| Author |
Message |
|---|
John Honorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
                 
|
|
April's PRPNet Project
The dual Sierpinski problem (Five or Bust) is PrimeGrid's PRPNet project of the month for April. The goal will be to complete the entire double check effort before their PROJECT COMPLETION CELEBRATION on 7 May 2011. With the double check complete, the Sloane sequence A067760, a(n) = "least positive k such that (2n+1)+2^k is prime, or 0 if no such k exists", will be established presuming that the 20 identified probable primes are actually prime.
To participate in wrapping this project up, please add the following server to your prpclient.ini file:
server=5OB:100:1:pgllr.mine.nu:13000
Stats
The project will run from 1 April to 30 April. It will be a great opportunity to be part of closing out the dual Sierpinski problem. Maybe its luck will rub off on some of the other conjectures PrimeGrid is participating in. :)
New to PRPNet? Please see Welcome to PRPNet
Testing times are currently ~30-35 minutes but will increase to ~45 hours by the end of the effort.
EDIT: Testing times as of this edit (25 Mar 2011 | 16:18:26 UTC) are ~1 hour now.
EDIT: Testing times as of this edit (28 Mar 2011 | 1:56:27 UTC) are ~2 hours now.
EDIT: Testing times as of this edit (4 Apr 2011 | 20:38:10 UTC) are ~2:45 hours now.
EDIT: Testing times as of this edit (10 Apr 2011 | 1:11:09 UTC) are ~4-6 hours for faster hosts and ~8-10 hours for slower hosts.
EDIT: Testing times as of this edit (18 Apr 2011 | 21:00:16 UTC) are ~7-9 hours for faster hosts and ~11-13 hours for slower hosts.
EDIT: Testing times as of this edit (27 Apr 2011 | 1:35:32 UTC) are ~9-11 hours for faster hosts and ~15-17 hours for slower hosts.
Background
Many may not realize that back in January a collaboration was formed between PrimeGrid and 5oB. By February, Mark Rodenkirch had already modified PRPNet to accept the new form and testing began in the PST forum. However, a few days before its official announcement in the PG forum, 5oB discovered the PRP to its final candidate.
This was excellent news and a welcomed completion to one of the Sierpinski conjectures. Now all that remains is to confirm that the found PRP's are the smallest PRP for each k. This requires a double check of all candidates below the found PRP's. Once this is complete, Sloane sequence A067760 can be established.
The participants of PrimeGrid have accomplished incredible achievements while at the same time contributing to the entire prime finding community. Along with project collaborations (PSP, SoB, SR5), PrimeGrid also provides sieve support for several riesel-side projects. The generosity of this sieve file is monumental as it frees up the other projects to focus strictly on prime finding.
Once again, PrimeGrid would like to extend its support to help another project. However, this time it is to facilitate the ending to a successful project.
Thank you for your consideration!
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
                   
|
|
PSA credit?
____________
Murphy (AtP)
 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
Good progress has been made in the run up to April. Testing times have increased from 45 minutes at the start to now approximately 2 hours long. They will continue to increase as n increases.
The project will run from 1 April to 30 April. Feel free to join in after the Marie-Sophie Germain Challenge (1-3 April) completes. :) It will be a great opportunity to be part of closing out the dual Sierpinski problem. Maybe its luck will rub off on some of the other conjectures PrimeGrid is participating in. :)
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
____________
|
|
|
 Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1224
ID: 18646
Credit: 877,860,382
RAC: 320,481
                      
|
|
How about a team challenge?
or
Who will reach PrimeSearchTeam?
have a look: stats
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
How about a team challenge?
or
Who will reach PrimeSearchTeam?
have a look: stats
I'm open to any suggestions. There's a nice increase in port activity after the Challenge, but we'll need a lot more users to stand any chance of reaching the goal.
At this stage in the testing, I reckon we need about 120 cores active. This number will only increase as testing times continue to increase. As of this post, a WU is taking about 2 hr 45 minutes to complete. At the end of the search, testing time will take about 48 hours. Therefore, we need to tackle as much of this as possible early on.
120 users donating 1 core
60 users donating 2 cores
40 users donating 3 cores
30 users donating 4 cores
20 users donating 6 cores
15 users donating 8 cores Please pass this along to your team forums if you're a member of a team. We need all the help we can get. :) PSA credit is available.
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username
clientid=clientID
teamid=
OS, 32/64bit, and how many cores you wish to run.
____________
|
|
|
|
|
|
Phil Moore here - I hope to submit a short paper on this project after this is all wrapped up, and I would welcome names of any contributors who would not mind being thanked for your work to include in the paper. Just send me a private message with your first and last names, and I will start a list. We already have about two dozen names so far from the work done on Mersenneforum, and it would be nice to add at least that many more! |
|
|
|
|
|
I've just attached my two cores to the project. I intend to leave them here until completion. I'm happy to help with stuff like this and I really ought to go back to the SR5 project once this is done.
Also: The Time per Bit for this project is about 6-8 ms for me in a 2.4 GHz C2D running Linux 64. Time per Bit on SR5 and ports 10K & 11K is much lower for me (max 2 ms per bit). Is this normal for this project?
____________
 |
|
|
|
|
I've just attached my two cores to the project. I intend to leave them here until completion. I'm happy to help with stuff like this and I really ought to go back to the SR5 project once this is done.
Also: The Time per Bit for this project is about 6-8 ms for me in a 2.4 GHz C2D running Linux 64. Time per Bit on SR5 and ports 10K & 11K is much lower for me (max 2 ms per bit). Is this normal for this project?
Times per bit increase for larger numbers when the FFT lengths increase. I'm not sure how much the k values affect the runtimes.
____________
|
|
|
|
|
|
One of my tests took five hours, the other six hours, I was under the impression that they were supposed to be about three hours long right now, so either a lot of progress has been made very recently or my computer is just older and slower than I thought it was.
____________
|
|
|
|
|
I've just attached my two cores to the project. I intend to leave them here until completion. I'm happy to help with stuff like this and I really ought to go back to the SR5 project once this is done.
Also: The Time per Bit for this project is about 6-8 ms for me in a 2.4 GHz C2D running Linux 64. Time per Bit on SR5 and ports 10K & 11K is much lower for me (max 2 ms per bit). Is this normal for this project?
Times per bit increase for larger numbers when the FFT lengths increase. I'm not sure how much the k values affect the runtimes.
John updates the very first post with approximate run times probably an average of hosts used. Ignore the jump in dates because most people left for the SGS challenge.
Currently my 3GHz hosts are around 4 to 5 ms and the most recent tasks took about 4 hours. Just a hint at the increasing time required for the last 5 tasks from the same core (multiple cores but chip doesn't support hyperthreads) as the current one is between 5 to 6 ms per bit:
Starting probable prime test of 2^2622520+40291
2^2622520+40291 is not prime. RES64: 10B76D9BBBA57EB5. OLD64: 322648D332EF4156 Time : 16894.813 sec.
Starting probable prime test of 2^2627272+40291
2^2627272+40291 is not prime. RES64: A8C613F373A4AC63. OLD64: FA523BDA5AED67C3 Time : 17930.299 sec.
Starting probable prime test of 2^2633224+40291
2^2633224+40291 is not prime. RES64: FE10E9FE5AF8AB51. OLD64: FA32BDFB10E9648D Time : 19137.766 sec.
Starting probable prime test of 2^2643604+2131
2^2643604+2131 is not prime. RES64: 1928E9E8EDF78271. OLD64: 4B7ABDBAC9E676AA Time : 17978.307 sec.
Starting probable prime test of 2^2656895+41693
2^2656895+41693 is not prime. RES64: C8F6A1E365C5FF6E. OLD64: 5AE3E5AA3150B88D Time : 18500.919 sec.
|
|
|
|
|
|
With my two beasts, the Phenom2 quad @3.3Ghz is doing them in around 13,600secs, and my 1090T @3.7Ghz is doing them in around 13,200secs.
I suspect the original suggested average timing looks to be affected by i7's in the early stages of fewer volunteers crunching, with a higher proportion of that number having i7's (maybe) as 2 3/4 hours sounds right for an i7 when set against my 1090T times.
EDIT:
Had three go through the 1090T at 12,200secs so the times do change, dont know if the the sizes have started increasing yet. I have MW GPU WUs running on both, dont know if that affects the 5oB WUs much or not - might be a factor.
Regards
Zy |
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
|
My Core2 Duo (1.73 GHz) is taking almost 8 hours per WU.
Just started one on the (2.00 GHz) machine and it looks to be ~ 6.5 hours.
____________
Murphy (AtP)
|
|
|
|
|
|
Thanks, knowing this means that my times probably are normal.
____________
|
|
|
Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1224
ID: 18646
Credit: 877,860,382
RAC: 320,481
|
|
My Phenom at 2.2 GHz :
date | runtime
04/04 | 15.9k seconds | 4.4h
04/05 | 19.6k seconds | 5.4h
04/06 | 20.6k seconds | 5.7h
04/07 | 21.2k seconds | 5.8h
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
A third of the way through the month and steady progress continues. Testing times are currently ~4-6 hours for faster hosts and ~8-10 hours for slower hosts. They will continue to increase as n increases.
There's been a change of plans. We received the final installment of double check work for the final k. There were an additional ~9000 candidates in this batch. :( The current goal was quite a stretch as it was but adding this additional work makes it highly improbable.
Therefore, the goal has been updated to the following:
- Complete double check work of 2131 to n=4582936
- Complete double check work of 41693 to n=5146239
- Complete double check work of 40291 up to n=6M
Even these goals are quite challenging...and I've been told, simply not possible by 5 May 2011. It sure would be nice to prove the naysayers wrong. ;)
Thank you for the help and please spread the word to keep the support coming. There are ~40 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.[/quote]
____________
|
|
|
|
|
|
The most disappointing thing about this project is zero results
Primes Found by User
No primes found
Primes Found by Team
No primes found
GFN Divisors Found
No GFN divisors found
____________
wbr, Me. Dead J. Dona
 |
|
|
|
|
|
Just wait. Someone will end up spending 45 hours on a work unit and find a monster prime. And I'm willing to bet it won't be me. ;)
____________
FUCK THIS PROJECT AND THEIR COMMIE BULLSHIT POLITICS |
|
|
|
|
Just wait. Someone will end up spending 45 hours on a work unit and find a monster prime. And I'm willing to bet it won't be me. ;)
I thought this was a double check sub-project and finding a prime would be a disaster?
____________
35 x 2^3587843+1 is prime! |
|
|
|
|
|
I was under the impression that we were trying to find that 78557 is the smallest k such that 2^n+k is composite for all n, so conceivable we are finding a prime eventually for each of the k we are testing, I was just under the impression that we were double checking everything to see if we found the smallest primes for each k and then double checking the actual prime that was already found.
____________
|
|
|
|
|
|
Every k<78557 has already had a prime (technically a probable prime) found. This PRPNet project is just to double-check that the n's found for last three k's are actually the smallest n to make k+2^n prime (All other k's have already been fully double-checked).
The actual PRP's found have already been double-checked many times using several different methods, so we are quite confident that they are indeed PRP's, and almost certainly prime (but we have no way of actually testing the largest of them for primality with current methods & technology)
It would not be a disaster to find a smaller PRP for one of these k, it would just mean that there was a lot of unnecessary work done. Although it is highly unlikely, the Seventeen or Bust project has already proven that primes can be found via double-checking.
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
HALF way through the month and steady progress continues. Testing times are currently ~5-6 hours for faster hosts and ~9-10 hours for slower hosts. They will continue to increase as n increases.
We are definitely making the naysayers sweat a little. The first goal is definitely within reach. A HUGE thanks to everyone who has answered the call for help. Let's keep the pressure up!!!
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936
- Complete double check work of 41693 to n=5146239
- Complete double check work of 40291 up to n=6M
Please keep spreading the word. We still need more support to reach the other two goals. There are ~55 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.[/quote]
____________
|
|
|
|
|
|
BIG Thank you to John for helping get PRPNet working, had a few Linux issues that he guided me through and I am now working. He is a great admin person.
Also thank you to Steve Martin who also gave guidance in this issue to me.
I have added 2 computers and 10 cores to the 5 or Bust problem, I will get a PSA badge yet..
Conan
____________
 |
|
|
|
|
|
From a 2.66Ghz core of an Intel Xeon X5355
2^3860692+2131 is not prime. RES64: D67CC082D2D8B6B1. OLD64: 83764188788A1BBD Time : 32423.231 sec.
[2011-04-17 00:36:47 CDT] 5oB: 1*2^3860692+2131 is not prime. Residue D67CC082D2D8B6B1.
[2011-04-17 00:36:47 CDT] Total Time: 9:00:27 Total Tests: 1 Total PRPs Found: 0
64-bit linux |
|
|
|
|
|
From a 3.2Ghz AMD Phenom II X6 1090T (linux 64):
2^3899080+40291 is not prime. RES64: 1812ABAC851C1525. OLD64: 483803058F5304A6 Time : 26293.885 sec.
[2011-04-17 08:11:24 CDT] 5oB: 1*2^3899080+40291 is not prime. Residue 1812ABAC851C1525.
[2011-04-17 08:11:25 CDT] Total Time: 7:18:19 Total Tests: 1 Total PRPs Found: 0 |
|
|
|
|
|
I wonder if the drive to get better badges on the 321 sieve is diverting resources from this project...
____________
|
|
|
|
|
|
That is possible for some, as I had gained my Silver and no real chance of getting a Gold before 22nd I switched to this project and joined PRPNet for the very first time.
I had not run this project before so I am a new one that has not crunched any PRPNet work before.
My 2 computers and 10 cores have already completed 53 tests since Saturday night so I am doing my bit.
So far I am enjoying the experience and would be very happy if I came across a Prime of any sort as I only have one double checker to my name.
The tests are getting longer, from 5 to 6 hours they are now taking over 8.5 hours and increasing in time a lot quicker now.
Conan
____________
|
|
|
|
|
|
I'm not just watching the times increase but watching this status. 1*2^n+40291 with max N=9091912 looks fairly daunting. |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1957
ID: 352
Credit: 6,139,270,864
RAC: 2,271,114
                                     
|
|
Yes, now with bleeding edge ~4M and FFT=448K, it's getting tough.
The good thing is that 1*2^n+2131 goes only to 4,58M and 1*2^n+41693 to 5,14M.
It leaves 1*2^n+40291 to 9,09M, we can expect times similar to Cullen Prime Search (?)
And this range has more candidates in upper range...
____________
My stats |
|
|
|
|
|
Strange. I have set teamid=SETI.Germany in prpclient.ini, but in the stats I show up as teamless.
____________
There are only 10 kinds of people - those who understand binary and those who don't
 |
|
|
|
|
|
The instructions say you must replace spaces with underscores. |
|
|
|
|
|
Please forgive my ignorance as this is my first time using PRPNet but on the pending tests page there appear to be outstanding work units from up to 4 days ago. What sort of timeout and recycle/retry measures are in place? It seems particularly important given that these challenges run for a matter of a few weeks. |
|
|
|
|
|
I believe for this one, 96hrs is timeout & recycled, as you will not see one on the list longer than 96hrs. It needs to be a long timeout to cater for slower machines, especially during the latter stages when normal crunch times get long.
I picked a short duration one this morning (around 6hrs to finish estimate, current crunch time for me is around 11hrs), so they do come through recycled.
Regards
Zy |
|
|
Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1224
ID: 18646
Credit: 877,860,382
RAC: 320,481
|
Strange. I have set teamid=SETI.Germany in prpclient.ini, but in the stats I show up as teamless.
perhaps an older PRPNet-client, which doesnt support teamid yet??
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
|
|
|
|
|
|
Ah, you raise a valid point as the post indicates that the work is expected to reach 45 hours for a single unit towards the end. Though, that raises the other problem that we're in for a lot more resends at the end. |
|
|
|
|
|
Shouldnt be too bad. There are around 100 floating at the moment, but most look like machine setup/start problems looking at the level of Machine & Cruncher. If there are - say - 50 left floating at the end, shouldnt take long with 300+ Cores to knock them off.
Regards
Zy |
|
|
|
|
|
Might not help a lot but I finally got 2 cores up and running over here.
____________
@AggieThePew
 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Might not help a lot but I finally got 2 cores up and running over here.
Every bit helps!!!
The tests are getting longer, from 5 to 6 hours they are now taking over 8.5 hours and increasing in time a lot quicker now.
Yes, as Honza pointed out we have entered the next FFT length, thus increasing the testing time. Times are now 7-9 hours for faster cores and 11-13 hours for slower cores. As mentioned in the first post, we can expect times to increase to 45 hours by the end which is k=40291 n=9091912.
As we enter this slowdown, the first goal is approaching (Complete double check work of 2131 to n=4582936). Please continue to recruit new participants. We are still shooting for that 120 active user level.
Good news, once the first goal is reached, we'll add 1/3 extra speed to the remaining goals. And once the second goal is reached, we'll double the speed to the last goal. Unfortunately, we'll hit another FFT length by that time so it won't quite be the doubled rate.
Thank you to everyone participating!!!
EDIT: BTW, yesterday, we passed the half way point overall for the double check effort (currently 22531 completed out of 43296 total candidates). Unfortunately, it was the easier half. :D
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
EDIT: BTW, yesterday, we passed the half way point overall for the double check effort (currently 22531 completed out of 43296 total candidates). Unfortunately, it was the easier half. :D
Reminds me of the old management axiom:
The first 90% of the project takes 90% of the time, and the last 10% of the project takes the last 90% of the time!
____________
Murphy (AtP)
|
|
|
|
|
|
When I said I moved 2 cores... well those 2 took 15:22 to run a wu each. I also moved 2 more from from one other system and those took over 9 each so that's running them much faster. Still not contributing much to the relief effort. |
|
|
|
|
|
To help the effort I have just downloaded the PPRNet Client to my AMD X4 955 Windows machine, aborted my other PrimeGrid CPU jobs and I am into 5oB.
So my core total is now 14 on 5oB.
As I am also running Proth Prime Sieve on my graphics cards 5oB runs a lot slower than on my Linux machines but will still be a lot faster than my Windows based Opteron 285 (around twice as fast), so I wont be adding that one.
3 computers will have to do, that is all I can spare.
Since late Saturday when I joined PRPNet, I have now done around 120 (if I haven't I soon will) Test Results and shot through 1.5 million PRPNet credits.
Will soon be worth equal to 100,000 Boinc credits, so a Bronze badge already before I even get a Badge awarded and soon to be Silver. So I might just skip right past Bronze.
Conan.
____________
|
|
|
NeoVolunteer tester
Send message
Joined: 28 Oct 10
Posts: 710
ID: 71509
Credit: 91,178,992
RAC: 0
            
|
|
I just wanted to say that the stats page for 5oB is totally awesome.
Neo
____________
 |
|
|
|
|
|
Just noticed a curious thing.
I have 14 cores now working on this project, however when I counted my running jobs I could only find 13.
So I wrote down all running jobs to find the one not listed.
I found the missing WU but it is showing up as being processed by a different person.
I am processing 1*2^4164376+40291 and will finish it in the next half hour but it showing on the job list as being processed by "Wolley68" on computer "clientID" and a member of "CzechNationalTeam" issued at 06:00:35 on Wed 20 2011.
How can that be?
The job is being processed by Conan on computer AMDX4955 and member of team Cobar_Spiders.
I will check to see if I get the credit or not.
Conan
____________
|
|
|
|
|
|
Well the Work Unit finished on my computer after 34,701.350 seconds.
I did not get any credit for this work unit as it says the WU was ignored and the WU could not be found.
I would like my 15,000 or so credit as I have worked on this WU and found it to NOT be Prime after 9.6 hours.
I took a screen shot but not sure how to place in the forum.
Disappointing
Conan
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
...
Will soon be worth equal to 100,000 Boinc credits, so a Bronze badge already before I even get a Badge awarded and soon to be Silver. So I might just skip right past Bronze.
Conan.
Due to the "trickle effect" of converting/awarding points to Boinc, you will have Bronze for quite a few days before it upgrades to Silver (unless, of course, your current RAC is considerably over 200,000 Boinc). PSA Silver = 200,000 Boinc = (approx.) 3,600,000 PRPNet.
____________
Murphy (AtP)
|
|
|
|
|
|
Well my computer has completed the WU and returned it or at least tried.
I did not get any credited test result or and credit.
My test was ignored plus No candidate could be found or the Test could not be found.
The Test ran for 34,701.350 seconds
I would like my 15,000 or so credits as I processed this WU for over 9.6 hours and found it to be NOT Prime.
How can I be given a work unit and also someone else be given the same WU ??
Very disappointing and a lot of wasted effort.
Conan.
EDIT: Bugger I thought this post had not gone through so I posted it again, Sorry about the double post.
____________
|
|
|
|
|
|
Conan, same thing happened to me too.
One of my cores was assigned a wu that was showing up as being processed by a different person (sent out a few min earlier). I aborted my wu and was assigned a new one, which was listed properly.
Just thought I'd share... I wonder if this is happening more frequently?
It was just by chance that I individually checked the wu's assigned to the detailed stats page. |
|
|
|
|
Conan, same thing happened to me too.
One of my cores was assigned a wu that was showing up as being processed by a different person (sent out a few min earlier). I aborted my wu and was assigned a new one, which was listed properly.
Just thought I'd share... I wonder if this is happening more frequently?
It was just by chance that I individually checked the wu's assigned to the detailed stats page.
I've noticed, when I check for in-progess times on the various cores that some are not showing up in the listing. They appear to crunch normally and when they cycle through appear to be accepted when they finish.
I just figure that the next one will be OK even if this one doesn't take.
____________
 |
|
|
|
|
Conan, same thing happened to me too.
One of my cores was assigned a wu that was showing up as being processed by a different person (sent out a few min earlier). I aborted my wu and was assigned a new one, which was listed properly.
Just thought I'd share... I wonder if this is happening more frequently?
It was just by chance that I individually checked the wu's assigned to the detailed stats page.
Thanks pabliedung,
How do you abort a WU?
On Linux ?
On Windows ?
I have not seen anything about this.
Mine was 96% done at the time I found it so it was a bit late by then.
Conan
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
How do you abort a WU?
On Linux ?
On Windows ?
I have not seen anything about this.
Mine was 96% done at the time I found it so it was a bit late by then.
Conan
Control-C to shut down.
Edit prpclient.ini
// This option is used to default the startup option if the PREVIOUS
// SHUTDOWN LEFT UNCOMPLETED WORKUNITS. If no previous work was left
// this will act like option 9.
// 0 - prompt
// 1 - Return completed work units, abandon the rest, then get more work
// 2 - Return completed work units, abandon the rest, then shut down
// 3 - Return completed, then continue
// 4 - Complete in-progress work units, abandon the rest, then get more work
// 5 - Complete in-progress work units, abandon the rest, then shut down
// 6 - Complete all work units, report them, then shut down
// 9 - Continue from where client left off when it was shut down
startoption=2
Start prpclient.exe
WU aborts ...
Re-edit prpclient.ini
(return to original "3")
Start prpclient.exe
(off and running with new WU)
____________
Murphy (AtP)
|
|
|
|
|
|
@Conan,
That sounds like the problem discussed in the Mystery Destination thread. For some reason, the server is not aware that it has given you the workunit so it goes to someone else, and you won't get credit for it if you let it finish. It doesn't happen often, but it is still happening.
In the future, if you catch one of these just Ctl-C it, delete the data files from the folder, and restart. It won't cause an abandoned WU since the server doesn't have it assigned to you.
|
|
|
|
|
|
Hm..., mfbabb2, where are you found this wide description?
My prpclient.ini has only the next:
// This option is used to default the startup option if the previous
// shutdown left uncompleted workunits.
// 0 - prompt
// 1 - Return completed work units and abandon the rest
// 2 - Complete assigned work units
startoption=2
// This option is used to default the stop option when the client
// is terminated
// 0 - prompt
// 1 - Return completed work units and abandon the rest
// 2 - Return completed work units
// 3 - Do nothing with current work units and terminate the process
stopoption=3
EDIT: Now I understand, why my prpclient
1) needs internet after restart
2) closes first time after shutdown
Fixed.
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Conan, same thing happened to me too.
One of my cores was assigned a wu that was showing up as being processed by a different person (sent out a few min earlier). I aborted my wu and was assigned a new one, which was listed properly.
We still have yet to find what's causing this. 4.3.1 and 4.3.2 have been released, but I don't believe either addressed this issue.
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
9 days remain till the end of the month...plus the first 5 days of May. Testing times are currently ~7-9 hours for faster hosts and ~12-14 hours for slower hosts. They will continue to increase as n increases.
~1000 WU's remain to reach the first goal and ~3000 WU's to reach the second goal. After 7 straight days of daily decline, yesterday we had an increase. Thank you to everyone who added more cores.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936
- Complete double check work of 41693 to n=5146239
- Complete double check work of 40291 up to n=6M
Please keep spreading the word. There are ~59 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.[/quote]
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
Hm..., mfbabb2, where are you found this wide description?
My prpclient.ini has only the next:
// This option is used to default the startup option if the previous
// shutdown left uncompleted workunits.
// 0 - prompt
// 1 - Return completed work units and abandon the rest
// 2 - Complete assigned work units
startoption=2
// This option is used to default the stop option when the client
// is terminated
// 0 - prompt
// 1 - Return completed work units and abandon the rest
// 2 - Return completed work units
// 3 - Do nothing with current work units and terminate the process
stopoption=3
EDIT: Now I understand, why my prpclient
1) needs internet after restart
2) closes first time after shutdown
Fixed.
I am running 4.3.0beta; it added additional (and very useful) options.
____________
Murphy (AtP)
|
|
|
|
|
We still have yet to find what's causing this. 4.3.1 and 4.3.2 have been released, but I don't believe either addressed this issue.
Didn't know 3.1 or 3.2 were available... now that I do, I might just wait until another version comes out :)
Rick
____________
@AggieThePew
|
|
|
NeoVolunteer tester
Send message
Joined: 28 Oct 10
Posts: 710
ID: 71509
Credit: 91,178,992
RAC: 0
|
|
Just completed my first two 5oB w/u's.
Can anyone tell me why the RES64 residue is different than the OLD64 residue? I thought they had to match up?
Neo
____________
|
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
We still have yet to find what's causing this. 4.3.1 and 4.3.2 have been released, but I don't believe either addressed this issue.
Didn't know 3.1 or 3.2 were available... now that I do, I might just wait until another version comes out :)
Rick
I suggest using a 4.2 client. The protocol was changed between 4.2 and 4.3 and I suspect the new protocol is related to the issue. The 4.3 server will use the 4.2 protocol with 4.2 clients. |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
Just completed my first two 5oB w/u's.
Can anyone tell me why the RES64 residue is different than the OLD64 residue? I thought they had to match up?
Neo
The way the residue is derived was changed, thus they are different. RES64 should match for both PFGW and LLR. |
|
|
|
|
|
The daily decline on WUs per day completed was reversed for 3 days but the larger N just need more cores.
From a 3.2Ghz AMD Phenom II X6 1090T (x86_64 linux):
2011-04-17 2^3899080+40291 07:18:14
2011-04-23 2^4474648+40291 10:30:30
From a 2.66Ghz Intel Xeon X5355 (x86_64 linux):
2011-04-17 2^3857212+2131 11:18:59
2011-04-23 2^4487896+2131 15:02:13 |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
Nice progress since the last update. 6 days remain till the end of the month...plus the first 5 days of May. Testing times are currently ~8-10 hours for faster hosts and ~13-15 hours for slower hosts. They will continue to increase as n increases.
All WU's for k=2131 have been sent out!!! :) ~140 WU's remain "in progress" to reach that first goal. Hopefully the second goal will follow soon now that there's only two k's left. However, testing times jumped up another hour. Also, the search took a hit yesterday. As quel mentioned earlier, the daily trend was reversed for 3 days but now it's on the decline again. Looks like it's going to be tough to maintain the daily rate with the increasing testing times.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - all work in progress
- Complete double check work of 41693 to n=5146239 - ~1900 WU's outstanding
- Complete double check work of 40291 up to n=6M
Please keep spreading the word. There are ~60 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.[/quote]
____________
|
|
|
|
|
|
+2 cores.
Expect the first results in 8 hours.
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
Thank you to everyone who continues to up the ante. The addition of two users today has helped recover some of the hit we took two days ago. 5 days remain till the end of the month...plus the first 5 days of May.
Only 3 WU's for k=2131 remain "in progress". The approach to the second goal shows signs of improvement now that there's only two k's remaining. If there's anyone left on the sidelines still watching this, please consider donating 24 to 48 hours in the next 5 days.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - 3 WU's remain in progress
- Complete double check work of 41693 to n=5146239 - ~1500 WU's outstanding
- Complete double check work of 40291 up to n=6M - ~3800 WU's outstanding to 6M, 13500 WU's outstanding to completion
Please keep spreading the word. There are ~61 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.[/quote]
____________
|
|
|
|
|
|
OK
as I just found one more SGS 2h ago
will paticipate with 2 cores
greetings ,, parabol |
|
|
|
|
April's PRPNet Project - 5oB Update
If there's anyone left on the sidelines still watching this, please consider donating 24 to 48 hours in the next 5 days.
OK! OK! Enough with the rolled up newspaper already! ... you can have TheDawgz 4 cores until the 5th ... a couple of jerky treats would have got you the same result!
____________
There's someone in our head but it's not us. |
|
|
|
|
|
Sorry but I am pulling my 14 cores out.
Have done 277 tests, I have achieved the goal of getting a PSA badge and with 4.5 million PSA points I have enough for a Silver badge.
Been going since the 16th so have given hundreds of hours.
Need to work on a few other PrimeGrid sub-projects before they run out, so switching to PSP, SGS and will do a few 321 Prime Sieves even though I need 1,000 of them to get Gold, which wont happen any time soon.
Have helped to the first goal, good luck with the next two goals.
Conan
____________
|
|
|
|
|
|
Don't forget, I am still collecting names to acknowlege when a paper is submitted on this problem, so if you don't mind your name being among those credited with helping solve this problem, pm me with your first and last names! And thanks! |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
OK! OK! Enough with the rolled up newspaper already! ... you can have TheDawgz 4 cores until the 5th ... a couple of jerky treats would have got you the same result!
:D Jerky treats...good to know. ;) I'll go stock up on them now. Anyone else have a carrot they'd like to receive?
SPECIAL NOTE TO EVERYONE: Please reduce your cache to just 1 WU per core. Testing times remain ~8-10 hours for faster hosts and ~13-15 hours for slower hosts. However, they will continue to increase as n increases. Having a 1 WU cache is now recommended.
server=5oB:100:1:pgllr.mine.nu:13000
The reason for this request is that we've dropped the deadline to 72 hours.
Thanks again to everyone helping out.
____________
|
|
|
|
|
....SPECIAL NOTE TO EVERYONE: Please reduce your cache to just 1 WU per core. Testing times remain ~8-10 hours for faster hosts and ~13-15 hours for slower hosts. However, they will continue to increase as n increases. Having a 1 WU cache is now recommended.
server=5oB:100:1:pgllr.mine.nu:13000
The reason for this request is that we've dropped the deadline to 72 hours.
Thanks again to everyone helping out.
Typical, I reduced cache yesterday, and the Router went down for 16 hours - Murphy is Toast if I ever meet up with the pesky Gremlin :)
........ Jerky treats...good to know. ;) I'll go stock up on them now. Anyone else have a carrot they'd like to receive? ...
£0.5 a litre petrol subsidy for the duration ?? {ahhh one can only dream - would solve the problem of number of crunching Cores though ... :) }
Regards
Zy |
|
|
|
|
Anyone else have a carrot they'd like to receive?
A carrot badge for the participants could lure in a few additional badge hunters ;)
____________
|
|
|
|
|
Anyone else have a carrot they'd like to receive?
A carrot badge for the participants could lure in a few additional badge hunters ;)
Thats a good one ..... I could see a number of Vegetarian converts coming PG's way :)
Regards
Zy |
|
|
|
|
Anyone else have a carrot they'd like to receive?
A carrot badge for the participants could lure in a few additional badge hunters ;)
Thats a good one ..... I could see a number of Vegetarian converts coming PG's way :)
Regards
Zy
Lennart already got his carrot badge. Just take a look at his profile picture :)
____________
|
|
|
|
|
Anyone else have a carrot they'd like to receive?
A carrot badge for the participants could lure in a few additional badge hunters ;)
+1
Or some more general broom badge for helping with this type of cleanup or tasks like the past PPS LLR assistances? It could also be extended to challenge cleanups as well to encourage people to finish their WUs. [/i] |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
3 days remain till the end of the month. Yesterday was the lowest output in two weeks even with the addition of 5 new users. Thank you to everyone who continues to trudge through these WU's. The decline just goes to show you how long these WU's are becoming. Testing times are currently ~9-11 hours for faster hosts and ~15-17 hours for slower hosts. They will continue to increase as n increases.
NOTE: These are very challenging goals we are striving towards. Even if we fall a little short, this has been the most productive PRPNet Project of the Month by a long shot. Thank you to everyone for contributing.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - 2 WU's remain in progress
- Complete double check work of 41693 to n=5146239 - ~1100 WU's outstanding
- Complete double check work of 40291 up to n=6M - ~3300 WU's outstanding to 6M, 13000 WU's outstanding to completion
Please keep spreading the word. There are ~65 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.
____________
|
|
|
|
|
|
Adding another 2 cores to this effort.
My Laptop has runtimes up to 59000 secs, lets see how my phenom II X4 crunches through...
____________
|
|
|
|
|
|
Server seems to be down. |
|
|
|
|
|
I cannot access the server either. It's listed as down here. I woke up to a new silver PSA badge but I have a longer road to gold :) |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Server seems to be down.
Yes, it appears that communication with pgllr.mine.nu is down.
FYI...SGS is a good backup port for this...short WU's.
server=SGS:0:1:prpnet.primegrid.com:12000
0% and 1 WU. This will only get work if other ports are down and will allow work to resume on 5oB as soon as Lennart can get it back up.
Sorry for the inconvenience. It's the last thing we needed so close to the end. ;(
____________
|
|
|
|
|
|
Looks like its up and running - I got one from 5oB server at 1432 GDT
EDIT: Definitely ok now, just got another
Regards
Zy |
|
|
|
|
Looks like its up and running - I got one from 5oB server at 1432 GDT
EDIT: Definitely ok now, just got another
Regards
Zy
The stats are working again. Good. I have 2 WUs between 85% and 90%. I had problems to reach free-dc.org too. I guess this was no server problem.
____________
|
|
|
|
|
|
For some reason after the server went back online three instances of llr where running (only two of them are supposed to) and now I have a WU I can't neither complete nor abort it.
Is there any way of aborting it? If it helps, the WU is the one whose number is 1*2^4917928+40291.
rembrecron. |
|
|
|
|
For some reason after the server went back online three instances of llr where running (only two of them are supposed to) and now I have a WU I can't neither complete nor abort it.
Is there any way of aborting it? If it helps, the WU is the one whose number is 1*2^4917928+40291.
rembrecron.
Depends on your settings and how you are running it.
Perhaps the easiest and most extreme is just to kill all of the prpnet related processes. Then just restart the two you want as these should continue given the default settings. Then you should 'cleanup' the unwanted WU by changing the 'startoption' to '2' in the appropriate prpclient.ini file ("Return completed work units, abandon the rest, then shut down").
There are better ways given in the master_prpclient.ini file or individual prpclient.ini files. So in the directory of unwanted WU, edit the file 'prpclient.ini' in the directory of the job you do not want and change the stopaspoption to '2' ("Return completed work units and abandon the rest").
|
|
|
|
|
|
Ha, my systems are running these wu's so slowly they slept right through the outage. |
|
|
|
|
For some reason after the server went back online three instances of llr where running (only two of them are supposed to) and now I have a WU I can't neither complete nor abort it.
Is there any way of aborting it? If it helps, the WU is the one whose number is 1*2^4917928+40291.
rembrecron.
Depends on your settings and how you are running it.
Perhaps the easiest and most extreme is just to kill all of the prpnet related processes. Then just restart the two you want as these should continue given the default settings. Then you should 'cleanup' the unwanted WU by changing the 'startoption' to '2' in the appropriate prpclient.ini file ("Return completed work units, abandon the rest, then shut down").
There are better ways given in the master_prpclient.ini file or individual prpclient.ini files. So in the directory of unwanted WU, edit the file 'prpclient.ini' in the directory of the job you do not want and change the stopaspoption to '2' ("Return completed work units and abandon the rest").
It has been solved. It looks like the server though I was computing that WU when I actually didn't have the WU.
rembrecron. |
|
|
|
|
|
Server just dropped again. I noticed I couldn't load the stats page and http://u-g-f.de/PRPNet/ shows it red.
UPDATE up again... |
|
|
|
|
|
Added 1/2 of a i7 2600K (running at stock) to the effort. Four more (HT) cores running.
Prpnet might get a little more help if a few kinks in the install process were fixed. I downloaded the 4.3.0 Linux .zip version. After running the install and update scripts I still had to "chmod +x" the prpclient script and all of the executables in the prpclient directories.
--Gary |
|
|
|
|
|
zip can't contain unix file perms as far as I know. The 7z gives the correct file perms so maybe the answer is to stop producing a zip for *nix. |
|
|
|
|
*Complete double check work of 2131 to n=4582936 - 2 WU's remain in progress
Done!
*Complete double check work of 41693 to n=5146239 - ~1100 WU's outstanding
305 of 468 are in progress - they are dropping like flies!
|
|
|
|
|
|
Wow, These WU's are getting long. ...... Points are good 20K +
____________
From the High Desert in New Mexico
 |
|
|
|
|
|
It looks like we just crossed into a new FFT. My n=5.11mil task is taking 2ms/bit longer than my n=5.08mil task.
____________
|
|
|
|
|
|
Yes, ouch 512K -> 560K. |
|
|
|
|
|
BUT in good news, all of the remaining 41693 work units are now in progress!
____________
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - 5oB Update
Less than 2 days remain till the end of the month. As mentioned by others in previous posts k=2131 has been COMPLETED and all of k=41693 WU's are now in progress.
Also, we've entered the next FFT length and the testing times have made a jump. Testing times are currently ~11-14 hours for faster hosts and ~18-24 hours for slower hosts. They will continue to increase as n increases. With the longer WU's, the average will be about 800 WU's/day; so the 3rd goal is still within reach by 5 May. :)
NOTE: These are very challenging goals we are striving towards. Even if we fall a little short, this has been the most productive PRPNet Project of the Month by a long shot. Thank you to everyone for contributing.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - COMPLETE!!!!
- Complete double check work of 41693 to n=5146239 - ~175 WU's remain in progress
- Complete double check work of 40291 up to n=6M - ~2400 WU's outstanding to 6M, 12150 WU's outstanding to completion
Please keep spreading the word. There are ~65 active participants. We are still shooting for 120 active participants...tell a teammate or two or three. :)
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.
____________
|
|
|
|
|
|
I know its late, but i just added my i5 quad in the mix now too, everything helps. :)
____________
|
|
|
|
|
|
1*2^n+41693 => 17 left then we are done
Then on to 1*2^n+40291
____________
From the High Desert in New Mexico
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - Finished
The month has ended. Congratulations to everyone who participated in the most successful PRPNet Project of the Month. 86 individuals from 45 teams joined forces to help wrap up the The dual Sierpinski problem (Five or Bust). We still have a ways to go, but we completed over 3/4ths of the work.
Approximately 16.6M cobblestones were awarded and one of three k's completed. The second k has only 17 WU's outstanding. There's still the third goal which might be completed before the 5th of May. It will be a close call.
This last FFT length increase really weighed us down. However, good progress is now being made towards the third goal. We expect at least two more FFT length increases before reaching the ultimate goal of n=9091912.
While no longer the PRPNet Project of the Month, 5oB will remain open until the last WU is double checked. Feel free to drop back in at anytime to help with the cause.
Once again, a HUGE thanks to all who participated. Check back in June for another PRPNet Project of the Month.
Testing times are currently ~11-14 hours for faster hosts and ~18-24 hours for slower hosts. They will continue to increase as n increases. With the longer WU's, the average will be about 800 WU's/day; so the 3rd goal is still within reach by 5 May. :)
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - COMPLETE!!!!
- Complete double check work of 41693 to n=5146239 - ~17 WU's remain in progress
- Complete double check work of 40291 up to n=6M - ~1800 WU's outstanding to 6M, 11550 WU's outstanding to completion
---------------------------------
For users reading this for the first time, please consider donating a core or two to the effort. If you are in search for your PSA badge (or upgrade the one you have), now is a good opportunity to get that done as well as give to a worthy cause. :) PSA credit is available.
To participate, please add the following server to your prpclient.ini file:
server=5oB:100:1:pgllr.mine.nu:13000
Stats
Thank you for your consideration!
New to PRPNet
If anyone needs help setting up PRPNet, PM me and I can provide an individualized "pre-set" package. All you'll have to do is download, double click a couple of times and be quickly contributing to the project. All I need is the following:
email=your@email.here
userid=PG_username (Make sure it's unique. If you don't have a unique PG username, add your PG userID to it. For example, for me, John would become John_2449)
clientid=clientID (can be anything you want...does NOT have to be PG clientID)
teamid=teamname (see here for teams already participating)
OS, 32/64bit, and how many cores you wish to run. Check out stats to see progress.
____________
|
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1957
ID: 352
Credit: 6,139,270,864
RAC: 2,271,114
|
|
Another increase in FFT lenght.
Starting probable prime test of 2^5729416+40291
Using zero-padded Core2 type-3 FFT length 640K, Pass1=640, Pass2=1K, a = 3
Last test using FFT 576k took 94k secs, expecting this to go over 100k secs on Q6600.
Latest tests ~9M will use FFT 960K and +50% of processing time per bit...long way to go.
____________
My stats |
|
|
|
|
Another increase in FFT length.
Starting probable prime test of 2^5729416+40291
Using zero-padded Core2 type-3 FFT length 640K, Pass1=640, Pass2=1K, a = 3 Last test using FFT 576k took 94k secs, expecting this to go over 100k secs on Q6600.
Latest tests ~9M will use FFT 960K and +50% of processing time per bit...long way to go.
2^5732920+40291 will be my last WU under this project as it's taking my 2GHz laptop days to complete: 17.361 ms per bit. The machine only runs for approx. 12 hours each day so I'm guessing this WU will finish Thursday-ish.
I'll go back to the short WUs I usually do on this machine.
Good luck all.
Pete. |
|
|
|
|
|
It looks like we polished off the +41693 form tests a few hours ago...
The last few WUs I completed were in the 60K sec. range on a 2600K with HT on.
--Gary |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - Epilogue
As mentioned in the previous post, k=41693 has been completed. Congratulations! Two out of three goals achieved. However, we've taken quite a hit since the end of April and the third goal is starting to slip.
Testing times are currently ~12-15 hours for faster hosts and ~20-26 hours for slower hosts. They will continue to increase as n increases.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - COMPLETE!!!!
- Complete double check work of 41693 to n=5146239 - COMPLETE!!!!
- Complete double check work of 40291 up to n=6M - 687 WU's outstanding to 6M, 10105 WU's outstanding to completion
____________
|
|
|
|
|
|
Looks like the The Cinco de Mayo Challenge is taking it toll.
Finishing up the downloads and not getting any more.
I will be back, .....
Looks like we will finish this up by the 15th of May for sure.
Good Luck on the Challenge, See ya there.
Steve
____________
From the High Desert in New Mexico
|
|
|
|
|
|
There are two challenges running from 5 April - PG's, and the BOINC Team Pentathelon (finishes 18 May). I'll be back on this 18 May after the Team Pentathelon, if still running, I suspect quite a few others will as well.
Regards
Zy |
|
|
Sysadm@NbgVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1224
ID: 18646
Credit: 877,860,382
RAC: 320,481
|
|
Last workunit ran longer than 24 hours on my Phenom.
The actual running will be the last for a while.
Maybe I will be back, too.
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
|
I, too, will be back after the Challenge. Hopefully, the deadline will be extended, since it is now taking well over 24 hrs per on my machines.
____________
Murphy (AtP)
|
|
|
|
|
|
The deadline had been dropped to 72 hours but last I looked it appears it is 96 hours again. |
|
|
|
|
|
I have scaled back my effort mainly because I am trying to catchup on another project. Also I find the new LLR app a little too hot to take the challenge seriously enough. So I will probably keep 2 or 3 cores on this at least some of the time - hopefully the shorter wus :-) |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
It's a bit late to be mentioning this but for those wanting to empty their cache (which hopefully is just 1 WU) and close PRPNet, simply update the master_prpclient.ini file as follows:
// stopasapoption= tells the client that it needs to be shutdown automatically, i.e. without
// a CTRL-C. It is evaluated after each test is completed. It should be 0 upon startup.
// The accepted values are:
// 0 - Continue processing work units
// 2 - Return completed work units and abandon the rest
// 3 - Return completed work units (keep the rest)
// 6 - Complete all work units and return them
stopasapoption=6
Then update the clients. Once the client finishes the cache, it will shut down. I set mine to 6 a few hours ago which means I'll be late for the Challenge.
If you have more than 1 WU in your cache, then option 2 is a better choice. This allows the current WU to be completed and the rest abandoned.
Remember, when returning to PRPNet, change stopasapoption back to 0.
BTW, the current deadline is 96 hours.
____________
|
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
...
Remember, when returning to PRPNet, change stopasapoption back to 0.
If start-up would treat "6" as if it were "0", then we would not have to keep changing it back and forth when we must do a control-C before the WU finishes. If a WU is in progress, it continues. If no WU is in progress, it shuts down immediately. When current WU finishes, it returns it and shuts down.
____________
Murphy (AtP)
|
|
|
|
|
2^5732920+40291 will be my last WU under this project...
I'll go back to the short WUs I usually do on this machine.
Famous last words, the new PPSELow must be empty so it went and got a WU from a back up server: 2^5875864+40291 :)
Oh well, 4 days or so before it'll complete (18.9ms per bit lol)
Pete
|
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - Epilogue
We didn't quite reach the third goal before the start of the Challenge. However, we came close, n~5.88M. The good news is that there are several participants still active. Therefore, 6M should only be a few days away. :)
Testing times are currently ~14-18 hours for faster hosts and ~24-30 hours for slower hosts. They will continue to increase as n increases.
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - COMPLETE!!!!
- Complete double check work of 41693 to n=5146239 - COMPLETE!!!!
- Complete double check work of 40291 up to n=6M - 370 WU's outstanding to 6M, 9988 WU's outstanding to completion
____________
|
|
|
|
|
|
Sorry John. had to pull the 4 cores off this to put on the challenge. Will try to return once the challenge is done.
Rick |
|
|
|
|
Sorry John. had to pull the 4 cores off this to put on the challenge. Will try to return once the challenge is done.
Rick
Me too!
--Gary |
|
|
|
|
Sorry John. had to pull the 4 cores off this to put on the challenge. Will try to return once the challenge is done.
Rick
Me too!
--Gary
Me three...
____________
|
|
|
|
|
|
Just an update - I'm estimating that around 150 tests are needing triple-checks, I am about a quarter of the way through these so far, but I will post an update once I get some solid statistics.
Phil |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
|
April's PRPNet Project - Epilogue
Yesterday, we completed the third goal. Thank you to everyone who helped us achieve this GREAT milestone. Now it's a long slow crawl to the final goal of finishing off the last k to 9091912. Currently we have 8932 candidates remaining between n=6033256 and n=9091912.
Testing times are currently ~16-20 hours for faster hosts and ~26-32 hours for slower hosts. They will continue to increase as n increases. We expect times will increase to ~45 hours by the end of the effort.
To participate in wrapping this project up, please add the following server to your prpclient.ini file:
server=5OB:100:1:pgllr.mine.nu:13000
Stats
Goals by 5 May 2011
- Complete double check work of 2131 to n=4582936 - COMPLETE!!!!
- Complete double check work of 41693 to n=5146239 - COMPLETE!!!!
- Complete double check work of 40291 up to n=6M - COMPLETE!!!!
____________
|
|
|
|
|
|
That's great news and shows how the PG community works to get something done.
Moved a core over this morning. |
|
|
|
|
|
Leading edge is at 6310504 and the next FFT change is 6325001 from 640K to 672K. Keep those cores going! |
|
|
|
|
|
What is the timeout limit now?
I just had one task wasted because the limit is insufficient:
Starting probable prime test of 2^6412600+40291
2^6412600+40291 is not prime. RES64: 97FE338A05F4593A. OLD64: C7FA9A9E11DD0BAB Time : 115983.230 sec.
[2011-05-23 19:46:00 CDT] 5OB: 1*2^6412600+40291 is not prime. Residue 97FE338A05F4593A.
[2011-05-23 19:46:00 CDT] Total Time: 99:29:44 Total Tests: 5 Total PRPs Found: 0
[2011-05-23 19:46:01 CDT] 5OB: Returning work to server pgllr.mine.nu at port 13000
[2011-05-23 19:46:01 CDT] 5OB: INFO: Test for 1*2^6412600+40291 was ignored. Candidate and/or test was not found
[2011-05-23 19:46:01 CDT] 5OB: INFO: 0 of 1 test results were accepted
[2011-05-23 19:46:02 CDT] 5OB: Getting work from server pgllr.mine.nu at port 13000
[2011-05-23 19:46:04 CDT] 5OB: PRPNet server is version 4.3.2
Also, will the timeout period automatically adjust for the apparent jumps in N that has to occur?
That is that the leading edge is now 6462424 but max N is 9091912 with only 8125 tasks to do. |
|
|
|
|
What is the timeout limit now?
I just had one task wasted because the limit is insufficient:
Starting probable prime test of 2^6412600+40291
2^6412600+40291 is not prime. RES64: 97FE338A05F4593A. OLD64: C7FA9A9E11DD0BAB Time : 115983.230 sec.
[2011-05-23 19:46:00 CDT] 5OB: 1*2^6412600+40291 is not prime. Residue 97FE338A05F4593A.
[2011-05-23 19:46:00 CDT] Total Time: 99:29:44 Total Tests: 5 Total PRPs Found: 0
[2011-05-23 19:46:01 CDT] 5OB: Returning work to server pgllr.mine.nu at port 13000
[2011-05-23 19:46:01 CDT] 5OB: INFO: Test for 1*2^6412600+40291 was ignored. Candidate and/or test was not found
[2011-05-23 19:46:01 CDT] 5OB: INFO: 0 of 1 test results were accepted
[2011-05-23 19:46:02 CDT] 5OB: Getting work from server pgllr.mine.nu at port 13000
[2011-05-23 19:46:04 CDT] 5OB: PRPNet server is version 4.3.2
Also, will the timeout period automatically adjust for the apparent jumps in N that has to occur?
That is that the leading edge is now 6462424 but max N is 9091912 with only 8125 tasks to do.
If you watch the pending tasks lists then it seems that after the challenge (when it was dropped to 72 hours that it was returned to 96 hours.
115983/60/60 = 32.2175 hours
I have slower procs that I'm averaging 34 hours for a WU on with no issues on the tasks being accepted. My faster procs are averaging 20 hours at this time.
I'm using "server=5oB:100:1:pgllr.mine.nu:13000" with only 1 task fetched at a time. Do you need to adjust your work queue length?
Next big jump is at N=6625001 when the FFT goes from our present 672K to 768K.
Someone should double check my numbers but I believe FFT sizes for the ranges of N for this project are as follows:
N-min N-max FFT size
5090001 5705000 576K
5705001 6325000 640K
6325001 6625000 672K
6625001 7535000 768K
7535001 7845000 800K
7845001 8460000 864K
8465001 8775000 896K
8775001 9400000 960K |
|
|
|
|
Also, will the timeout period automatically adjust for the apparent jumps in N that has to occur?
That is that the leading edge is now 6462424 but max N is 9091912 with only 8125 tasks to do.
Oh, I meant to add the timeout is manually adjusted by the admin. |
|
|
|
|
What is the timeout limit now?
I just had one task wasted because the limit is insufficient:
Starting probable prime test of 2^6412600+40291
2^6412600+40291 is not prime. RES64: 97FE338A05F4593A. OLD64: C7FA9A9E11DD0BAB Time : 115983.230 sec.
[2011-05-23 19:46:00 CDT] 5OB: 1*2^6412600+40291 is not prime. Residue 97FE338A05F4593A.
[2011-05-23 19:46:00 CDT] Total Time: 99:29:44 Total Tests: 5 Total PRPs Found: 0
[2011-05-23 19:46:01 CDT] 5OB: Returning work to server pgllr.mine.nu at port 13000
[2011-05-23 19:46:01 CDT] 5OB: INFO: Test for 1*2^6412600+40291 was ignored. Candidate and/or test was not found
[2011-05-23 19:46:01 CDT] 5OB: INFO: 0 of 1 test results were accepted
[2011-05-23 19:46:02 CDT] 5OB: Getting work from server pgllr.mine.nu at port 13000
[2011-05-23 19:46:04 CDT] 5OB: PRPNet server is version 4.3.2
Also, will the timeout period automatically adjust for the apparent jumps in N that has to occur?
That is that the leading edge is now 6462424 but max N is 9091912 with only 8125 tasks to do.
If you watch the pending tasks lists then it seems that after the challenge (when it was dropped to 72 hours that it was returned to 96 hours.
115983/60/60 = 32.2175 hours
I have slower procs that I'm averaging 34 hours for a WU on with no issues on the tasks being accepted. My faster procs are averaging 20 hours at this time.
I'm using "server=5oB:100:1:pgllr.mine.nu:13000" with only 1 task fetched at a time. Do you need to adjust your work queue length?
Yup, I am thinking to change it from one (where it has been since the start) to zero!
I am very aware of what this host is doing and has done (38 successfully since 2011-04-25 22:42:18 CDT until that one - ranging between 40362 and 99147 seconds). So clearly a conflict between prpnet and boinc for it not to get the time allocated (both are meant to be niced at 19).
Next big jump is at N=6625001 when the FFT goes from our present 672K to 768K.
Someone should double check my numbers but I believe FFT sizes for the ranges of N for this project are as follows:
N-min N-max FFT size
5090001 5705000 576K
5705001 6325000 640K
6325001 6625000 672K
6625001 7535000 768K
7535001 7845000 800K
7845001 8460000 864K
8465001 8775000 896K
8775001 9400000 960K
Thanks, but we should be hitting these fairly quickly with only 8K tasks left. |
|
|
|
|
|
I may be wrong, but I think this can happen if the previous cruncher of that test times out and the subsequently returns a result before you do. The first result is accepted and the second is not because it's already been received. |
|
|
|
|
|
I may have missed the answer but what is the timeout period now? I am fixing to start several jobs on a system that only runs 12 hours at a time and want to make sure I'm not wasting the cores and getting the wu's done to help.
Rick |
|
|
mfbabb2Volunteer tester
Send message
Joined: 10 Oct 08
Posts: 510
ID: 30360
Credit: 20,784,268
RAC: 731
|
|
At least 82 hours.
See: http://pgllr.mine.nu:13000/pending_tests.html
____________
Murphy (AtP)
|
|
|
|
|
At least 82 hours.
See: http://pgllr.mine.nu:13000/pending_tests.html
Thanks Murph.. I will have 8 cores running but the system is fairly slow and I would've hated to have those go to waste. Hopefully over the weekend I can get it running 24/7 to insure they all get done.
Rick |
|
|
|
|
|
Rick
A comparitor for you on current WUs at the slower end of the scale - one of my machines is a laptop running one WU 24x7, its a VIAO 2Ghz (T7250), I usually run one WU from 5oB on it. It started another one at 0545 GDT this morning, and at time of post is at 25% done. So it looks like it will take 40hrs in all running 24x7 - or 80hrs 12x7
Stepping up a bit speed wise, another one, a PhenomII x4 @3.3Ghz started one at 0240 GDT this morning, currently is 36% done at time of post, also running 24x7
Regards
Zy |
|
|
|
|
|
Intel(R) Core(TM) i7 CPU 860 @ 2.80GHz [Family 6 Model 30 Stepping 5]
(8 processors)
Here's what I'm going to be running. I've had it running the ppselow port for a while. Based on what you gave it looks like it's right between your systems.
thanks for the info
Rick |
|
|
|
|
|
I think it is 96 hours right now.
On systems I have going 24/7 some timing data:
Intel Xeon 5160 @ 3.00GHz: 20 hours (completed 2011-05-26 11:28:13 UTC)
Intel Xeon X5355 @ 2.66GHz: 31 hours (completed 2011-05-26 10:11:40 UTC)
Those both represent the best case numbers as for the 2nd proc some are running to 34 hours.
Earlier into my start I had a AMD Phenom II X6 1090T @ 3.2 Ghz (with turbo-boost disabled) core running for comparison and it beat the two Xeons by 20-30%. That proc is now back to full BOINC as I'm chasing the long LLR badges. |
|
|
|
|
Intel(R) Core(TM) i7 CPU 860 @ 2.80GHz [Family 6 Model 30 Stepping 5]
(8 processors)
Be wary of the hyper threading for such projects. That cpu is 4 real cores and 8 threads. Benchmarking is always best of course but generally turning off HT and running only 4 instances is going to be a good idea. Depending upon scheduling you may do fine with HT on as long as you only run 4 instances.
Even the latest generation of HT when I benchmark for pure cpu bound tasks like this project, file server tasks, or database tasks then the performance is always worse with HT on. It's not nearly as crushing as the original HT but seems to only be a good fit for a system with a very diverse and/or non-cpu intense work load. By non cpu-intense I mean if you have a 4 core proc with 8 threads total then your general workload must not exceed 4 "threads" fully busy. On that proc if you run 5 llr processes at the same time then you should expect all 5 to run slower to the point that running 4 will be faster overall.
I'm happy to hear feedback from others on HT v no HT v limiting cpu cores used to real cores. I'm only talking about the new HT in core iN procs or newer Xeons and not the trash that was HT years ago. Most of my testing focus is more on the server side and of course suffers from a small sample set. Also, I have quite the bias from earlier HT that makes me very wary of these "threads." |
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2392
ID: 1178
Credit: 18,654,341,482
RAC: 6,969,388
                                            
|
|
Rule of thumb at PG for HT generally (at least from my experience) is as follows:
1) Always run it on sieves. The work runs slower for each workunit, but the total machine throughput is always greater--ranging from as little as 20-30% more to as much as 70% to nearly double throughput (**note: these numbers include some sieves that are now deactivated projects).
2) Always run HT on for small LLR (i.e., PPS and SGS). Though not as big a gain as with sieves, total throughput increase has ranged from about 5% to 25% depending on the particular machine.
3) Mid-range LLR (i.e., TRP, 321, etc.)...HT typically doesn't matter. Generally (at least for me), the workunit times double almost exactly with HT on.
4) Big LLR (especially SoB, but including some others), HT on tends to result in slightly longer than double time. Thus, with these having HT turned off is generally better.
5) What if it is a mixed bag (i.e., sieves and LLR together)? This is where it gets tricky, but if you can get both to run in balance (which is virtually impossible), it seems that HT will give a bit of an overall boost in total throughput--even if the LLR are of the big variety.
So what does that mean for the 5oB work? Well, the earlier work would be more along the lines of #3 above, so HT on probably make sense. However, as the project progresses toward completion, workunit times are increasing substantially, so a shift to #4 above is likely.
____________
141941*2^4299438-1 is prime!
 |
|
|
|
|
|
I would add the following things:
No HT if you want to be the Initial finder for PPS and SGS LLRs (or increase the probability of it)
No HT if you need to control heat
No HT if there is a deadline like the end of a challenge
Leave a core for a GPU when running a GPU (sub)project
I have never tested beyond the small LLRs in boinc. I usually enable HT and then just restrict myself to the number of cores that give the best performance. Under Linux, that usually means the number of real cores with LLRs because I don' consider the different enough to use all core. But I use 99% of HT cores on my windows i7 host with SGS so one core if left for a GPU project. Otherwise it really messes the performance of the GPU tasks. |
|
|
|
|
1) Always run it on sieves. The work runs slower for each workunit, but the total machine throughput is always greater--ranging from as little as 20-30% more to as much as 70% to nearly double throughput (**note: these numbers include some sieves that are now deactivated projects).
Interesting that would indicate that the sieves may not be making optimal use of caches and pipelines. Though, for some sieves the memory usage can be high in which case the hit will be all the transfer to/from main memory. Now, you have me wondering what oprofile would tell us.
2) Always run HT on for small LLR (i.e., PPS and SGS). Though not as big a gain as with sieves, total throughput increase has ranged from about 5% to 25% depending on the particular machine.
This may be similar to my remark in 1 or may just be the nature of short and fast WUs.
3) Mid-range LLR (i.e., TRP, 321, etc.)...HT typically doesn't matter. Generally (at least for me), the workunit times double almost exactly with HT on.
If you have 8 threads then with HT off on 4 cores you would get say 5 hours each with 4 going and with HT and 8 going you'd get 10 hours for each? I'm actually surprised that it is not 2x + overhead.
4) Big LLR (especially SoB, but including some others), HT on tends to result in slightly longer than double time. Thus, with these having HT turned off is generally better.
Odd, this is what I would have expected for 3.
HT is is an obnoxious problem for all distributed computing projects since in general the OS will present threads as standard cpu cores. My search across BOINC upstream wiki, bug tickets, etc. didn't yield much. If anyone has information, including links, to anything especially for BOINC upstream on hyperthread "thread v real core" detection or handling then let me know. Specifically anything that allows tuning without the user doing something like manually forcing core use to half and hoping the scheduler is smart. |
|
|
|
|
|
FFT range for N
672K 6325001-6625000
768K 6625001-7535000
Leading edge is 6624184. Prepare for the next slow down! |
|
|
|
|
|
Great :( the 1 pc I have on it now takes days already. Life in the slow lane. |
|
|
|
|
|
Just noticed on 5oB that the visible cut off time appears to have been reduced. Longest WU is around 44hrs at time of post. Could be that there is a long gap between the current longest running WU and actual timeout ..... but just being cautious as I have some that may nudge up and maybe slightly over 48hrs.
The latter would be a tight timeout given the ever lengthening WUs, and the fact that we are a couple of months away from completing it all, so long(ish) timeouts hardly matter at this point.
72hrs would be good - apologies in advance if it in fact already is 72hrs+
EDIT: Just thought ..... have seen this question asked a few times in different ways over the weeks/months. Would be handy to insert a small bit of additional code for the individual WU update page build, to insert into the top line to the effect that "The work timeout is currently xxhrs"
Regards
Zy |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1256
ID: 12001
Credit: 18,565,548
RAC: 0
|
EDIT: Just thought ..... have seen this question asked a few times in different ways over the weeks/months. Would be handy to insert a small bit of additional code for the individual WU update page build, to insert into the top line to the effect that "The work timeout is currently xxhrs"
I've considered doing something on the pending tests page. I just haven't gotten around to it yet. |
|
|
|
|
Great :( the 1 pc I have on it now takes days already. Life in the slow lane.
My post was meant to encourage, not depress :)
Numbers on my machines:
2011-05-29 20 hours - 35 hours
2011-06-01 26 hours - 41 hours
The numbers on the 2nd are after the FFT increase. Both ranges are my faster to slowest cpus. |
|
|
|
|
Great :( the 1 pc I have on it now takes days already. Life in the slow lane.
My post was meant to encourage, not depress :)
Numbers on my machines:
2011-05-29 20 hours - 35 hours
2011-06-01 26 hours - 41 hours
The numbers on the 2nd are after the FFT increase. Both ranges are my faster to slowest cpus.
lol don't get me wrong, I'm glad the wu's task list is going down.. |
|
|
|
|
Rule of thumb at PG for HT generally (at least from my experience) is as follows:
1) Always run it on sieves. The work runs slower for each workunit, but the total machine throughput is always greater--ranging from as little as 20-30% more to as much as 70% to nearly double throughput (**note: these numbers include some sieves that are now deactivated projects).
2) Always run HT on for small LLR (i.e., PPS and SGS). Though not as big a gain as with sieves, total throughput increase has ranged from about 5% to 25% depending on the particular machine.
3) Mid-range LLR (i.e., TRP, 321, etc.)...HT typically doesn't matter. Generally (at least for me), the workunit times double almost exactly with HT on.
4) Big LLR (especially SoB, but including some others), HT on tends to result in slightly longer than double time. Thus, with these having HT turned off is generally better.
5) What if it is a mixed bag (i.e., sieves and LLR together)? This is where it gets tricky, but if you can get both to run in balance (which is virtually impossible), it seems that HT will give a bit of an overall boost in total throughput--even if the LLR are of the big variety.
So what does that mean for the 5oB work? Well, the earlier work would be more along the lines of #3 above, so HT on probably make sense. However, as the project progresses toward completion, workunit times are increasing substantially, so a shift to #4 above is likely.
Here are my test results for PPS/Boinc on a i7 950 (no OC) running Ubuntu 64 running no GPU tasks.
Results are interesting.
8 PPS Tasks (4 core W/HT On) 26:36 per task
4 PPS Tasks (4 core W/HT On) 13:45 per task
4 PPS Tasks (4 core W/HT Off) 12:55 per task
It seems to me that HT off is the way to go for even PPS
____________
|
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2392
ID: 1178
Credit: 18,654,341,482
RAC: 6,969,388
|
Here are my test results for PPS/Boinc on a i7 950 (no OC) running Ubuntu 64 running no GPU tasks.
Results are interesting.
8 PPS Tasks (4 core W/HT On) 26:36 per task
4 PPS Tasks (4 core W/HT On) 13:45 per task
4 PPS Tasks (4 core W/HT Off) 12:55 per task
It seems to me that HT off is the way to go for even PPS
Interesting results. My own results testing this were on Win XP/Vista/7 machines, so there may be an OS difference with this. Also, when I tested was quite some time ago, and PPS has moved to higher ranges now such that HT might indeed be worse for all LLR (though SGS is now the smallest LLR, so HT might still be okay). Also, have you tried it with mixed LLR and sieve to see if HT performs better under mixed results?
____________
141941*2^4299438-1 is prime!
|
|
|
|
|
Here are my test results for PPS/Boinc on a i7 950 (no OC) running Ubuntu 64 running no GPU tasks.
Results are interesting.
8 PPS Tasks (4 core W/HT On) 26:36 per task
4 PPS Tasks (4 core W/HT On) 13:45 per task
4 PPS Tasks (4 core W/HT Off) 12:55 per task
It seems to me that HT off is the way to go for even PPS
Interesting results. My own results testing this were on Win XP/Vista/7 machines, so there may be an OS difference with this. Also, when I tested was quite some time ago, and PPS has moved to higher ranges now such that HT might indeed be worse for all LLR (though SGS is now the smallest LLR, so HT might still be okay). Also, have you tried it with mixed LLR and sieve to see if HT performs better under mixed results?
I have never tried that. Usually I dedicate my PCs to a subproject at a time.
I'm going to do this same test with TRP Sieve later.
____________
|
|
|
|
|
|
Very interesing.
I tested some time ago and PPS / SGS with HT on was better than without HT. Win 7 64b, so maybe linux is different in this case.
Combination of 4 llr and 4 sieve may be the best option. |
|
|
|
|
|
Yesterday I shut down my rig (i7 2600K o/c at 40/100, gtx570 at stock, ubuntu 10.10) and rebooted with HT off in BIOS. I put 2 cores on 5oB and 2 on boinc TRP Sieve, and let the GPU run PPS Sieve. 5oB times were consistently between 7.3-7.4 ms/bit, and TRP sieving times were about 7100s per WU.
Shut down again and booted with HT on, with no other changes. 4 threads on 5oB, 4 on boinc TRP sieve, and 1 GPU running PPS sieve. 5oB is running consistently under 14.0 ms/bit (a *little* under... 13.8 or so) and sieving times are 12000-13000s.
Seems like "HT on" is the choice at least for this combination.
--Gary |
|
|
VatoVolunteer tester
Send message
Joined: 2 Feb 08
Posts: 851
ID: 18447
Credit: 713,640,206
RAC: 1,642,230
                     
|
|
Yes, "modern" (ie i7) HT is good for throughput on the right mix of load, but only if you have an integer and a floating point task on each physical core, which can sometimes be hard to guarantee.
____________
 |
|
|
|
|
|
Can You Extend the Time out Limit by 24hr.
So I can put All cores into the Challenge?
Steve
____________
From the High Desert in New Mexico
|
|
|
|
|
Can You Extend the Time out Limit by 24hr.
So I can put All cores into the Challenge?
Steve
Delay is raised to 144 hr's
Lennart |
|
|
JohnHonorary cruncher
Send message
Joined: 21 Feb 06
Posts: 2875
ID: 2449
Credit: 2,681,934
RAC: 0
|
Can You Extend the Time out Limit by 24hr.
So I can put All cores into the Challenge?
Delay is raised to 144 hr's
Lennart also upgraded to PRPNet 4.3.4 so now you can see the "Expires" column in pending. :)
____________
|
|
|
|
|
|
Thanks, Lennart & John.
The "Expires" column is great
Steve
|
|
|
|
|
|
Expires column is great indeed. I wish I had known this was going to be raised sooner as I have been having all my clients complete their work and shutdown in preparation for the challenge. |
|
|
|
|
|
2-3 days and the next FFT jump from 768K to 800K should take place. |
|
|
|
|
2-3 days and the next FFT jump from 768K to 800K should take place.
So This should jump up WU time by what percent?
Just your best guess
TIA, Steve
____________
From the High Desert in New Mexico
|
|
|
|
|
|
Looks like my first WU on the new FFT length on 1 system brought me from 118623.704 sec to 111964.431 sec. So a 5.95% increase but some of that is also larger N and not just the FFT increase. |
|
|
|
|
Looks like my first WU on the new FFT length on 1 system brought me from 118623.704 sec to 111964.431 sec. So a 5.95% increase but some of that is also larger N and not just the FFT increase.
Thanks,
I don't even want to think about the last 5% of these WU's on time each will take.
At least the points keep going up as with the N.
I have a hard time with WU's that take over a day or two. SOB's are just killers for me.
Thanks again for the info,
Steve
____________
From the High Desert in New Mexico
|
|
|
|
|
|
SoB waiting for your dc (or inevitable dc timeouts) is the most killer part of it. I should have my silver CUL LLR whenever all the dc completes and I just switched over to working on my silver SoB. Saving my other bronze upgrades for the challenges.
768K to 800K wasn't a major jump as the last jump was 672K to 768K. At N=7845001 is when we move to 864K. |
|
|
|
|
|
Next FFT jump is at N=7845001 and is 800K to 864K. This will run through N=8460000. A very crude guess says WU times will jump about 12-15%. |
|
|
|
|
|
Does anyone know where pgllr.mine.nu:13000 went or when it will return? It seems I haven't been able to connect for 24 hours now...would like to get units completed and turned in ahead of the challenge coming up. |
|
|
|
|
|
working on it...
Lennart
EDIT: Should work now :) |
|
|
|
|
|
For those that have not noticed it. This thread gets a new home as the PRPNet October Project. |
|
|
