to thank all the PG Admins, and volunteer programmers on this project for all their hard work and hope they have a very nice summer period. |
|
|
|
|
If all you are worried about are credits, then this sucks for you, but oh well, life moves on, and so will the project, the project was doing just fine with its efforts before the GPU people came and overtook everyone in the rankings (Note that even with my laptop crunching full time, even with its GPU crunching, my ranking only went down).
My reason for moving to another CPU only project. I might come back now :)
____________
 
Join team BOINCstats |
|
|
|
|
|
I am sorry, but i am really just annoyed of the fact, that the credit reduction and the according information came nearly at the same time (Maybe even in the wrong order). This is Ghaddafi-Dictator-Style. If you want to play fair, keep the people informed about your tactical moves. If you are planning changes, why can't anyone of the administration post it one or two weeks before the change happens?
This medal like every other has TWO sides, you (the maintainer or the project) and us (the people who invest money, time, electricity).
We would like to be treated fair. In sportive way we seek the competition.
but imagine two people with 100%exactly gpu/cpu-runtime, 100%exactly hardware but 2 different starting-dates:
szenario a) they have the same credit
szenario b) they have different credit
So vote a) or b) which you think is fair? |
|
|
|
|
|
Not only has credit decreased to 4327/wu but my runtimes have increased from 18.25 min to 40.25 min, which means daily credit will drop from 356K to 154K, more than 50% and will drop further in the next few weeks.
More importantly, with the cost of electricity due to increase by 10% in the UK, apart from the fact that the credits have dropped, the number of completed WUs per day has dropped by more than 50% and therefore the electricity cost per WU has more than doubled.
____________
 |
|
|
|
|
And BTW, even at 50%, PG is still No.1 PPD for NV cards. And it will make the competition of ATI/NV more interesting...
GTX 570 Primegrid 650k PPD
Primegrid(later) 325k PPD
DNETC/Moo 190k PPD
GPUGrid 180k PPD
Collatz 140k PPD
5870 Primegrid 200k PPD
Primegrid(later) 100k PPD
DNETC/Moo 300k PPD
Collatz 200k PPD
Milkyway 200k PPD
I run a GTX 580 on PG and get about 650K. A 5970, which is roughly on par with a 580 gets about 600K on Moo or DNET and 300K on MW or Collatz. I'm giving my Nvidia to my son and going all ATI. Much better points, plus MW is real science. |
|
|
|
|
Not only has credit decreased to 4327/wu but my runtimes have increased from 18.25 min to 40.25 min, which means daily credit will drop from 356K to 154K, more than 50% and will drop further in the next few weeks.
More importantly, with the cost of electricity due to increase by 10% in the UK, apart from the fact that the credits have dropped, the number of completed WUs per day has dropped by more than 50% and therefore the electricity cost per WU has more than doubled.
No changes is done to the work. If there are more user that can confirm this please post it or PM me.
Lennart
|
|
|
|
|
Not only has credit decreased to 4327/wu but my runtimes have increased from 18.25 min to 40.25 min, which means daily credit will drop from 356K to 154K, more than 50% and will drop further in the next few weeks.
Possibly the NVIDIA downclock problem but we need more information... driver version, clock rates, etc.
____________
|
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
I haven't noticed any differences in the run times ...
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
If you are planning changes, why can't anyone of the administration post it one or two weeks before the change happens?
I completely concede that this was a gross error in communication to the users created by miscommunication between the admins. My apologies. However, this process is expected to take 12-16 weeks to unfold. Hopefully that's adequate time for everyone to adjust their computing interests.
____________
|
|
|
|
|
I'm giving my Nvidia to my son and going all ATI. Much better points, plus MW is real science.
Real science? You made my day! What do they examine? Downtimes?
____________
|
|
|
|
|
[quote]No changes is done to the work. If there are more user that can confirm this please post it or PM me.
Lennart
Theres no downclock problem here, not using the problem driver, it was fixed in later versions.
____________
|
|
|
|
|
|
I participate in PG because it performs important work in the field of primes and because I enjoy the badges. Therefore, for me nothing has changed. The importance and fun are still there.
____________
Thanks,
Jim
 |
|
|
|
|
Theres no downclock problem here, not using the problem driver, it was fixed in later versions.
The downclocking still happens here with the 275.33 drivers installed. Did you install 275.50?
____________
|
|
|
|
|
|
Still using 275.33 with no problems but I have rebooted and the runtimes have gone back to normal. There was no sign of any downclock before I rebooted but I'll give driver 275.50 a try
____________
|
|
|
|
|
I'm giving my Nvidia to my son and going all ATI. Much better points, plus MW is real science.
Real science? You made my day! What do they examine? Downtimes?
My science said:
I go ! - perhaps to Collatz / Dnetc
I am one, I know, but how many goes ?
In my Country, we pay 15 % more for Energie, and then 50 % lower Credits ? No ! Tschau tschau !
____________
|
|
|
Scott Brown Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2540
ID: 1178
Credit: 29,799,904,808
RAC: 11,912,495
                                                            
|
|
Wow...looks like many people need to stop and take a deep breath. John's first post has been updated (with a more detailed update still to come) to note the transition to calibration to a different sieve application. The new application for the PPS sieve will then award credit comparable to the other GPU app here at PG, the GCW sieve. I take that to mean that the new PPS sieve application (and new range being examined) transitions things such that work done is now more comparable with the level done in the GCW sieve. I am sure that John's forthcoming detailed explanation will provide a more satisfactory description of how this works.
Let's also nip a couple of things in the bud here before the "credit wars" and other nonsense from other areas of BOINC enter and ruin the overall pleasurable experience at PG:
1. GPU apps get huge credit because they do huge work in certain applications. The couple of sieves here are great examples of how much work can really be done (especially with single-precision apps). GPU crunchers are just as legitimate as any other cruncher...they should not be demeaned nor should they feel superior.
2. Sieves are applications doomed to end (or at least suspension) upon optimization at which point there will be no credit at all. Hopefully, additional sieves will become available for the GPU by then or the LLR GPU apps will be in force (though the latter are much less efficient than the sieve apps on GPU and require double-precision...i.e., only newer GPUs and much lower throughput/credit compared to sieves).
3. Credits are not meaningless...anyone who states differently is simply ignorant (*please note that I did not say stupid...ignorant, as in not informed*). A full century or scientific work in the fields of anthropology, sociology, psychology, political science, and economics (among other fields) has provided and continues to provide great detail on why and how humans volunteer and participate in supposedly altruistic behaviors. Simply put, credits seem to appear to function similarly to arbitrary donation level designations common in many charitable organizations (i.e., give more money to be a "sponsor" rather than a "friend" or a "benefactor" vs. a "sponsor", etc.). For more detail, go to your local library and do some reading.
4. All projects change...period. If you buy any equipment with the expectation that change will not happen, than you are being a bit naive. An excellent example of this sort of thing is found in the GPUGrid project. Once they used the PS2 only (Indeed, that used to be the project name), and then transitioned to GPUs over time. They further adjusted their applications such that many GPUs were no longer useful at that project. Such is the nature of distributed computing projects (and in many cases, projects have disappeared entirely). We should be thankful that PG is particularly good at showing its participants respect by providing reasonable transition periods (e.g., when subprojects end, there is a time of essentially useless computation allowed such that users can complete a badge level, etc.).
5. John has already apologized more than once for the communication error with this transition...let's all give that part a rest.
6. PG is indeed not "real science"...it is actually real math! If you are looking for a science project, then you are in the wrong place (though, of course, math is essential to most scientific endeavors...both types of projects are quite important).
PG has always been one of the (if not THE) most enjoyable projects in which to participate. Credits are great, but someone will always come along at some point who will pass you in the rankings (I know quite well...I was 7th at PG before the PPS sieve GPU app...and that does not bother me at all). If you want to make a more permanent mark, then crunch some of the LLR to find a big prime, etc.
____________
141941*2^4299438-1 is prime!
 |
|
|
|
|
6. PG is indeed not "real science"...it is actually real math! If you are looking for a science project, then you are in the wrong place (though, of course, math is essential to most scientific endeavors...both types of projects are quite important).
Your right it is real math, I guess I meant to say relevant math. It's kinda like a spelling bee is to a language. A nice oddity.
But what do I know, I'll ask my daughter and her boyfriend next week how much they use prime numbers in their PhD work. Of course here I go again assuming Physics and Astronomy use math. |
|
|
|
|
|
I have noticed an increase in run times on my GTX 295s they went from about 18 - 20 min to 40 - 45 min. I am running 275.33
Not only has credit decreased to 4327/wu but my runtimes have increased from 18.25 min to 40.25 min, which means daily credit will drop from 356K to 154K, more than 50% and will drop further in the next few weeks.
More importantly, with the cost of electricity due to increase by 10% in the UK, apart from the fact that the credits have dropped, the number of completed WUs per day has dropped by more than 50% and therefore the electricity cost per WU has more than doubled.
____________
|
|
|
|
|
I have noticed an increase in run times on my GTX 295s they went from about 18 - 20 min to 40 - 45 min. I am running 275.33
Not only has credit decreased to 4327/wu but my runtimes have increased from 18.25 min to 40.25 min, which means daily credit will drop from 356K to 154K, more than 50% and will drop further in the next few weeks.
More importantly, with the cost of electricity due to increase by 10% in the UK, apart from the fact that the credits have dropped, the number of completed WUs per day has dropped by more than 50% and therefore the electricity cost per WU has more than doubled.
Number of tasks today: 77
Consective valid tasks: 0
I suggest a reboot...
Update:
Consective valid tasks: 0 doesn't seem to indicate an error. You might check the clock rate of your GPU. The reboot suggestion still applies ;)
____________
|
|
|
|
|
|
Thank you Pooh Bear and Scott Brown; I've been trying to figure out how to politely voice my feelings on this issue - you folks did it beautifully.
I will NOT be leaving PG. I hope no one does.
I've enjoyed watching you GPU folk chew through work almost as fast as the admins can load it into the queues.
I've enjoyed being able to help some of you get things going and I've appreciated your help when I've gone adrift w/o a clue as to how to fix my mess.
My reason for coming here was to find TOP 5000 primes (the bigger the better).
I found a home - a project without the usual politics, a project where Admins and cunchers actually took time to help me get up and running, a project that actually stuck around, had a constant flow of work, with proven apps and servers that stayed up, a project that gave ME credit for MY discoveries.
I came this project before the CUDA app was released - I was one of many who helped Ken to test his gpu based app (all without receiving any BOINC credit).
I admit I sure have enjoyed the credit I've earned since the app went live.
I like credits just like everyone else - I wish I had more - once upon a time I was ranked 38th (now I'm 336th) - I admit to having really bad cases of high-end GPU and big pharm envy.
I just have a single system with a single GTX275.
I also wish to thank all the PG Admins, volunteer programmers and testers as well as all the CPU and GPU based crunchers on this project for all your hard work, your patience, your help, and your support!
And remember its all relative - some of you folks have RACs that are 100 times greater than mine - even at 50% less credit per wu you'll still be earning 100 times more a day than I can.
If you do feel that you must leave - may your RAC and kWh be inversely proportional.
TheDawgz
____________
There's someone in our head but it's not us. |
|
|
|
|
|
Thank you Pooh Bear, Scott Brown and of course TheDawgz for your kind words.
____________
|
|
|
|
|
|
I know that credits are fun and they enable stats and races and all such.
But are they really the only reason to crunch for a specific project? Regardless of what the project is doing?
I did a lot of CPU work for Prime Grid and back in those days I was rather high up in the stats. But my main reasons for crunching here were that I liked the project's goals and I got extraordinary support from the admins to make best use of my rather unusual setup. That's also why I will return and do more work here when I finish crunching projects where I don't get any credits at all. They're fun and for me that's what it's all about.
I'll stress the point once more: the late RCN was probably the only project with an admin being as close to the users as Rytis, Lennart and John are here. Whichever problem or question I had, they took pains to support me. Which is why I will continue supporting their project, regardless of credits.
If you absolutely feel you have to switch projects or stop crunching for good because of the credit reduction, that's a pity. I enjoy watching open-mouthed as the GPU people plow through massive amounts of work in short time. But of course, in the end each has to decide for himself.
Take care,
Peter
____________
There are only 10 kinds of people - those who understand binary and those who don't
 |
|
|
|
|
|
perhaps CW sieve is better, after 12 weeks. If the projektadmin do the credit there a little bit higher.
____________
|
|
|
|
|
|
Primegrid is actualy only project with good linux CUDA app.
I will continue even with lower credits.
____________
 |
|
|
|
|
6. PG is indeed not "real science"...it is actually real math! If you are looking for a science project, then you are in the wrong place (though, of course, math is essential to most scientific endeavors...both types of projects are quite important).
Your right it is real math, I guess I meant to say relevant math. It's kinda like a spelling bee is to a language. A nice oddity.
But what do I know, I'll ask my daughter and her boyfriend next week how much they use prime numbers in their PhD work. Of course here I go again assuming Physics and Astronomy use math.
Try this...
prime_numbers_get_hitched
____________
Thanks,
Jim
|
|
|
|
|
|
I will continue crunching PPSsieve on my GPU(s) and other sub-projects (PrimeGrid ones) on the CPUs. But, I will be moving some of my GPUs to other projects when I get the dust bunnies firmly removed and the other 2 closed down servers back producing. |
|
|
|
|
|
I will continue to crunch, although there is an obvious need for John and the other Admins to coordinate announcements.
____________
 |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
I will continue to crunch, although there is an obvious need for John and the other Admins to coordinate announcements.
As a last reminder, We are in "summer hours".
The first snafu of "summer hours" is miscommunication. The credit reduction was a bit premature; however, a reduction is coming for PPS (Sieve). Please see this post for more details.
____________
|
|
|
|
|
|
I will stay on PG, the GPU credits did help me, but I enjoy the project. I may bounce my GPU around to other projects a little.
____________
 |
|
|
|
|
|
One more thought.
The teams at PG are competitive. By cutting the credits in half, you've effectively double the leads each team has on teams with less credit. Will you cut each teams points in half or do you favor giving other teams this windfall. |
|
|
|
|
|
I'm not going anywhere, I see the credits I received from PPS Sieve on my GPU as a bonus, thank you!
____________
35 x 2^3587843+1 is prime! |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
One more thought.
The teams at PG are competitive. By cutting the credits in half, you've effectively double the leads each team has on teams with less credit. Will you cut each teams points in half or do you favor giving other teams this windfall.
Your Kidding Right ???
The Credits each Team has/had were earned under what the Project was giving & every Team had the same chance to earn more or less Credit than what they have now.
If a 777 Member Team can't keep up with a 118 Member Team you shouldn't come crying for a Team Credit Reduction when the Credits are lowered ... IMO
____________
|
|
|
ich_eben* Volunteer tester
Send message
Joined: 14 Oct 10
Posts: 101
ID: 70063
Credit: 468,624,789
RAC: 79,438
          
|
One more thought.
The teams at PG are competitive. By cutting the credits in half, you've effectively double the leads each team has on teams with less credit. Will you cut each teams points in half or do you favor giving other teams this windfall.
Each team had the same credit per WU until now and they will always have the same credit per WU in the future.
I see your point that it gives leading teams more "buffer" but every team had the opportunity to crunch the project when it rewarded the higher credits.
____________
|
|
|
|
|
|
no change in run times.
602 seconds PPS Sieve.
GTX 570 Linux Mint 11 x86
____________
|
|
|
|
|
One more thought.
The teams at PG are competitive. By cutting the credits in half, you've effectively double the leads each team has on teams with less credit. Will you cut each teams points in half or do you favor giving other teams this windfall.
Besides, we are changing the credit scheme for just one project, right? No big deal. |
|
|
|
|
|
another 4 cents needed?
everyone is free to crunch what he wants and when he wants to - or simply not to do anything.. |
|
|
|
|
|
I must say, I am disappointed as well as most of you. I specifically purchased a number of NVIDIA cards at considerable expense so that I may get more credit at PG. It seems now I have to change to ATI and chase another project that does not half their credit rating. If I stay with PG, it will now take twice as long to catch up to somebody who is already ahead of me. Are we that stupid that we do not warrant a warning about this reduction in the open but hide the fact in the Message Boards? At the same time run a campaign of canvassing for donations? I say again. Are we really that stupid? |
|
|
|
|
|
I just spent £550 on a GTX 590 to crunch this project.
So I should be particularily irked at this change in the credits. Or should I?
Oddly I'm not.
Accept it dudes- the Admins giveth & the Admins taketh away: 'Twas ever thus and so it shall, and should, remain.
Whatever happens my 590 is going to out perform the 2 x 260s it replaces several-fold. If the PG returns sink low enough then that will free me up to explore a return to -say- GPUGrid safe in the knowledge that this will not leave my rivals free to exploit my easy-going promiscuity.
PG had got a little overgenerous and this seems to be a laudible attempt to check the credit inflation that has afflicted Boinc to the extent that I can now antipate that my output from now until Xmas might double what has taken me 7 years on a now slimmed down farm to build up.
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
                      
|
|
EDIT: Nah, too easy, like shooting fish in a barrel. Move along, this is not the sarcasm you're looking for.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Anyone who plans to buy NVIDIA cards for the sole purpose of accumulating tons of credits at this single project should consider this:
The lifetime of the GCW sieve subproject is limited. The lifetime of the PPS sieve is limited, at least for every exponent range that is tackled by the project.
The project and it's admins may decide to (or not to) continue the PPS sieve work for higher exponent ranges or suspend the (sub) project for extended periods of time to give the LLR people time to catch up with the sieving wavefront.
While spending tons of credits, eh bucks, for new hardware, clearly shows your dedication and willingness to help the science, your team or your stats, it maybe a wise move to inform yourself beforehand about the goals and various schedules of the relevant sub projects.
____________
|
|
|
|
|
The lifetime of the GCW sieve subproject is limited. The lifetime of the PPS sieve is limited, at least for every exponent range that is tackled by the project.
How limited is time for GCW sieve?
____________
|
|
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2703
ID: 1
Credit: 279,408,510
RAC: 373,669
         
|
|
See the forum thread for information about project life expectancy.
____________
|
|
|
|
|
another 4 cents needed?
everyone is free to crunch what he wants and when he wants to - or simply not to do anything..
+1. |
|
|
|
|
|
Another litttle "Fact" is.
Now my ATI's get about the same ppd as on Collatz.
If thats cut in half.
1 : Lots of Nvidia will move to other projects.
2 : ALL Ati users will move to other projects.
So i will have to find another project.. prolly back to Collatz/MW
Not going to put my Quad GTX570 and sr2 on this anymore after the cut its useless.
Ah well good luck with the CPU crunchers that you will have left after the credit cut.
Like moost the rest i dont mutch give a hoot what i crunsh for.
Making nuclair bombs or saving baby's.
It the points that count.
In the end its actualy all useless anyways.
____________
 |
|
|
|
|
|
Everyone with nvidia cards who wants points will stay in PG because it pays the most of all.
Most of ati users will move to dnetc, cc, mw, moo.
I don't think that the drop in power will be enormous. Besides, all sieves will end pretty fast in compare to llr. |
|
|
|
|
[quote]I must say, I am disappointed as well as most of you. I specifically purchased a number of NVIDIA cards at considerable expense so that I may get more credit at PG. It seems now I have to change to ATI and chase another project that does not half their credit rating.
If I stay with PG, it will now take twice as long to catch up to somebody who is already ahead of me.
I am surprised you take that view?
It will take you the same amount of time to catch up as the other crunchers will be cut back exactly the same amount of credit given as you will be given.
The gap between you will, I agree, need more given units crunched to close. But, in a competition between the two, or more, of you they will pull ahead at half the speed they were, and you will catch up in proportion.
So, clearly there is no change, and no need to take the path you have. All are affected equally, and not you alone.
Are we that stupid that we do not warrant a warning about this reduction in the open but hide the fact in the Message Boards?
I have to admit some people had to give the Admins a prod to find out what was happening, and that immediate reduction was reversed. We are now in the 200 given credit reduction each week for the next 12-16 weeks, as explained.
So, you have been advised on the reduction, how it is being done and over what time frame.
Regarding your comment about thinking the Admins think you are stupid .... from this post, and your reaction, the answer has to be ... YES. I certainly think so.
At the same time run a campaign of canvassing for donations? I say again. Are we really that stupid?
My apologies if this offends, but you did walk in to this reply and clearly did not think it through. |
|
|
|
|
|
Current RAC ~5.4m. Time to 1b at current RAC ~100 days.
Future RAC ~2.8m. Time to 1b at future RAC ~ 200 days.
Not going make it by new year but eventually will.
Move to non-PG project, future RAC <2m/day. Time to 1b even longer. Staying where I am.
My non-GPU RAC ~125k so LLR Ruby Badges every 16 days and Ruby Sieve Badges every 32 days. No change there. I will still get my full set by new year.
All in all the admins are equitably balancing things and I remain happily committed to the credits, the badges and the finding of primes.
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
1 : Lots of Nvidia will move to other projects.
2 : ALL Ati users will move to other projects.
That assumes that all participants only care about credits. While I think it's likely that most do care at least a little bit about credit, I think more than a few, and probably a majority, are primarily concerned with the science or math. Granted, even die hard math geeks like myself pay attention to the credits for various reasons, not the least of which is that the "competition" is fun. But I'd still do the crunching for little or no credit.
On the day I can walk into a deli and buy a pastrami on rye and pay with credits, I'll consider them worthy of being a factor in deciding what I am going to crunch. Until then, they're no more important to me than the score on a pinball machine or any other game.
____________
My lucky number is 75898524288+1 |
|
|
Sysadm@Nbg Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1289
ID: 18646
Credit: 1,126,600,531
RAC: 178,927
               
|
|
well my2cobbelstones:
I am not honored with any Cobblestone, while I spent my time here reading your comments :( In this time I should find a prime by llring in my head ...
I am crunching here to find some Primes or do something so any other user is able to find a prime much faster, nothing else matters!
so please dont cry, go prime searching or spent your GPU-/CPU-time/money to an other scientific project ...
____________
Sysadm@Nbg
my current lucky number: 456121176^131072 + 1
 |
|
|
mfbabb2 Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 21,825,457
RAC: 0
             
|
|
The loud rumbling in the background you hear is the sound of those of us with NO GPUs rolling on the floor laughing at this thread!
Credit is what it is! (And other than ego, quite worthless!)
GPUs get hundreds to thousands of time the amount of credit per hour than a CPU gets on the same task, so being upset about that being reduced is just plain silly.
(If all the GPUs leave -- and I doubt that will happen --, then that will just leave work for those of us with CPU only systems.)
Just my 2 credits worth.
____________
Murphy (AtP)
 |
|
|
|
|
The loud rumbling in the background you hear is the sound of those of us with NO GPUs rolling on the floor laughing at this thread!
Credit is what it is! (And other than ego, quite worthless!)
GPUs get hundreds to thousands of time the amount of credit per hour than a CPU gets on the same task, so being upset about that being reduced is just plain silly.
(If all the GPUs leave -- and I doubt that will happen --, then that will just leave work for those of us with CPU only systems.)
Just my 2 credits worth.
Well said! I'll throw in some hard earned CPU-only credits.
|
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1291
ID: 12001
Credit: 18,565,548
RAC: 0
|
Credit is what it is! (And other than ego, quite worthless!)
That gives me a thought. BOINC should allow people to sell credit via ebay and take a cut of the profit. I could imagine a few hundred farms sprouting earning credit then selling to the highest bidder.
As funny as that seems, it has happened with games thus leading to people selling virtual money for real money. |
|
|
mfbabb2 Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 21,825,457
RAC: 0
|
|
Let me know if you find a bidder -- mine are available for real cash! ;-)
____________
Murphy (AtP)
|
|
|
Vato Volunteer tester
Send message
Joined: 2 Feb 08
Posts: 982
ID: 18447
Credit: 2,340,591,191
RAC: 819,083
|
|
Look at the recent Bitcoin/MtGox fiasco for another example...
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
Credit is what it is! (And other than ego, quite worthless!)
That gives me a thought. BOINC should allow people to sell credit via ebay and take a cut of the profit. I could imagine a few hundred farms sprouting earning credit then selling to the highest bidder.
As funny as that seems, it has happened with games thus leading to people selling virtual money for real money.
LOL
That would be interesting -- that is, if you think hyper-inflation is interesting.
If a mechanism existed to transfer credits, that would permit you to sell them on eBay. Lots of people would start selling credits. Some would start buying them.
People would then start flocking to projects that offer the most credits. Project admins would start increasing the amount of credit disbursed, not just to attract new volunteers, but to keep the ones they have from going elsewhere. Other admins would be forced to do the same, which would make still other admins raise the credits even higher. This would go on until. quite rapidly I would assume, the number of credits being offered exceed what can be stored in a 64 bit signed number, and people's credit totals go negative.
Then everyone starts scratching their heads, thinks "WTF?!?", and start pointing fingers trying to figure out whose brilliant idea this all was. :)
(The difference between BOINC credits and game currencies is that game publishers have a vested interest in keeping the amount of currency present in the game to a minimum, whereas project admins would have a vested interest in pumping out as much credits as they can -- and there's no limit to how much credit they can create other than what can fit in a 64 bit number.)
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
"People would then start flocking to projects that offer the most credits"
They already do if i did not notice that :P
PrimeGrid 42,338 +26 132,869 +50 2,149 0 180 0 85,420,617,741 +441,359,788
Collatz Conjecture 20,845 +13 46,790 +50 1,106 0 152 0 68,223,040,520 +90,071,185
MilkyWay@home 102,325 +100 200,050 +198 2,701 +1 197 0 91,416,054,323 +108,905,579
Why u think primegrid has the most credits PPD of all boinc projects ?
____________
|
|
|
Lumiukko Volunteer tester
Send message
Joined: 7 Jul 08
Posts: 169
ID: 25183
Credit: 1,232,820,667
RAC: 1,284,957
        
|
...If a mechanism existed to transfer credits, ....
I think you can transfer credits. But not partially, you need to transfer all your credits in a project.
Go to your user account, change your password and email into some dummy ones.
Sell your credits and give the buyer those dummy email and password, so that buyer can then log in and change the username, password and email into his. Then he will soon get all those credits added into his account.
I have not tested, but it might work, or it could claim that the user account already exists with different ID.
Why not test it:
fastest one of you gets my Seti credits.
Goto Seti, log in with
email: anon@invalid.none
password: passwrd
change email, password and username into yours and you should get all my 230834 credits at Seti.
(Note: it will take some time before the CPID's are syncronized).
--
Lumiukko |
|
|
pschoefer Volunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 720
ID: 845
Credit: 3,968,155,877
RAC: 1,633,333
       
|
Why not test it:
fastest one of you gets my Seti credits.
Goto Seti, log in with
email: anon@invalid.none
password: passwrd
change email, password and username into yours and you should get all my 230834 credits at Seti.
(Note: it will take some time before the CPID's are syncronized).
It's not possible ("There's already an account with that email address").
____________
 |
|
|
|
|
you should get all my 230834 credits at Seti.
So do I hear a starting price of $.01?
Lumiukko, I don't really want them, nor do I think people should really start doing this, so I've pm'd you the new details. |
|
|
|
|
|
Sounds just like quantitative easing when they printed $billions and sent it in to the economy, to go no-one knows where.
Perhaps we can all donate all of our credits to the Greek economy and then no crisis! |
|
|
Lumiukko Volunteer tester
Send message
Joined: 7 Jul 08
Posts: 169
ID: 25183
Credit: 1,232,820,667
RAC: 1,284,957
|
It's not possible ("There's already an account with that email address").
So you can not have an existing account with same email in that project. You must first change your email in your current account (and sell/give/throwaway your current credits. Maybe someone just starting who does not have an account in that project yet would like to have them).
Meaning that you can't add credits to your existing credits in that project, but you can change to an account with higher credits, and so get higher total BOINC credits.
--
Lumiukko
|
|
|
|
|
This would go on until. quite rapidly I would assume, the number of credits being offered exceed what can be stored in a 64 bit signed number, and people's credit totals go negative.
...
and there's no limit to how much credit they can create other than what can fit in a 64 bit number.)
64 bit numbers for credits are no long term solution :D We should use the GNU Multiprecision Library (GMP) to stay on the safe side until the first QPUs (Quantum Processing Units) arrive :)
I can already hear the complaints of the GPU crunchers about the crunchers using that newfangled stuff: "That's no accomplishment!..." You can imagine how the rant will continue... ;)
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
This would go on until. quite rapidly I would assume, the number of credits being offered exceed what can be stored in a 64 bit signed number, and people's credit totals go negative.
...
and there's no limit to how much credit they can create other than what can fit in a 64 bit number.)
64 bit numbers for credits are no long term solution :D We should use the GNU Multiprecision Library (GMP) to stay on the safe side until the first QPUs (Quantum Processing Units) arrive :)
We'll have to start expressing credits in the form k*b^n+/-1.
____________
My lucky number is 75898524288+1 |
|
|
|
|
This would go on until. quite rapidly I would assume, the number of credits being offered exceed what can be stored in a 64 bit signed number, and people's credit totals go negative.
...
and there's no limit to how much credit they can create other than what can fit in a 64 bit number.)
64 bit numbers for credits are no long term solution :D We should use the GNU Multiprecision Library (GMP) to stay on the safe side until the first QPUs (Quantum Processing Units) arrive :)
We'll have to start expressing credits in the form k*b^n+/-1.
I suggest Fermat number based milestones: 2^2^n+1
____________
|
|
|
|
|
This would go on until. quite rapidly I would assume, the number of credits being offered exceed what can be stored in a 64 bit signed number, and people's credit totals go negative.
...
and there's no limit to how much credit they can create other than what can fit in a 64 bit number.)
64 bit numbers for credits are no long term solution :D We should use the GNU Multiprecision Library (GMP) to stay on the safe side until the first QPUs (Quantum Processing Units) arrive :)
We'll have to start expressing credits in the form k*b^n+/-1.
I suggest Fermat number based milestones: 2^2^n+1
But only if they are prime.
____________
|
|
|
|
|
|
My GTX 470 is crunching the first WUs in the 3M-6M range. Runtimes are up to around 16.5 minutes from 14.5 minutes per WU.
Update: Runtimes increased by ca. 110 seconds per WU. "Average factors per workunit" are rising slowly...
____________
|
|
|
|
|
|
Same here and i see it with more peeps like steve.
went from 580 seconds to 680 seconds..
Ah well thats i nice way to get the same credit it just takes longer..
But noone will notice it i guess :P
____________
|
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
Same here and i see it with more peeps like steve.
went from 580 seconds to 680 seconds..
That could come from your driver 275.33 and the still unsolved down-clock problem.
Try a reboot like others did and your times should be back to normal.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
Same here and i see it with more peeps like steve.
went from 580 seconds to 680 seconds..
Ah well thats i nice way to get the same credit it just takes longer..
But noone will notice it i guess :P
Some crunchers have already noticed the increased runtimes but they didn't draw the right conclusions. I guess they think this is another diabolic scheme to scare the GPU crunchers away :D
____________
|
|
|
|
|
Same here and i see it with more peeps like steve.
went from 580 seconds to 680 seconds..
That could come from your driver 275.33 and the still unsolved down-clock problem.
Try a reboot like others did and your times should be back to normal.
Not this time. The WUs are 3M-6M WUs with lower p ranges.
____________
|
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
Not this time. The WUs are 3M-6M WUs with lower p ranges.
You are right. I see now run times with 5.5ksec instead of the 4.8ksec from yesterday...
Is the problem now solved with 275.50 or not?
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
Not this time. The WUs are 3M-6M WUs with lower p ranges.
You are right. I see now run times with 5.5ksec instead of the 4.8ksec from yesterday...
Is the problem now solved with 275.50 or not?
I run 275.33 here on Windows 7 and it still occurs. I haven't tested 275.50. Jason Gee (one of the experts) over at the Seti@Home forum talked about a bug(?) in the boinc libs. I'm not sure if this is the sole reason because I've managed to reproduce this problem with other (non BOINC) CUDA applications.
____________
|
|
|
|
|
Not this time. The WUs are 3M-6M WUs with lower p ranges.
You are right. I see now run times with 5.5ksec instead of the 4.8ksec from yesterday...
Is the problem now solved with 275.50 or not?
Hi
run times have increased, for me my 580s have gone from 9.00mins to 10.30mins for lesss credit!!!!
So can someone tell me what is changed?
Cheers
Ross*
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
The initial 3M-6M range has been released and as expected, the run times at the lower P are slightly longer. This will not alter the original plan below. Adjustments will be made to the final number after the pause at 3607.
Credit will be decreased by ~200 cobblestones per week until those numbers are achieved. Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks.
Status for July
31 July - 3607
24 July - 3719
17 July - 3923
10 July - 4127
3 July - 4327
2 July - 4523
____________
|
|
|
|
|
The initial 3M-6M range has been released and as expected, the run times at the lower P are slightly longer.
slightly?
it went up from about 45ksec to 63ksec on my E8400.. :(
|
|
|
|
|
The initial 3M-6M range has been released and as expected, the run times at the lower P are slightly longer. This will not alter the original plan below. Adjustments will be made to the final number after the pause at 3607.
Credit will be decreased by ~200 cobblestones per week until those numbers are achieved. Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks.
Status for July
31 July - 3607
24 July - 3719
17 July - 3923
10 July - 4127
3 July - 4327
2 July - 4523
LLR tests on numbers in this range will take longer to complete.* How does this effect the optimial sieve depth for this range?
*I'm blatantly ignoring the possible availability of an improved AVX LLR version which might speed up LLR tests considerably - at least on Sandy Bridges, the future Ivy Bridges and (hopefully) the Bulldozers or their improved successors - when we start LLR tests on numbers in this range.
____________
|
|
|
|
|
|
My ATI cards now take 30 minutes to complete a task since a few hours.
That used to be 24 minutes.
So all my ATI card r going back to Collatz.
Have fun with the Nvidia's that still run "for now"
____________
|
|
|
|
|
|
Run times for PPS sieve on my humble cuda during the last day or so have been 10-15% higher than before.
Is this also part of the plan ? |
|
|
|
|
|
Read the first post in this tread.
"Credit will be decreased by ~200 cobblestones per week until those numbers are achieved. Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks."
Status for July
31 July - 3607
24 July - 3719
17 July - 3923
10 July - 4127
3 July - 4327
2 July - 4523[/quote]
Lennart |
|
|
|
|
|
Translation "We will make you work longer for half the points" :)
Everyone knows about the point reduction.
Not everyone knows it now takes longer to finish work units to.
None even told us (again) that my work units now take 6 minutes longer to complete.
____________
|
|
|
|
|
Translation "We will make you work longer for half the points" :)
Everyone knows about the point reduction.
Not everyone knows it now takes longer to finish work units to.
None even told us (again) that my work units now take 6 minutes longer to complete.
It is what it is. It may be a harsh response but either live with it or go.
____________
|
|
|
|
|
|
A recalibration to 3607.. yeah that is do-able (as stated :: recalibration).
But the second iteration to effectively go to 2251 points (less then 50% of the original 4523)..
is killing a whole subproject. No matter how you reason about it.. this is definitely not explainable.
Then to make matters WORSE: Increasing the time to process WU's for less then 50% of the points.. make no sense at all.
(Especially when you view it from a competitive viewpoint). |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
A recalibration to 3607.. yeah that is do-able (as stated :: recalibration).
But the second iteration to effectively go to 2251 points (less then 50% of the original 4523)..
is killing a whole subproject. No matter how you reason about it.. this is definitely not explainable.
Then to make matters WORSE: Increasing the time to process WU's for less then 50% of the points.. make no sense at all.
(Especially when you view it from a competitive viewpoint).
Have been thinking that myself but it is what it is, everybody will either have to live with it or move on to some other project.
____________
|
|
|
|
|
..
Have been thinking that myself but it is what it is, everybody will either have to live with it or move on to some other project.
Well I only recently joined PrimeGrid.. and got enthusiastic about it! Even so much that I donated some Euro's for the project!
If I had of known that the project-leaders/management (or however they are called) made such (un-explainable) decisions..
then I would NOT have donated... |
|
|
|
|
|
Any company owners out there?
Next week lets tell our workers that over the next few weeks their pay is going to be gradually cut in half and oh, you guys are also going to have to work a few extra hours each week!!
The reason I am doing this is totally arbitrary, my house my rules.
Life is nothing but a series of choices and I am now choosing to cut my own daily BOINC RAC in half because I just don't like arbitrary nonsense!
|
|
|
|
|
Any company owners out there?
Next week lets tell our workers that over the next few weeks their pay is going to be gradually cut in half and oh, you guys are also going to have to work a few extra hours each week!!
The reason I am doing this is totally arbitrary, my house my rules.
Life is nothing but a series of choices and I am now choosing to cut my own daily BOINC RAC in half because I just don't like arbitrary nonsense!
Yawn...get over it already. It's only the boinc point collectors that seem to be crying the most. I'm here to find primes, period. Go collect points somewhere else.
____________
|
|
|
|
|
Yawn...get over it already. It's only the boinc point collectors that seem to be crying the most. I'm here to find primes, period. Go collect points somewhere else.
There is quite clearly two camps here the credit crunchers and the prime crunchers, now I have not seen anybody from the credit camp tell the prime crunchers to 'live with' 'yawn' 'go somewhere else' blah blah bllllahhhhh, however you as part of the second party are getting quite a rep for it.
I know I can go as do other people, the posts where to make a case to the admins to reconsider the credit change |
|
|
|
|
Yawn...get over it already. It's only the boinc point collectors that seem to be crying the most. I'm here to find primes, period.
This.
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
Yawn...get over it already. It's only the boinc point collectors that seem to be crying the most. I'm here to find primes, period.
This.
::ponders a bit::
I was going to just "+1" this, then it occurred to me...
Instead of halving the credit, double it.
The credit hounds will flock to the sieves, letting the prime finders run LLRs instead. Everyone's happy. :)
____________
My lucky number is 75898524288+1 |
|
|
|
|
Yawn...get over it already. It's only the boinc point collectors that seem to be crying the most. I'm here to find primes, period.
This.
::ponders a bit::
I was going to just "+1" this, then it occurred to me...
Instead of halving the credit, double it.
The credit hounds will flock to the sieves, letting the prime finders run LLRs instead. Everyone's happy. :)
HA HA HA!!! I like it!
____________
|
|
|
KEP Send message
Joined: 10 Aug 05
Posts: 304
ID: 110
Credit: 13,646,296
RAC: 0
   
|
|
Could someone please do a benchmark of their CUDA and tell us exactly how long their PPS sieve units run for? Then when these numbers is presented, could someone then tell us exactly how much credit (according to the good old BOINC bencmark decided cobblestone value) a WU is worth?
This thread is getting quite lengthy and boring, with people complaining for various reasons. But there is no reason to spark a war between the admins and those of you complaining. It is true, as stated, that there has apparently been some sort of miscommunication, however it is also true and has been it for several years, that most fixed credit WUs in the various BOINC projects has tended to grant more credit to each user than the actual worth of their productivity. In other words, where the cobblestone, should resemble the actual TeraFlops produced, it is no longer, due to the widespread Fixed Credit system, the case. So please guys and girls, put your weapons down and start focusing on productivity rather than credit.
If you're only in it for the credits, then try and do 4 years of CRUS work, as I have done until writing moment, and all the credit it has given me is 2 Top5000 primes and my nick on a very short list of contributors. Thats all the credit I've been granted, as so all the credit any other CRUS users has been granted. There is no cobblestones or anything else, for those of us not contributing to the PRPnet part of the project.
Well this is just my 2cents, but I can see it from both sides, since I'm both a sieve and Prime guy, however please be aware CUDA guys that your work is very supportive and productive, and it actually helps the CPU guys save a ton of time and energy, and henceforward helps this project leep further than it else would have done. So please stay and try and resolve this issue, wich is solveable without slamming with the doors.
So who of you can contribute the numbers to solve the solution in paragraph 1. Maybe 100 different computersystems with different CUDA cards, should be totalled and then an average of these 100 systems, should be the credit given for the future.
Also please be adviced, that the time it takes for you to complete a single sieveunit will decrease as the sievedepth increases, since fewer Prime-p's will have to be tested per range. This is btw also what happens in sr2sieve. In this case be aware that the difference between a WU prior to timeincrease was sieving at around p=100P and now it sieving around p=3P, wich is a difference of more than 33 times. Though I'm not sure about this, my experience with srsieve tools, seems to indicate that the n range also has some sort of influence on the sievespeed. Maybe others can confirm or reject.
Take care everyone
KEP |
|
|
|
|
|
Hi All,
Several people have said they are leaving PG.
I hope you enjoyed your time here and continue to enjoy BOINCing elsewhere.
I respect your choice, Good Luck elsewhere.
Now on to silver up my bronzes.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 |
|
|
|
|
|
2,767.78 seconds on stock clock gtx 295: credits: 4,327.00
as of July 7: 3:03 utc reporting time.
____________
|
|
|
mfbabb2 Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 21,825,457
RAC: 0
|
|
Run time (sec) CPU time (sec) Credit Application
55,686.91 ___ 462.33 ___ 4,327.00 ___ Proth Prime Search (Sieve) v1.39 (cuda32)
NVIDIA GeForce 9400 (253MB)
____________
Murphy (AtP)
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Microcruncher* wrote: LLR tests on numbers in this range will take longer to complete.* How does this effect the optimial sieve depth for this range?
It will require a deeper sieve.
JohnMD wrote: Run times for PPS sieve on my humble cuda during the last day or so have been 10-15% higher than before.
Is this also part of the plan ?
It's the nature of the sieve at lower p. Times should improve as p gets deeper. Final credit will be adjusted for the longer run times.
KEP wrote: Could someone please do a benchmark of their CUDA and tell us exactly how long their PPS sieve units run for? Then when these numbers is presented, could someone then tell us exactly how much credit (according to the good old BOINC bencmark decided cobblestone value) a WU is worth?
Credit is not calibrated to tpsieveCUDA. It is calibrated to 32 bit tpsieve.
----------------------------------------------------------------------------
I apologize for the confusion this is causing people and obviously for the completely false conclusions that are being drawn. My English is not that good. ;) I hope the following translations will help.
Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks.
We will pause once 3607 is reached. This will give us time to evaluate the impact of the 3M-6M sieve. If testing times are shorter, we'll adjust the final credit lower. If the testing times are longer, we'll adjust the final credit higher. Once the final credit is established, we'll continue from 3607 and decrease 200 each week until the final credit is reached. The entire process is expected to take 12-16 weeks.
For the Proth Prime Search, originally, the applications used were srsieve and sr2sieve. Now the applications are srsieve, ppsieve, and tpsieve.
sr2sieve is no longer used in the Proth Prime Search (Sieve).
Unlike most other BOINC projects, it has been PrimeGrid's policy to adopt the optimized applications as the standard application. This has happened to all of PrimeGrid's applications (LLR, AP26, sr2sieve, gcwsieve). Each significant update to an optimized application is followed by a recalibration of credit. While this has never made PrimeGrid popular, it has provided participants easy access to the most efficient applications without having to fumble around with app_info files.
sr2sieve is no longer used in the Proth Prime Search (Sieve). Credit in the the n<3M phase was calibrated to 32 bit sr2sieve. 32 bit tpsieve was the optimized application. Since sr2sieve is no longer used in the Proth Prime Search (Sieve), 32 bit tpsieve has become the STANDARD application and credit is now calibrated to that.
____________
|
|
|
|
|
...
There is quite clearly two camps here the credit crunchers and the prime crunchers, now I have not seen anybody from the credit camp tell the prime crunchers to 'live with' 'yawn' 'go somewhere else' blah blah bllllahhhhh, however you as part of the second party are getting quite a rep for it.
Indeed :: The guys that don't do it for the credits.. Please stay out of this thread?! We know your standpoints.. and in your view you don't give a damn about the credits. You just want to find a titan prime (once in a lifetime)
BUT we ARE getting credits for the amount of work we have done.. and we DO have competition going on. Changing the rules (that is how I view the second iteration from 3607 to 2251) during the game doesn't make sense.
And it is not about leaving or not leaving PG.. it is about managing the project.
Clearly that is done in a ill way |
|
|
|
|
|
The runtimes will decrease over time.* They also decreased during the 0-3M phase but it was nearly impossible to notice for the casual viewer because the WU length and workload was increased several times (from 1G/1M/1200k/+1 sr2sieve WUs to 6G/3M/10000k/+-1 tpsieve WUs - leading to a roughly 300 times larger sieving space) and the credits were adapted to reflect the changes.
The sieving application switched from sr2sieve to ppsieve to tpsieve (ppsieve + Riesel sieving), the WU lengths were increased (partly to reduce the server load) and the number of participants who followed the developments was much smaller.
Nowadays every cruncher and is dog is alerted by even the tiniest change in run times but last fall things were different.
*The question is: How much? That's the thing the admins will examine and act accordingly. If the higher exponent range leads to higher average runtimes credits will be adjusted according to the numbers.
____________
|
|
|
|
|
..
Have been thinking that myself but it is what it is, everybody will either have to live with it or move on to some other project.
Well I only recently joined PrimeGrid.. and got enthusiastic about it! Even so much that I donated some Euro's for the project!
If I had of known that the project-leaders/management (or however they are called) made such (un-explainable) decisions..
then I would NOT have donated...
How much did you give Camulos? I might well be prepared to give you the money back. |
|
|
|
|
The credits issued were higher than they would have been due to the benchmark being 32bit sr2sieve (which was done to be fair in the first place). The reduction is getting rid of the artificially high score. So, we've been getting a bonus all this time, and now it's going to reduce. If that makes anyone so unhappy that they leave, that's sad, but is their choice. Personally, I'm here for the project purposes, rather than credit.
I agree. I enjoyed the benefit of high credits here but when there is no longer a valid rational for the credit level it has to change. That change is upon us and it seems well thought out and fair. I'm glad the project is basing the credits on standards and not just pulling them out of their hat. It's been a fun credit run. Now it's time to adjust. I'll still be here crunching away as usual.
Thanks for the great project!
Regards/Beyond |
|
|
|
|
Any company owners out there?
Next week lets tell our workers that over the next few weeks their pay is going to be gradually cut in half and oh, you guys are also going to have to work a few extra hours each week!!
The reason I am doing this is totally arbitrary, my house my rules.
Your statement doesn't work here cncguru... at your job--you ARE EMPLOYED. You are a volunteer here. Yes we may not like the change, but it IS their house and it IS their rules. Your analogy doesn't work.
____________
|
|
|
|
|
|
Microcruncher* wrote: The runtimes will decrease over time.*
...
*The question is: How much? That's the thing the admins will examine and act accordingly.
...
Within 3 days the runtimes decreased >2%.
On fast GPU's you will rather notice this, but on my not so fast GT240 this is about 2 minutes.
____________
 |
|
|
|
|
Microcruncher* wrote: The runtimes will decrease over time.*
...
*The question is: How much? That's the thing the admins will examine and act accordingly.
...
Within 3 days the runtimes decreased >2%.
On fast GPU's you will rather notice this, but on my not so fast GT240 this is about 2 minutes.
The runtimes on my GTX 470 (Win 7/275.33/stock clocks) sank from 16:25 (friday) down to 16:05 (yesterday), that's also ca. 2%.
____________
|
|
|
|
|
|
First, thank you VERY much for the recalibration announcement being formalized.
Given the tools at PG's disposal, is it now practical to use sr1sieve and single K/file sieves in BOINC? Have the candidate list(s) been sieved to sufficient depth to permit this finer grain sifting or is this not practical at PG which would permit highly efficient memory utilizationand wildly fast CPU and GPU work?
I've been doing theoretical math and even the Riesel Sieve project since the early days. I've also a minor in Mathematics (with my double major EE/CE). I'm doing Quantum work now at work... That will make your head spin around a few times... (LOL).
Thank you for the rapid reply and I hope PG can remain the premier project it's always been.
Thanks,
C
____________
|
|
|
Ken_g6 Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 969
ID: 3110
Credit: 365,802,398
RAC: 79,633
        
|
Given the tools at PG's disposal, is it now practical to use sr1sieve and single K/file sieves in BOINC? Have the candidate list(s) been sieved to sufficient depth to permit this finer grain sifting or is this not practical at PG which would permit highly efficient memory utilizationand wildly fast CPU and GPU work?
Well, first of all, I don't believe there is an srsieve for GPUs. I've thought about developing one that would be like sr1sieve, using Pollard Rho for Lograrithms; but I haven't done anything about it yet. If you find one out there, let me know and I'll BOINCify it!
Even if there was one, for PPS Sieve we're working with 4997 K's in a range 5-9999. tpsieve is faster for this range than any of the srsieves. See this post for the basic tpsieve algorithm; though it's been slightly modified since then, and highly modified with Montgomery math for GPUs.
____________
|
|
|
|
|
|
:) |
|
|
|
|
:)
I agree, the complainers are indeed.
____________
|
|
|
|
|
|
It is good to see the credit ratio being set to fair, i allways wondered how come my gts 450 (~450 sp gflops) was able to outperform my ati 6870 (~2000 sp gflops) card in terms of credit, so now i understand.
People who are complaining: please xxxx. |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
:)
I agree, the complainers are indeed.
People who are complaining: please xxxx
A little Civility Please, everybody should be free to express themselves as they Please without being called :) & told to xxxx ... You may disagree with their Opinion but they are entitled to express their Opinion none the less as much as you are ...
____________
|
|
|
|
|
:)
I agree, the complainers are indeed.
People who are complaining: please xxxx
A little Civility Please, everybody should be free to express themselves as they Please without being called :) & told to xxxx ... You may disagree with their Opinion but they are entitled to express their Opinion none the less as much as you are ...
+1
There's a worldwide etiquette crisis going on for at least 20 years now. And all this negativity and rudeness expressed in those three posts, tell a lot about their authors.
____________
  |
|
|
|
|
|
Im sorry if i misrepresented or misspoke myself, everybody are free to speek afcourse, as long as they dont start complaining and whining.
Imo it was cheating anyways, when weaker cards outperformed the more powerful cards. I was just wondering for how long is it going to continue (i wasnt complaining about it anywhere). |
|
|
Vato Volunteer tester
Send message
Joined: 2 Feb 08
Posts: 982
ID: 18447
Credit: 2,340,591,191
RAC: 819,083
|
|
if you're unhappy about the performance on ATI cards, the best things to do would be to get AMD to produce better OpenCL implementations, or port the psieve apps to CAL (which AMD are going to end support for anyway). It's not cheating.
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
fyi...PrimeGrid is certainly always open to improvements to any of its applications. Any optimized applications will gladly be implemented. All the code is open source, easily accessible, and available for download.
As for ppsieve/tpsieve, Ken has been brilliant with his improvements. Additionally, I'm sure he's open to any suggestions or ideas. Please feel free to review the code here:
Finally, PrimeGrid has a team of volunteer testers with various hardware ready and willing to test any improvements. Below are several active threads that have been used to test any changes:
* Mac support by Iain.
____________
|
|
|
|
|
fyi...PrimeGrid is certainly always open to improvements to any of its applications. Any optimized applications will gladly be implemented. All the code is open source, easily accessible, and available for download.
As for ppsieve/tpsieve, Ken has been brilliant with his improvements. Additionally, I'm sure he's open to any suggestions or ideas. Please feel free to review the code here:
Finally, PrimeGrid has a team of volunteer testers with various hardware ready and willing to test any improvements. Below are several active threads that have been used to test any changes:
* Mac support by Iain.
The link for the OpenCL source points to the cuda source
____________
|
|
|
|
|
It is good to see the credit ratio being set to fair, i allways wondered how come my gts 450 (~450 sp gflops) was able to outperform my ati 6870 (~2000 sp gflops) card in terms of credit, so now i understand.
People who are complaining: please xxxx.
Sir Sent, I believe your demand to xxxx is not a fair comment because you happen to think that way. Please do not forget others have a right to express their opinion.
With regards to your overall statement, I happen to disagree with you, so please do not shoot me or tell me to xxxx, I have as much right to express my opinion as you.. Some of the earlier guys to this project before the credit cut back or CreditNew whatever it is called were on eve ground and were credited evenly. Suppose I just happen to join this project with the same equipment as you have, do you think it is fair that it should take me twice as long to get to the same points as you currently have?, not forgetting that cost of equipment and the cost of electricity is rising all the time.
In my opinion, the only fair way to solve is to close the current project and start the new one at a score that the project admins deep appropriate and fair. It is not fair to change that project score in the middle of the stream. This is like moving the goal post in the middle of the game and in football it will cause world war three...
|
|
|
Ken_g6 Volunteer developer
Send message
Joined: 4 Jul 06
Posts: 969
ID: 3110
Credit: 365,802,398
RAC: 79,633
|
The link for the OpenCL source points to the cuda source
Allow me to clarify:
CUDA
OpenCL
And slightly off-topic:
Cullen/Woodall CUDA
Yes, they're all in the same repository. They're more similar than you might think.
____________
|
|
|
|
|
|
CreditNew did this have any baring on the recalibration and will you be using CreditNew now/future? |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
n my opinion, the only fair way to solve is to close the current project and start the new one at a score that the project admins deep appropriate and fair. It is not fair to change that project score in the middle of the stream. This is like moving the goal post in the middle of the game and in football it will cause world war three...
Changing the Credit has happened a lot at a lot of other Projects too. Take primaboinca, they recently Doubled the Wu Length without any Credit increase so effectively you could say they cut the Credit in half.
Hardly anybody at primaboinca said anything other than ask if the increase in Wu Length was intentional or something else was wrong. The only reason there's a stink here at PrimeGrid on the Credit Reduction is because of the Amount of Credit that could be had in such a short period of time.
If you think the Project should start a New Project just because the Credits are going to drop (already have actually) then maybe all the BOINC Projects should start a New Project every time somebody new joins BOINC. Is it fair to them that they have to try and catch everybody ahead of them with all the Credits they have already ?
If a new Project had to be started at all the BOINC Projects every time there was a Credit change there would be 2000 or more BOINC Projects already I suspect with all the Credit Changes that have been done since BOINC started ... ;)
____________
|
|
|
|
|
In my opinion, the only fair way to solve is to close the current project and start the new one at a score that the project admins deep appropriate and fair. It is not fair to change that project score in the middle of the stream.
Project credit policies change constantly according to application efficiency, WU type and various other factors. For recent examples look at Spinhenge, AQUA, Moo! and Yoyo to name a few. The Buddha said change is inevitable. You're going to have to get used to it :) |
|
|
|
|
|
..To those of you (us?) who are only in it for the Mega-Primes, and don't give a fig for the credits..
Please remember your CPUs are surfing the crest of an incredibly huge wave, built for you by the credit-hungry GPU-Sievers, so I wouldn't be so hasty to give them all the boot!
- You may just find yourself chewing sand before long...
____________
 |
|
|
|
|
|
Regarding GPU version of srsieve / tpsieve / sr1sieve / et ali.
Gang,
I develop GPU apps for a living. My career path is winding down from the huge demand it's been in the past few years and why absent so much since 2005. Doing modeling + graphics at approx 128 Gigabits/frame (multi-panel) with the frame rate of 75 fps makes for some interesting challenges (lighting, hidden line removal, aka... GL-like stuff) without dropping a frame over udp. You can find examples of this on the net... Take a F/A-18 and fly (emulation not simulation) it at mach 2+ and do a high speed barrel roll. That makes for some interesting stresses on the GPUs and CPUs not to mention the LANs (it's all DV until it hits the display system).
Math side of this is (from the DV perspective only) (1920×1080×24×4×9×75)/1048576 = 128144.53125 Mbit/sec
However, with that now 'stable product', and me having moved on to a new task, I have more time (hence my quiet return to PG).
I still have my tools, both ATI and nVidia GPUs, so will be, unless directed/requested otherwise to see what, if anything, I can help port over from Geoff's great code to the GPUs to 'BOINCify' . Geoff took over when I left for work and Joe finished the code. Since then, Geoff has been doing his normal 'fantastic work'.
Those of you who have access to my email (or can get it) please don't hesitate to contact me off-board directly with your thoughts, experiences, or direction on what is best to do. I ask for your guidance.
Thanks,
C.
edit: provide the math behind the work.
____________
|
|
|
|
|
|
Rules:
• Posts must be 'kid friendly': they may not contain content that is obscene, hate-related, sexually explicit or suggestive.
• No commercial advertisements.
• No links to web sites involving sexual content, gambling, or intolerance of others.
• No messages intended to annoy or antagonize other people, or to hijack a thread.
• No messages that are deliberately hostile or insulting.
• No abusive comments involving race, religion, nationality, gender, class or sexuality.
Lennart |
|
|
|
|
CreditNew did this have any baring on the recalibration and will you be using CreditNew now/future?
Really, it sounds like the phase in of DA's well thought out control and domination of credit called CreditNew. It is forced on projects if they want to upgrade to the new BOINC server software. Forced as in if you want the new BOINC software youll have to use it, its probly hackable, but will it be just as good?...Slippery slope, slippery slope.
Any light on the subject or an answer.
I will still continue to crunch here, I found a prime in SG, so that was cool.
____________
My first Prime
20208428036625*2^666666-1 (SGS)
PROUD MEMBER OF TEAM CARL SAGAN |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
CreditNew did this have any baring on the recalibration and will you be using CreditNew now/future?
Any light on the subject or an answer.
CreditNew had no impact on this current recalibration. tpsieve replaced sr2sieve as the standard application and credit was adjusted accordingly.
____________
|
|
|
|
|
|
I've read the thread (felt like in a roller-coaster) and I'll stay... Sad is the fact that two of my preferred projects (PG and Aqua) and, simultaneously, two of the best "paying" projects seem to have been overcomed by an urge to re-evaluate their retributive (because, yes, points are my salary as a volunteer) policy.
Is this a coincidence?
But I'll keep on crunching because PG is well administered and shows results.
Thanks |
|
|
|
|
I've read the thread (felt like in a roller-coaster) and I'll stay... Sad is the fact that two of my preferred projects (PG and Aqua) and, simultaneously, two of the best "paying" projects seem to have been overcomed by an urge to re-evaluate their retributive (because, yes, points are my salary as a volunteer) policy.
Is this a coincidence?
But I'll keep on crunching because PG is well administered and shows results.
Yep, AQUA has had no end of problems partly due to creditnew and partly to misconfiguration. PPS Sieve credits have been re-evaluated for the reasons you've read above. |
|
|
|
|
Microcruncher* wrote: The runtimes will decrease over time.*
...
*The question is: How much? That's the thing the admins will examine and act accordingly.
...
Within 3 days the runtimes decreased >2%.
On fast GPU's you will rather notice this, but on my not so fast GT240 this is about 2 minutes.
The runtimes on my GTX 470 (Win 7/275.33/stock clocks) sank from 16:25 (friday) down to 16:05 (yesterday), that's also ca. 2%.
My oc-ed gts 450 512 mbt (shaders @1820mhz), runtimes now for pps sieve cuda 32.13 min and before the change 30.30 min. Gtx 570 oc-ed (shaders 1590mhz) runtime now 11.30 and before about ~11 minutes (dont remember exactly). Drivers and clocks for all cards are the same at all times. And yes the memory speed for all cards is running at 1/4-th. |
|
|
|
|
|
...pardon me for not reading every post. I just noticed a lot of complaints from the credit-hungry contingent and went oh, this again?
Aren't credits/cobblestones based on the amount of FLOP/s in a given unit? Something like that. The new credit system is totally different, I know, and don't claim to know anything about it as I'm a bit out of the loop. The current (or I guess past) version of pps sieve was very very efficient, hence the absurdly high credits.
Now the apps are changing, and I'm afraid I don't quite understand (though I can imagine that since we will be sieving bigger and bigger numbers, in order to keep runtime the same the amount of numbers per sieve unit will be decreased). Is this all because of the transition to a bigger, deeper, and as yet untested sievefile?
That said, I don't care about credit either way; I care about the tangible results produced by crunching this particular project. If cutting it in half is the right way to go, then so be it. It'll still "pay" twice as well as Moo! or MW for me, anyway.
____________
 |
|
|
|
|
Aren't credits/cobblestones based on the amount of FLOP/s in a given unit? Something like that. The current (or I guess past) version of pps sieve was very very efficient, hence the absurdly high credits.
Doesn't a more efficient algorithm mean less FLOPs required to get the same result? e.g.: FFT is O(nlog(n)) doing a job that would be O(n^3) without the ingenuity in the algorithm. So is credit based on the amount of FLOPs that would be required to brute force a problem?
____________
|
|
|
|
|
|
I guess I need new glasses, I would swear the 1st post in this thread says wu should be paying 3607 now but in reality they are paying 3507. I guess I'm just not seeing one of the numbers correctly. Yes, this is sarcasm! |
|
|
|
|
Credit will be decreased by ~200 cobblestones per week until those numbers are achieved. Once 3607 is reached, we'll pause while we review the impact of the 3M-6M sieve to testing times. We'll adjust the 2251 number, if needed, and then continue the ~200 per week decrease until the final number is reached. The entire process is expected to take 12-16 weeks.
Status for July
31 July - 3607
24 July - 3719
17 July - 3923
10 July - 4127
3 July - 4327
2 July - 4523
Looks like a typo. I wondered why the drop was supposed to be only 112 this week when it was between 196 and 204 for all the others. |
|
|
|
|
|
I have been in it for the credits ever since I went with Prime Grid and 2 GTX 260's,I'm currently with just one(GTX 560) as my son's motherboad died due a lightning strike(he had just bought a GTX 580). I have been on vacation and noticed the reduction in credits on return and was directed to this thread. I hope Prime Grid will continue in some way go back to the credits pre July 2nd because I have really enjoyed looking at my daily stats(going over a million a day at times). Will Prime Grid have a project that will very closely follow pre july numbers? I was a long time distributed.net user(10 years) and it took me forever just to get to one million and beyond just going cpu(never again!). When I saw how Prime Grid was doing in daily credits with my GPU's I fell in love with this project. Now I am going to have to re-evaluate my participation in this wonderful project,but am going to wait til the bottom falls out(which I'm afraid it will at some point),then move on to something more attractive if it is out there. |
|
|
|
|
..To those of you (us?) who are only in it for the Mega-Primes, and don't give a fig for the credits..
Please remember your CPUs are surfing the crest of an incredibly huge wave, built for you by the credit-hungry GPU-Sievers, so I wouldn't be so hasty to give them all the boot!
- You may just find yourself chewing sand before long...
Amen to that,brother! |
|
|
|
|
..To those of you (us?) who are only in it for the Mega-Primes, and don't give a fig for the credits..
Please remember your CPUs are surfing the crest of an incredibly huge wave, built for you by the credit-hungry GPU-Sievers, so I wouldn't be so hasty to give them all the boot!
- You may just find yourself chewing sand before long...
Amen to that,brother!
I don't think so.
Even after planned credit decreasing PG GPU projects will be enough attractive for credit-hunters.
For example, today your GTX560 Ti @ PG PPS Sieve gives 3.4 credits/s
after all it will give 2.2 credits/s
Instead, today Collatz, GPUGRID give you only ~1.25 credits/s, other projects even less.
So, a wicked generation, you will not get anywhere.
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
|
I find the idea of "being in it for the credits" somewhat amusing, given that credits have no value whatsoever. I do understand that mentality, however, since I share it to some degree. The competition is fun. The acquisition of pretty colored pixels (badges) is fun. Nonetheless, I find it amusing, even if I'm laughing at myself.
That being said, however, for the record, I have to agree with everyone who is complaining about the reduction in credits. I don't necessarily share other people's views on what credits should represent, or how they should be assigned, but I do believe that they should represent a given unit of work.
We're doing the SAME work as before. We should receive the SAME credit as before. End of story.
The reason the GPU tasks are/were cranking out so many credits isn't because they were miscalculated. The reason is because of the incredible programming done by Ken to make them so efficient. The credits were so high because the amount of work being done was immense. Coupled with the drastic increase in GPU performance as new chips came out (and people started buying newing cards), the credit counters went through the roof.
Reducing the credits is not only a slap in the face to those who care about it (I'm not one of those, btw), but, more important to me, it represents a break in the continuity of the "work-to-credit" equation. If credits are to have any meaning at all,they need to represent a consistent measure of the work being done. This change violates that most basic of concepts.
____________
My lucky number is 75898524288+1 |
|
|
|
|
The basis for the BOINC credit system is the cobblestone named after Jeff Cobb of SETI@home. The basis of the system is the concept that 200 cobblestones would be claimed for one day of work on a computer with the following specifications:
1,000 double-precision MIPS based on the Whetstone benchmark.
1,000 VAX MIPS based on the Dhrystone benchmark.
3507 cobblestones must be equal to 17.5 days = 420 hours of execution any PPS Sieve workunit @ standard (model) CPU.
Here you can see, what is a real time of PPS Sieve execution @ different CPU's
Take, for example, Intel Core2 Quad Q6600 @ 2.40GHz. CPU spent ~18 hours per WU.
Look at Intel Core2 Quad Q6600 @ 2.40GHz benchmark result:
1856 floating point MIPS (Whetstone)
6952 integer MIPS (Dhrystone)
So, Q6600 @ 2.40GHz ~7 times faster than standard (model) CPU.
We can estimate that model CPU will need 7*18 hours = 126 hours per WU.
420 hours vs 126 hours.
Even the rough estimate shows that credits per PPS Sieve workunit are over 3 times bigger than they should be. For both, CPUs and GPUs.
____________
|
|
|
|
|
We're doing the SAME work as before. We should receive the SAME credit as before. End of story.
This is where you got it wrong. We moved on to a new Sieve and so the credits are being reduced because of thew new Sieve.
Quoting from the first post: "PPS (Sieve) cobblestone per WU is being decreased to reflect the majority use of the tpsieve application."
So the change is because the work has changed.
I have no qualms with the reduction. Sieving is not the real work here. LLR is. Sieve just reduces stress on the LLR by removing candidates not previously removed.
As it has been pointed out before, the reduction will still make this the highest paid GPU project on NVidia. |
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
|
My understanding -- and I might have this completely wrong -- is that the GPU application didn't change. It's the CPU version that changed. As far as I know, the GPUs are doing the same work as before.
____________
My lucky number is 75898524288+1 |
|
|
|
|
My understanding -- and I might have this completely wrong -- is that the GPU application didn't change. It's the CPU version that changed. As far as I know, the GPUs are doing the same work as before.
GPU and CPU does exactly the same work, there is no difference in the pool of work. It's all the new tpsieve work. |
|
|
Scott Brown Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2540
ID: 1178
Credit: 29,799,904,808
RAC: 11,912,495
|
My understanding -- and I might have this completely wrong -- is that the GPU application didn't change. It's the CPU version that changed. As far as I know, the GPUs are doing the same work as before.
GPU and CPU does exactly the same work, there is no difference in the pool of work. It's all the new tpsieve work.
Actually, I think you are both off a bit here in the way things are being put. As I understand it (John, Rytis, Lennart please correct me if I am wrong), the sequence/reasoning is as follows...
1) Original application developed for CPU and calibrated such that it is fair (relatively) with credit generated for other sub-projects at that time (no GPU apps available then).
2) GPU app added and calibrated to represent a fair match to amount of work done with CPU (though in truth, this was always a bit on the high side).
3) Switch to new sieve application results in change in how the CPU is able to complete work such that the gap between CPU and GPU is less than before for this sieve.
4) GPU credit is adjusted downward to reflect this reduced efficiency gap.
...so why not raise CPU credits. Probably because that would throw the CPU credit balance relative to other sub-projects out of whack requiring a credit adjustment across most of the PG sub-projects. Clearly, there were probably some options in how credit would/could be adjusted given the necessary application changeover, and it appears that PG chose one of the simplest options requiring only the recalibration of GPU credit in this case.
____________
141941*2^4299438-1 is prime!
|
|
|
|
|
...so why not raise CPU credits.
I do not get what people do not understand. The credit has dropped for both CPU AND GPU. it's a new sieve sub-project under the PPS project. There is no difference in credit. Check out this unit done on a CPU WU 198830589. It received the exact same 3507.00 credit as GPU currently does.
They are the same work. Neither one gets more or less credit. It just takes longer on a CPU. |
|
|
|
|
...so why not raise CPU credits.
I do not get what people do not understand. The credit has dropped for both CPU AND GPU. it's a new sieve sub-project under the PPS project. There is no difference in credit. Check out this unit done on a CPU WU 198830589. It received the exact same 3507.00 credit as GPU currently does.
They are the same work. Neither one gets more or less credit. It just takes longer on a CPU.
They returned the same result, they received the same credits.
But they don't did the same work.
CPU app was changed, it's true, but GPU app doesn't changed, app sieves the same amount of numbers, spend the same time as before.
So the main motive to recalibrate credits is to bring the credits in correspondence to standard CPU productivity. It means to measure how many time standard CPU needs to execute workunit using ppsieve app.
It doesn't matter how long GPU app worked, how many kW it consumed, how many heat it produced. The only thing matters - how many integer & float operations standard CPU needs to do for the same correct result.
In fact Cobblestones are the measure of time of running WU @ standard CPU. Thus if, for example, you have optimized some code and execution time @ standard CPU reduced at twice, you need to reduce earned credits at twice too.
Remind, standard CPU means a computer with the following specifications:
1,000 double-precision MIPS based on the Whetstone benchmark.
1,000 VAX MIPS based on the Dhrystone benchmark.
____________
|
|
|
|
|
|
When we calculate credit on pps sieve we only use CPU work.
CPU,GPU get the same credit !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CPU,GPU do the same amount of work !!!!!!!
all WU has the same amount off work to do.
It is the CPU work that gets to much credit because we used sr2sieve as main app when we calculated credit.
GPU's will get the same credit as CPU
We do not use sr2sieve anymore !! We use TPSIEVE.
Tpsieve is much faster and thats our new main app on pps sieve.
We will countinue to only use CPU when calculating credit !!!
GPU's gets the same credit !!!
We don't even think about GPU when calculating credit !!
Lennart.
|
|
|
|
|
|
Lennart,
You said absolutely the same as I said, but in more explicit style.
____________
|
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
We don't even think about GPU when calculating credit !!
Lennart.
 ... lol ... j/k ... It's all good
____________
|
|
|
|
|
|
wow. |
|
|
|
|
|
Now that's an explanation...Oorah!!
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
I call on the GREAT masters of the English language whose precision and economy of words are unparalleled (Mark Twain, Ben Franklin, Henry David Thoreau).
The PPS (Sieve) no longer uses sr2sieve as its main application. tpsieve is now the main application.
Previously
Starter sieve - srsieve CPU (manual)
Main sieve -sr2sieve CPU (manual & BOINC) - credit calibrated to WU
Optimized sieve -ppsieve CPU (BOINC)
Optimized sieve - tpsieve CPU (BOINC)
Currently
Starter sieve - srsieve CPU (manual)
Transitional sieve - ppsieve CPU (manual)
Main sieve - tpsieve CPU (manual & BOINC) - credit calibrated to WU
In this case, treat ALL GPU applications as optimized applications.
p.s. Some of y'all may recall that early on, sr2sieve went through a few optimizations and credit recalibrations as well. :)
____________
|
|
|
|
|
|
In other words, as soon as an optimized application have appeared, skim the cream until it became a main!
____________
|
|
|
|
|
|
Why Project give credit?
Why Project have a Badge?
Why Projects have Cruncher Rank?
Why Projects have an CPD?
Why Project have Challenge?
Why Project are log on Boinc World & Country records?
Why project have a rank on World record?
Competition??? of course, without competition , why Credit , Challenge & Badges!!!!!!!!!
I did invest a lots of money on Video cards , GPU & CPU & electrical bill,
to have a 800k per day? Now to keep the same speed i will have to invest again the same amount of money?
That's are rewards....Badge , rank & ....
When a Project will find somethings and will make money out of it , do we '' Cruncher '' will have a piece of the cake? no of course......
Is there a war on Nvidia & Ati over the market????
I had FUN working on all projects and try to reach personal goal , Country rank & .....
Some of you have the nerve to write '' A- Live with that''..
You are lucky , they are rules on this Forums!!!
I do agree with Steve!!!
For me as soon another Project will rise with good credit , i will leave Primegrid , anyway it will not change anything for them , putting down the credit show they have no respect for us...
Luc Gauthier
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
|
Buckle up, boys and girls, this is going to be a long ride...
First of all, I'm writing this intending it to be in good humor and, perhaps, just a tiny bit informative. I've got my opinions, everyone else has theirs, and that's all fine and good and the way it should be.
If any part of this gets your blood boiling, I apologize for that; it's certainly not my intention. I'd prefer if you put a smiley face on while reading this. (Not an emoticon, but a real smile on your real face.) Life's too short to be angry over stuff like this!
To start, I'd like to thank John and Lennart and Rytis and everyone else who has made PrimeGrid as fantastic an organisation as it is. Well done guys!
I'd also like to thank D.A. and everyone who has contributed to the development of BOINC (and its predecessor, SETI@Home) over the last 10+ years. Again, well done guys! You have helped make the world a better place.
This whole discussion started with the admins saying they're going to lower the credits for [insert good reasons here].
Then people got upset.
My first urge was to make fun of those people; after all, as I'm prone to saying, you can't buy a sandwich with credits. Credits are worthless.
That's quite hypocritical of me because I *DO*, in fact, care about credits. I like the competition, in all of its forms. I like earning badges. But I also recognize that in the grand scope of things, it's the work that's truly important, not the credits. In my mind, those who think otherwise, that is to say, are in it solely for the credits, can't see the forest for the trees. But hey, to each their own. The world would be a boring place if everyone thought the same way as I do. Besides, the computing that's done is useful regardless of the reason it's done. Se la vie.
So, when people started complaining about the credits being reduced, I was amused. For about 3 seconds. That's about how long it took me to realize that I thought it was wrong to reduce the credits, too. But for a completely different reason.
Remember how I thanked those people up at the top? When someone starts out like that, it's a good bet that it's a prelude to presenting some sort of disagreement with those folks. And that's exactly what I intend to do.
It's not that I think they're stupid, or insensitive, or short-sighted, or anything at all like that. Quite the opposite. Like I said, they've done a great job. No doubt about that.
The problem lies with one of the earliest definitions of scope in the BOINC system: the definition of a cobblestone, which is defined as a certain amount of computation being performed. That's the genesis for why I disagree with the reduction in credits.
The standard for credits basically says, "If you do "X" number of calculations, you get one credit." That's simple enough, but it's got a horrible flaw. As anyone who has any actual training in writing software knows, there's different ways of writing the same algorithm, sometimes with incredible differences in efficiency. When you're talking about stuff that a computer can do, the numbers can be large, and that is true here too. When I say "differences in efficiency", it's very, very easy to come up with examples where the run-times of different algorithms can vary by a factor of a million. Or a billion. The simplest example is sorting a large set of numbers. This is basic stuff from a first or second year CS curriculum. Sort a large set of numbers with an efficient sort like quicksort and it will run in a decent amount of time. Use a linear selection sort, and, well, come and talk to me when it's done, ok?
So, when a credit is based on "the number of calculations performed", that means that my sort algorithm example would generate a million times more credits using a linear selection sort as compared to doing the same work using a quicksort. Even though both do exactly the same thing.
That's my problem with the system. It's based on the amount of calculations done. In my opinion, it should be based on the amount of work done. The two are not the same.
You should get X credits for sorting those billion numbers, regardless of whether you use a linear sort, a quicksort, or an abacus. But that's not the way BOINC is defined.
So, when the admins changed some stuff around, and, based on the rules, decided the right thing to do was to the lower the credits, they did the right thing.
The right thing according to their rules, that is. It's their rules, which are based on the definition of a cobblestone, that I disagree with. While what they did makes sense based upon the original D.A. benchmarks, I've never thought that was a reasonable way to measure the contributions that people make.
So when they reduced the credits, from their (and BOINC's) perspective, they did just what the title says: they recalibrated the credits based upon the way they calculate credits.
From my perspective (and the credit-rulez! dudes, too), they reduced the credit even though we're doing the same amount of work as before.
Their way of looking at it is the official way of doing things.
I just happen to think the official way of doing things is wrong.
(I'm almost done, btw.)
Now, I'm not a fool. (That's my opinion; once again, you're entitled to your own opinion.) I realize why the definition of a cobblestone is based on calculations rather than work. If it was up to me to make that decision, I'd probably have done it the same way. For one thing, it's a heck of a lot simpler and less subjective than trying to define what "work" means. It also gives you a fighting chance of having credits mean something similar across multiple projects. So there are advantages to the calculation method of defining credits.
But I still disagree with it. :)
Ok, now I'm done, except for one more set of thanks: Thank you to everyone who took the time to read this!
____________
My lucky number is 75898524288+1 |
|
|
|
|
I did invest a lots of money on Video cards , GPU & CPU & electrical bill,
to have a 800k per day? Now to keep the same speed i will have to invest again the same amount of money?
Many of us bought a few cards to take advantage of the very high credits on PPS Sieve. Truth is they were higher than any other project and it comparatively didn't take that many (or that fast) GPUs to build up a lot of credits. Now it seems things are being readjusted but PPS Sieve will still be among the highest credit applications. We've all hopefully read the reasons and they make sense. Just say thank you and know that you couldn't have done even close to as well credit-wise anywhere else. Maybe a sieve half full is better than a sieve half empty :) |
|
|
|
|
The problem lies with one of the earliest definitions of scope in the BOINC system: the definition of a cobblestone, which is defined as a certain amount of computation being performed.
.......
You should get X credits for sorting those billion numbers, regardless of whether you use a linear sort, a quicksort, or an abacus. But that's not the way BOINC is defined.
So if I start with an app that's not efficient and calibrate it to get 100 credits a day on a P100, then find/build an extremely efficient app that is 10,000,000 times faster. Then I should get 1,000,000,000 credits a day? Not a great idea IMO and not the way other projects such as MW and Collatz calculate their credits. Seems to me that the admins here are trying to do the right thing credit-wise. Lets cut the tantrums and pat them on the back for having a great project. |
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
So if I start with an app that's not efficient and calibrate it to get 100 credits a day on a P100, then find/build an extremely efficient app that is 10,000,000 times faster. Then I should get 1,000,000,000 credits a day
That depends entirely upon what you define "work" to be. In your example, the answer is an emphatic "Yes!".
It's sometimes hard to put credits in a reasonable framework because they're so intangible. But there are really, really good analogies that can be used for comparison.
Let's say you need to have the oil in your car changed. You take it to your local mechanic.
Ignoring the materials charges (the oil) and concentrating just on the labor. He can charge you one of two ways: He can charge you a flat rate for the job, or he can charge you based on the labor actually expended. Most places charge a flat rate for an oil change. If they have a new guy doing the work (presumably under supervision as he learns which end of the wrench to use), it might take five hours to do. Any other mechanic in the shop will probably do it in 15 minutes. Chances are, you'll get charged the same thing regardless of who did it, and how long it took. (This is my "work based" model.)
On the other hand, if the new guy ruined your engine while changing the oil because he didn't tighten the oil pan nut, resulting in the oil leaking out and the engine seizing, you might go and hire a lawyer to sue the mechanic. That lawyer will charge you by the hour. (At least in this example, he's charging by the hour.) You're paying for the amount of time he spends, regardless of how efficient he is. Presumably, you, in your due dilligence, select a lawyer whose billing rate is commensurate with his skills and efficiency. (That's the "calculations based" model).
Sometimes, of course, you pay the lawyer based upon the results (Insert TV lawyer commercial here: "We only get paid if you win!"). That's back t the "work based" model.
If you hire an employee, are you hiring a body to fill a chair between punching the time clock at 9 AM and 5 PM, or are you hiring him to perform a specific job?
Both models have their place. I just prefer the work-based model.
Sorry if you think this is throwing a tantrum. Trust me, I'm capable of ranting with the best of them. This ain't it. ;)
____________
My lucky number is 75898524288+1 |
|
|
|
|
So if I start with an app that's not efficient and calibrate it to get 100 credits a day on a P100, then find/build an extremely efficient app that is 10,000,000 times faster. Then I should get 1,000,000,000 credits a day
That depends entirely upon what you define "work" to be. In your example, the answer is an emphatic "Yes!".
First of all the tantrum comment wasn't aimed at you, just the general tone of this thread lately. Sure, some people feel something is being taken away but it has to be tempered with the knowledge that no project anywhere awarded even close to the credits that PPS Sieve did. That goes for CPU also.
In your scenario all I would have to do is design a really inefficient stock application and then replace it with an efficient one later. Voila, massive credits. Not a good idea and not workable in the long run.
Edit: BTW, look to the left and you can see that I'm losing a lot more credit/day than most of the people complaining. But the adjustment is fair and I think it brings PPS Sieve at least more in line with other projects. That's a good thing. |
|
|
|
|
|
Hi everyones...
i start crunching DATA the minute Seti Project was starting...Many years ago..
I am not stupid, of course Science evolved, of course it is something great to find more Prime...i am also willing that credit be erased, and Badges deleted, but it make less interesting that's all...
But remember this '' I am not a computer Scientist, the challenge i have with all project & stat & Credit & rank, is to find new way to improved my computer speed.. try 64 bits, install OS, Video card with more GPU performance, try to enhance my Pc'' when i see my rank decrase, i say to myself '' hey , it is time to find something faster, that is the fun i have..
I don't blame Primegrid for that , but i don't like when some people say '' This is not a competition..''or '' the competition is FUTILE''
So 1= Let the data cruncher having their fun!!!
or 2= Deleted credit & Primefinder list base on the top finder list to the lowest !!!BTW , when i saw one cruncher who was at the top, i check what kind of computer he was using to have a great performance, i was so surprise , we exchange email, and he explain to me about multiple core and GPU, so you see it is not everybody who went to Harvard but Boinc project is a great window to learn.......
Question? Is it better to lower Credit on GPU or to Increase Credit on CPU to make everyones Happy???
Regards
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
..the challenge i have with all project & stat & Credit & rank, is to find new way to improved my computer speed.. try 64 bits, install OS, Video card with more GPU performance, try to enhance my Pc'' when i see my rank decrase, i say to myself ''
Good point. Excellent point, in fact.
I should have mentioned that credits aren't completely worthless. They are an excellent way to compare the speed of different hardware/software combinations under real-life conditions.
And the competition is fun, too. :)
Is it better to lower Credit on GPU or to Increase Credit on CPU to make everyones Happy???[/b]
Another point I intended to address and forgot about.
I'm going to take a lot of heat for this one.
I think PG should drop the CPU-based PPS and CW sieves completely.
Remember, I look at all BOINC projects through the lens of the original SETI@Home project: people volunteering their idle computer time to help an organization with their computing goals. In that light, the use of CPUs to process those sieves is simply a waste of a perfectly good CPU. It could be doing something useful instead of spending its time (and electricity and carbon footprint and purchase cost) doing something it is no longer very good at (in comparison to what GPUs are capable of doing.)
Another analogy, this time sports. In baseball (sorry for the folks across the pond, but I'm a yank so you get a U.S. analogy) if you have a major league player who loves to pitch, but sucks at it, and is a fantastic hitter, you don't let him pitch because he wants to. His job on the team is to do what's best for the team.
That analogy falls apart somewhat because we're volunteers, and the admins do need to generate good will and keep the volunteers happy. But on the other hand, the purpose of this project is to find prime numbers, and utilizing the resources efficiently is certainly in the best interest of that goal. GPUs have immensely more computing ability than CPUs do, but their specialized design makes them more suited for some applications than for others. Given the current state of the technology, using a CPU for something that can be done 100 or a 1000 times faster by a GPU is clearly not the best use of that resource.
I'll save you the trouble of presenting "but what about the person with no video card who wants to get a badge?" argument. My response to that is even a dirt cheap bottom of the barrel CUDA capable gpu will be a vast improvement for them, and the almost negligible cost of the GPU will be made up by electricity savings. Of course, if they decide to crunch on both CPU and GPU, they won't save money on their utility bill, obviously, but they'll be getting far more credits and badges.
The admins, who are clearly nicer people than I am, believe that everyone should be free to crunch whatever they want. That's a perfectly reasonable position. I guess it's a good thing they're in charge! :)
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Well Said Michael...
Now this forum sound very positive!!!
Thank's for your good comment!!!
Now .. let crunch some Wu's and find THE PRIME...
BTW , All united , we can kick the word '' IMPOSSIBLE '' and if aliens ET, exist , they will be jalous of us, éhéh
Regards to All
Luc |
|
|
|
|
|
Hey Michael,
I'm sending lots of heat, and we have plenty in Virginia, your way!
Jim
CPU only and proud of it! :-)
EDIT: When I earn a color, that sucker is earned. |
|
|
|
|
|
I use CPU & GPU
3 Computer with 4 Cores Each = 12 CPU's full time
2 have additional GPU GTX 470
1 have Ati
Total 12 CPu + 3 Gpu
BTW Gpu gtx470-each contains 448 Cuda cores FYI |
|
|
|
|
|
2 hosts, 6 cores total that are shared over maybe 4 BOINC sites. I am NOT a heavy hitter ;-). |
|
|
|
|
I think PG should drop the CPU-based PPS and CW sieves completely.
I do not think this is a good idea. I would have taken ages to get my CW Sieve Ruby without my CPUs. Plus then you remove those people who have no GPUs ability to get any badges/credit/etc. on these projects.
Now .. let crunch some Wu's and find THE PRIME...
You will NEVER find a prime in Sieving. You just remove non-prime candidates. It takes LLR to find the prime. So GPU crunching is not finding primes in the BOINC world currently. Yes GPU crunching on PSA for Primes is happening, but that is a Staging/Test area and may take a long time before it comes to BOINC.
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
This all reminds me of a movie scene where words such as kannnnniggets and elderberries are used and a cow is catapulted over a castle wall. I think we are type-casted as the guards. All I can say is watch out...here comes the flying rabbit. :)
p.s. Please, I sure hope everyone finds the humor in this.
____________
|
|
|
|
|
|
I use CPU also, and i did find Primes!!! |
|
|
|
|
This all reminds me of a movie scene where words such as kannnnniggets and elderberries are used and a cow is catapulted over a castle wall. I think we are type-casted as the guards. All I can say is watch out...here comes the flying rabbit. :)
p.s. Please, I sure hope everyone finds the humor in this.
I didn't find this funny at all. In fact because of the lack of humor, you only get 2 instead of 3 questions. |
|
|
|
|
This all reminds me of a movie scene where words such as kannnnniggets and elderberries are used and a cow is catapulted over a castle wall. I think we are type-casted as the guards. All I can say is watch out...here comes the flying rabbit. :)
p.s. Please, I sure hope everyone finds the humor in this.
I didn't find this funny at all. In fact because of the lack of humor, you only get 2 instead of 3 questions.
Me neither. In fact I would suggest this thread be locked as people are simply kicking a dead horse.
Either you're going to keep crunching or not....pretty simple decision to me.
____________
|
|
|
|
|
This all reminds me of a movie scene where words such as kannnnniggets and elderberries are used and a cow is catapulted over a castle wall. I think we are type-casted as the guards. All I can say is watch out...here comes the flying rabbit. :)
p.s. Please, I sure hope everyone finds the humor in this.
I didn't find this funny at all. In fact because of the lack of humor, you only get 2 instead of 3 questions.
Me neither. In fact I would suggest this thread be locked as people are simply kicking a dead horse.
Either you're going to keep crunching or not....pretty simple decision to me.
hmmm kicking a dead horse... always reminded me of a c class western :)
ps John you do know I was just kidding... I actually did find the humor in your response :) |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
I didn't find this funny at all. In fact because of the lack of humor, you only get 2 instead of 3 questions.
ps John you do know I was just kidding... I actually did find the humor in your response :)
Yes...and lucky for me, the guard and bridgekeeper never meet. However, if they ever do, I think you made it easier for me to cross the bridge. My answer to the 2nd question "What is your quest" is "To seek primes!!!"
:)
____________
|
|
|
|
|
|
Primes, you say? ... all this time TheDawgz have been seeking a holey pail in which to plant shrubbery!!!
____________
There's someone in our head but it's not us. |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
Credit decrease has resumed. Please see the first post for an updated time table.
____________
|
|
|
|
|
Credit decrease has resumed. Please see the first post for an updated time table.
Hi John,
Basing credits on real world benchmarks is laudable. There has increasingly been a perhaps unfortunate effect. It seems many of the larger NVidia based producers are moving to a project that's banned by some teams and that apparently bases it's credits on the "with no apparent justification let's award credit higher than anyone else to attract users" philosophy. Not much to do about it but at least be aware of the issue... |
|
|
|
|
|
Ofcourse what else would you expect, the big bunch of top crunchers will move to the biggest rewarding project, and yes i do the same when i get much more credit on another project
So why am i doing primegrid well i find the given badges fun and they add some extra to the work done
Would i stay on the project with such low rewards, no i don't think so i get much better rewards elsewhere so thats what all projects have to deal with.
Because even though some yell that they are here for the science they jump on the credit train as fast as they can when such project appears
Its simple you do not pay the electricity bill so the mass moves where it gets the most out of the buck they spend on that bill
If teams ban a project they hurt themselfs which is my opinion not an option
I allways have been against these credit cops roaming over the projects to stop them giving the credits they think are acceptable.
There are idiot projects who pay 7 credits for 25000 seconds of cpu time ..... would you demand them to up the ante .....
And if boinc goes into assign credits on a much lower basis with set credits by the developers i think they can expect a huge loss of volunteers as well |
|
|
|
|
|
I've always thought credits should be either based on or directly proportionate (or proportionate to a reasonable degree) to FLOPs; that is, when you look at the certificate, it says "This user has contributed 43.28 quintillion floating-point operations" and that is applied to an algorithm that takes a certain number and makes it a cobblestone, ie 1 "credit" is awarded for every 1,024 FLOP, or something like that - much like how PRPNet works, actually - 1:18 and we think that's totally fair. (hint it is) - if I were any less professional I'd label the ones who will leave as "credit-something-or-others" but it seems silly AND justifiable.
Those who understand why you're reducing credits probably won't care too much. I certainly don't; I run this and DistRTGen, currently the two highest-paying projects, but only because my machine craps out on long WUs and frankly, I've never found a project better run than PrimeGrid. I'm staying here because it's been a SOLID year crunching for ya, never run into problems, and guess what, there's always work. And when it comes down to it, it's not like I'm getting paid. I also don't pay my electricity, as it's taken care of - that's a good point someone raised.
And if boinc goes into assign credits on a much lower basis with set credits by the developers i think they can expect a huge loss of volunteers as well
Despite what I said at the beginning of my post, I agree. Projects should be responsible for setting their own credits responsibly. (it's very political isn't it, ie states vs federal). I'm not saying the developers would do a poor job of standardizing credits, but it certainly would take away some of the interest and competitive aspect of distributed computing. Funny how we see it that way regardless.
____________
|
|
|
|
|
|
I'm beginning to relax now, guess the credit fever is passing. I was actually crunching gpugrid (and i was quite happy) before i was *told* to crunch primegrid. Soon i'll crunch some gpugrid again, have a 10 million point mark to break there. And then i have alot of mw, dnetc and moo! crunching to do. So much to crunch, so little to crunch on. |
|
|
|
|
|
MW is very inefficient on Nvidia cards. |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
nm
____________
|
|
|
|
|
|
For my part I forget this project.
I possess several graphs Nvidia 480 and 590 and beside spending many electricity, even above we lower the credits.
Bye. |
|
|
|
|
|
Bye. ;) |
|
|
|
|
|
Question: why challenge?
why Badges? why credit?
i am using GPU & CPU , my electric bill always rise, i did invest dollars on new Nvidia Card and to see my rac going down & also to receive mail asking me money '' donation '' ARE DONATION IS THE CRUNCH WE DO.. i now ask myself: should i move to another project? or should i resales my pc & accessories? after all i don't make money here, it is only a competition that give us the fun , but now, where is the fun????
BTW we are only the 13 october & credit will continu to go down until the end of the month.......Sad very Sad
In the futur many project will achieve their goals and some will make money out of it.. so please have little more respect for cruncher ..it is like a hockey game , one score = 1 point , they never change the rule of the game saying now for every score we will count .075 and if you hit the goaler you get .015, ... ... ...
|
|
|
|
|
|
It's for fun, science and for help discover something ( based on what project do ).
The time when you don't gain fun from it anymore, it's time to stop. |
|
|
|
|
It's for fun, science and for help discover something ( based on what project do ).
The time when you don't gain fun from it anymore, it's time to stop.
Yes, it has to be fun, cause for the cruncher there is not much else in it. |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
... & also to receive mail asking me money '' donation '' ...
This is complete fabrication. PrimeGrid does not do this. Please forward the email that you are referencing so we can investigate.
____________
|
|
|
|
|
I'm beginning to relax now, guess the credit fever is passing. I was actually crunching gpugrid (and i was quite happy) before i was *told* to crunch primegrid.
I'm sure I'm reading too much into that statement but it would never work with me to be *told* to do anything :) |
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
..."with no apparent justification let's award credit higher than anyone else to attract users" philosophy.
If this ever happens, then it will be an excellent tribute to the software developers of tpsieve...Geoff Reynolds and Ken Brazier. The code advancement was phenomenal (as quantified by the increased credit generation over the previous software). Their outstanding work revolutionized sieving, and they should be duly recognized. :)
The amount of sieving that was completed after their improvements was unimaginable over a year ago. However, since the introduction of tpsieve and its port to the GPU, decades of work have been completed in less than a year (thanks to all the volunteer crunchers). And the benefits extend beyond PrimeGrid as other prime search projects are now using these incredibly sieved files.
The prime community is forever indebted to Geoff and Ken...and to all the crunchers who made use of their software. :)
____________
|
|
|
|
|
..."with no apparent justification let's award credit higher than anyone else to attract users" philosophy.
If this ever happens, then it will be an excellent tribute to the software developers of tpsieve...Geoff Reynolds and Ken Brazier. The code advancement was phenomenal (as quantified by the increased credit generation over the previous software). Their outstanding work revolutionized sieving, and they should be duly recognized. :)
The amount of sieving that was completed after their improvements was unimaginable over a year ago. However, since the introduction of tpsieve and its port to the GPU, decades of work have been completed in less than a year (thanks to all the volunteer crunchers). And the benefits extend beyond PrimeGrid as other prime search projects are now using these incredibly sieved files.
The prime community is forever indebted to Geoff and Ken...and to all the crunchers who made use of their software. :)
Well said!
|
|
|
|
|
|
Sorry John,
Thank you for the precision...
Primegrid never ask me for donation, most of the time it is Seti Project and other.
I truly appologies...
Luc |
|
|
|
|
..."with no apparent justification let's award credit higher than anyone else to attract users" philosophy.
If this ever happens, then it will be an excellent tribute to the software developers of tpsieve...Geoff Reynolds and Ken Brazier. The code advancement was phenomenal (as quantified by the increased credit generation over the previous software). Their outstanding work revolutionized sieving, and they should be duly recognized. :)
The amount of sieving that was completed after their improvements was unimaginable over a year ago. However, since the introduction of tpsieve and its port to the GPU, decades of work have been completed in less than a year (thanks to all the volunteer crunchers). And the benefits extend beyond PrimeGrid as other prime search projects are now using these incredibly sieved files.
The prime community is forever indebted to Geoff and Ken...and to all the crunchers who made use of their software. :)
Well said!
Well said indeed. Over 15 years ago I got my BComp so I am not a total novice in programming. I thought I would have a go at modifying the TRP sieve for ATI, after an hour reading the AMD introduction to that form of programming I gave up and I still do not claim to know a thing about it.
I take my hat off to our programmers, they are top notch and we are lucky to have them.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 |
|
|
|
|
..."with no apparent justification let's award credit higher than anyone else to attract users" philosophy.
If this ever happens, then it will be an excellent tribute to the software developers of tpsieve...Geoff Reynolds and Ken Brazier. The code advancement was phenomenal (as quantified by the increased credit generation over the previous software). Their outstanding work revolutionized sieving, and they should be duly recognized. :)
The amount of sieving that was completed after their improvements was unimaginable over a year ago. However, since the introduction of tpsieve and its port to the GPU, decades of work have been completed in less than a year (thanks to all the volunteer crunchers). And the benefits extend beyond PrimeGrid as other prime search projects are now using these incredibly sieved files.
The prime community is forever indebted to Geoff and Ken...and to all the crunchers who made use of their software. :)
Well said!
Well said indeed. Over 15 years ago I got my BComp so I am not a total novice in programming. I thought I would have a go at modifying the TRP sieve for ATI, after an hour reading the AMD introduction to that form of programming I gave up and I still do not claim to know a thing about it.
I take my hat off to our programmers, they are top notch and we are lucky to have them.
Hmm, below is my post from which the snipped above was culled:
Hi John,
Basing credits on real world benchmarks is laudable. There has increasingly been a perhaps unfortunate effect. It seems many of the larger NVidia based producers are moving to a project that's banned by some teams and that apparently bases it's credits on the "with no apparent justification let's award credit higher than anyone else to attract users" philosophy. Not much to do about it but at least be aware of the issue...
Note that I was referring to another project that is basing it's credits on seemingly little other than a desire to pull users from elsewhere. I think PrimeGrid is taking the high road (like Collatz and MW) by basing credits on the amount of data being crunched. I like credits as much as anyone but think they should have some real world basis. |
|
|
|
|
|
Hello Everyone
It is very deceiving when i see my rac going from 920,000 540,000..
I bought 2 nvidia card at 300$ each to reach 920,000
Now with the 50% reducing the way it is going i will have at the end a rac of 460,000.
So for returning to a rac of 920,000 i will have to buy 2 more card???
Also i have 3 Cpu quadra core runnung with my gpu 24- 7!!!
Any idea guy's what project i should switch to..???
Regards
____________
|
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
If i remember correctly, there is no other better paying project.
On the other hand, PG is one of the most stable projects in the Boinc world...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
|
rroonnaalldd, you are right PG is one of the most stable projects, but isn't the most 'caloriest'.
DistrRTGen bestows relatively more credits for spent time (power):
4,220.00 for approx. 1600 sec = ~2.64 cr/sec @ GTX460 (810/1700)
while PG gives
2,609.00 for approx. 1400 sec = ~1.86 cr/sec @ GTX460
On the other hand, DistrRTGen is not so stable like PG
and generates slightly more data traffic.
If I remember correctly, about 5Mb per WU on input, 5Mb on output.
____________
|
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
x3mEn, thanks for pointing to. I confirm your values on my slower GTS450 eco/green-version.
DistrRTGen..2,600-3,000secs for 4220 credits
PG..............2,600secs for 2609 credits
You are right with 5MB input/output traffic for a DistrRTGen-unit. This is much more than the ~10kB for a PPSsieve-unit.
Anyway, i donated some money for PG to reward the stability and sufficiently Cuda work in the queue...
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
DoES Volunteer tester
Send message
Joined: 11 Oct 08
Posts: 784
ID: 30382
Credit: 75,064,140
RAC: 0
 
|
|
Another week and then the last calibration -- then I might get the 990 RAC pointed in the right direction --- keep up the good work PG -- we are after primes not credit --- though I will admit I had to add another 5 hosts to keep a respectable RAC showing.
____________
Member of AtP
Shown here is an Australian native rat (Ratus Kickarsus) |
|
|
|
|
Another week and then the last calibration -- then I might get the 990 RAC pointed in the right direction --- keep up the good work PG -- we are after primes not credit --- though I will admit I had to add another 5 hosts to keep a respectable RAC showing.
Hi All,
When the last step is taken we will get 2251 credit per WU.
I started crunching a few years ago and a lot of my PPS Sv work was done in my early days on PG. I have recently added a GPU and currently have 299 WU for 288276 credit, which gives 964 Cr/WU (and rising).
DoeS has crunched many more by GPU when the credit per WU was higher but still has an average of only 1984 Cr/WU. (I hope you don't mind me using you as an example).
After the last step we will both get more than we got a few years ago. I suspect that, except for recent crunchers or those who did a massive amount of work when the WUs were overpaid, this will be true for most users.
The recalibration is not so bad when looked at in this way.
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 |
|
|
DoES Volunteer tester
Send message
Joined: 11 Oct 08
Posts: 784
ID: 30382
Credit: 75,064,140
RAC: 0
|
Another week and then the last calibration -- then I might get the 990 RAC pointed in the right direction --- keep up the good work PG -- we are after primes not credit --- though I will admit I had to add another 5 hosts to keep a respectable RAC showing.
Hi All,
When the last step is taken we will get 2251 credit per WU.
I started crunching a few years ago and a lot of my PPS Sv work was done in my early days on PG. I have recently added a GPU and currently have 299 WU for 288276 credit, which gives 964 Cr/WU (and rising).
DoeS has crunched many more by GPU when the credit per WU was higher but still has an average of only 1984 Cr/WU. (I hope you don't mind me using you as an example).
After the last step we will both get more than we got a few years ago. I suspect that, except for recent crunchers or those who did a massive amount of work when the WUs were overpaid, this will be true for most users.
The recalibration is not so bad when looked at in this way.
no probs mate --- I should explain though my RAC drop is compounded by taking the 12 cpr cores off sieving on the 3rd of October leaving just the 560 gpu on sieving-- cpu's switched to PPS LLR-
looks like the cpu's were knocking out about 70K credits per day in PPS sieves so my graph looks like the side of Mt Everest (coming down) but I am finally out in the foot hills now.
____________
Member of AtP
Shown here is an Australian native rat (Ratus Kickarsus) |
|
|
BiBi Volunteer tester
Send message
Joined: 6 Mar 10
Posts: 151
ID: 56425
Credit: 34,290,031
RAC: 0
   
|
After the last step we will both get more than we got a few years ago. I suspect that, except for recent crunchers or those who did a massive amount of work when the WUs were overpaid, this will be true for most users.
The recalibration is not so bad when looked at in this way.
You are forgetting that the size of the WU was increased because of the increased speed of the GPU sieves and consequently the amount of credit per WU for the CPU. It just took a CPU more time to complete them.
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
The last credit adjustment occurred on 30 October and it is set at 2251 cobblestones per task. For now, this concludes the credit recalibration of the PPS (Sieve) project using tpsieve.
The current sieve will go deeper than any sieve to date. Should anything change in the speed of tpsieve, we'll be certain to notify everyone of any future adjustments.
____________
|
|
|
|
|
|
Been runnin cullen woodall cuda wu-s for a while. Time credit ratio for gts 450 is 1,1755 credit per sec. Pps sieve cuda ratio is 1,1288 credit per sec. The gts is oc-d. Wonder what results other crunchers get. |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
I ran a few of the CW CUDA Wu's on my 470 Box & I show 3.050 Credits Per Second for the PPS's CUDA Wu's & 2.450 Creidts Per Second for the CW CUDA Wu's. So the PPS CUDA Wu's even with the Credits cut in half still give better Credit than the CW CUDA Wu's do, at least for me anyway ... :)
____________
|
|
|
|
|
|
I have a 570 and the PPS Sieve units are, as of today, 24% more credit-productive per second than CW Sieve. I run both just for variety.
--Gary
Go AtP!
____________
"I am he as you are he as you are me and we are all together"
87*2^3496188+1 is prime! (1052460 digits)
4 is not prime! (1 digit) |
|
|
rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
I get 3160 credits per hour for PPS Cuda (2251 per unit in the moment) and 2916 credits per hour for CW Cuda on my GT 450 "green version".
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
errrrrrrrrrrr, any reason for the sudden 50% Credit Increase from 2,251 to 3,371 I seen overnight ... :0
____________
|
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
errrrrrrrrrrr, any reason for the sudden 50% Credit Increase from 2,251 to 3,371 I seen overnight ... :0
The WU size was increased by 50% to make it easier on the servers during the upcoming Winter Solstice Challenge. Credit per WU was therefore increase by the same amount.
There's an announcement about this around here somewhere.
If the WUs are the same size, it's probably still related to the WU change somehow.
____________
My lucky number is 75898524288+1 |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
So far I haven't seen an increase in Wu Length, but that could change over the next hours as I carry a full Cache which takes a minimum of about 12 hours to run out or before I would see any change in length ...
____________
|
|
|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
errrrrrrrrrrr, any reason for the sudden 50% Credit Increase from 2,251 to 3,371 I seen overnight ... :0
Yes, as Michael mentions, the WU's have increased by 50% from 6G's to 9G's and credit was adjusted accordingly. Post is here. It looks like we should have waited a little longer in updating credit to allow for the buffer to clear out. ;)
NOTE: This is not a recalibration...but only an adjustment based on increased WU length.
____________
|
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
|
Okay & Thanks John, thought that might be the reason myself when I seen the Increase ...
____________
|
|
|
|
|
|
I'm confused, my WUs are still completing in c. 11mins (GTX570) but I'm getting the 50% increase in credits. Nice but unexpected.
____________
35 x 2^3587843+1 is prime! |
|
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14665
ID: 53948
Credit: 1,033,756,273
RAC: 402,691
|
|
It seems they can't tie the credit to specific WUs, so they have two steps to do:
1) Feed bigger WUs into the system.
2) Increase the credit per WU.
I guess it's not possible to get the two perfectly in sync (or what's necessary to do so may be undesirable), so their options are to either increase the WU size first or increase the credit first.
If they increase the WU size first, some people will do 50% more crunching for the same credits as before, and be unhappy.
If they increase the credits first, some people will do the same crunching for 50% more credits, and be confused (but not unhappy).
Sounds to me like they used plan B. :)
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
I would like to complain about excessive credits ....
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
If they increase the credits first, some people will do the same crunching for 50% more credits, and be confused (but not unhappy).
DaveB is Unhappy, lol
____________
|
|
|
|
|
I would like to complain about excessive credits ....
LOL in that case might I suggest you stop running those excessive credit :) - HA
____________
@AggieThePew
 |
|
|
|
|
|
Keep those shorties coming! |
|
|
|
|
Keep those shorties coming!
well I think the shorties have run out... i'm getting the new ones now.. alls well that ends well
____________
@AggieThePew
|
|
|
|
|
|
...sigh... |
|
|
STE\/E Volunteer tester
Send message
Joined: 10 Aug 05
Posts: 578
ID: 103
Credit: 3,964,984,134
RAC: 237,743
|
If they increase the credits first, some people will do the same crunching for 50% more credits, and be confused (but not unhappy).
DaveB is Unhappy, lol
Dave's Happy again ... :P
____________
|
|
|
|
|
|
Hi Steve
credits and runtimes are bouncing around all over today.
Yesterday same run time bigger credit.
Before PM here today longer runtime [40%] for same credit ie 3371
Now back to shorter run time with 3371 credit.
Hope they sort it out before the Challenge starts.
Also increase in Temps on the longer run WUs
Hope you are keeping well.
Cheers
Ross*
____________
|
|
|
|
|
|
Now we have longer WU's, average precalculated duration 17min 13sec on my 470/570 host
hostid=173588
heinz
|
|
|
