PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise

Advanced search

Message boards : Number crunching : Are hardware requirements too high?

Author Message
Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55582 - Posted: 18 Jun 2012 | 15:03:52 UTC - in response to Message 55369.

well this is stupid.
I have invested over 2K into my machine, and frankly I'm sick of having software coders make stuff for Boinc that just can't work on high end gear unless you have a: b: c: d: e: f: g: and subsection x of every point, contingency and wish list you can think of.

I understand that there's need for technological advances, but I'm having a hard time swallowing why Wu's cannot "just work".

It's not up to us to donate our gear to your specs... It's up to you to make WU's to suit generic gear. Because when we do you still can't provide functional WU's without a mile of contingencies.

We are volunteering our machines for you to "get ahead" - it's not the other way around.

I'm sorry I don't have a Cray computer to suit your needs. I'm also sorry you can't make work units that aren't so freakin bloated.

I'll finish my current work units and I'm cancelling this project, just as I've cancelled others with ludicrous expectations for volunteer computing.
____________

Arjant2Project donor
Send message
Joined: 14 Aug 10
Posts: 193
ID: 65556
Credit: 163,239,345
RAC: 326,050
321 LLR Gold: Earned 500,000 credits (858,704)Cullen LLR Silver: Earned 100,000 credits (261,684)ESP LLR Silver: Earned 100,000 credits (109,285)PPS LLR Ruby: Earned 2,000,000 credits (2,326,228)PSP LLR Silver: Earned 100,000 credits (312,735)SoB LLR Silver: Earned 100,000 credits (154,103)SR5 LLR Silver: Earned 100,000 credits (139,377)SGS LLR Silver: Earned 100,000 credits (334,971)TRP LLR Amethyst: Earned 1,000,000 credits (1,232,766)Woodall LLR Silver: Earned 100,000 credits (441,996)321 Sieve Bronze: Earned 10,000 credits (22,692)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,124,289)PPS Sieve Double Bronze: Earned 100,000,000 credits (143,073,904)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Bronze: Earned 10,000 credits (49,211)TRP Sieve (suspended) Silver: Earned 100,000 credits (371,632)AP 26/27 Ruby: Earned 2,000,000 credits (4,115,774)GFN Ruby: Earned 2,000,000 credits (4,330,983)PSA Gold: Earned 500,000 credits (979,013)
Message 55584 - Posted: 18 Jun 2012 | 15:38:52 UTC
Last modified: 18 Jun 2012 | 15:49:06 UTC

What is your problem?
What is stupid?

According to my knowledge your '2K' computer can run sieves or not?
So why are you complaining in this unrespectful way?

If you are complaining about the fact your AMD cards cannot run all sub- projects on PG, find another project that suits AMD better.
There are projects enough where AMD cards are better than Nvidia cards.

By the way, a lot of people invested in their machines to run PG. If this project was that important to you, you made the wrong choice in hardware.
You could have bought a lot better for probably a lot less. Just my 2 cents.
____________

[DPC]Charley
Volunteer moderator
Project scientist
Send message
Joined: 15 Apr 11
Posts: 803
ID: 95137
Credit: 94,979,517
RAC: 0
321 LLR Amethyst: Earned 1,000,000 credits (1,009,885)Cullen LLR Amethyst: Earned 1,000,000 credits (1,440,123)ESP LLR Amethyst: Earned 1,000,000 credits (1,008,511)PPS LLR Ruby: Earned 2,000,000 credits (2,389,215)PSP LLR Amethyst: Earned 1,000,000 credits (1,080,442)SoB LLR Amethyst: Earned 1,000,000 credits (1,143,918)SR5 LLR Amethyst: Earned 1,000,000 credits (1,853,162)SGS LLR Amethyst: Earned 1,000,000 credits (1,344,963)TRP LLR Amethyst: Earned 1,000,000 credits (1,140,037)Woodall LLR Amethyst: Earned 1,000,000 credits (1,130,355)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,038,617)PPS Sieve Sapphire: Earned 20,000,000 credits (36,643,348)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Ruby: Earned 2,000,000 credits (2,000,717)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,298,244)GFN Jade: Earned 10,000,000 credits (17,332,530)PSA Sapphire: Earned 20,000,000 credits (20,125,451)
Message 55585 - Posted: 18 Jun 2012 | 15:42:11 UTC

Sorry, but what are you going on about?

Are you dismayed because this challenge can only be run on fast cpu's (ineffectively) and NVidia GPU's? Someone has been busy trying to port to OpenCL, but that takes a lot of time and effort.

The hardware configurations mentioned in the first post are meant to given an indication of how long each test will take, not as a list "you must have this hardware else you can't join!". Well, except for the NVidia and 64bit cpu bits. This has been constructed this way for practical reasons to the best of my knowledge. Details can probably be better provided by other people :)

This project is completely run by voluteers so I think it's amazing how well it's working compared to some others run by universities or companies.

Sorry to see you go, as for me it's basically the more the merrier on projects like these :)
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55586 - Posted: 18 Jun 2012 | 15:57:48 UTC - in response to Message 55582.
Last modified: 18 Jun 2012 | 16:02:16 UTC

Axiom,

I'm sorry you feel that way.

Of course, you should feel free to crunch (or not crunch) whatever you want, for whatever reason you want.

I'm not 100% sure what your complaint is about, since all three of your computers should be able to participate in this challenge. I see you've already run GFN WUs.

If you're upset about this only working on Nvidia GPUs and not ATI GPUs, then, well, I agree with you. It would be preferable if there was an ATI version of Genefer, but there isn't right now. If anything, I suspect that Genefer will run faster on ATI than on Nvidia, so it really is disappointing that we don't have an application for ATI. All of our software is written by volunteers, and we can only offer what exists. Generally speaking, developers write software for the hardware that they own.

Somebody is taking a look at writing an ATI version, but I don't have an ETA at this time.

It's not up to us to donate our gear to your specs... It's up to you to make WU's to suit generic gear.


Generic gear? That's pretty much what the CPU version of the app is. Yes, it's resticted to 64 bit CPUs, but that's exactly what your computers are. There's also a much faster version of the application that can take advantage of faster hardware for those that have it, but the generic (CPU) version is available for all. (OK, not quite all, but almost all. There are still some 32 bit CPUs out there -- I sometimes crunch with one myself -- but they're old and slow and not many people want to crunch such long WUs with such slow computers. Even the GFN "short" WUs are working on prime numbers that are nearly three million digits long, so the crunching times are going to be pretty long no matter what.)

Mike

EDIT: I hope you decide to stay -- if a new prime is found in the challenge it will be a huge one (and it looks like we just found one, BTW.) But whether you stay or go, good luck with whatever you decide to crunch. There's a lot of worthy projects out there.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55588 - Posted: 18 Jun 2012 | 16:47:32 UTC - in response to Message 55586.

Michael,
You're right of course, some of my machines can limp through the projects, and some run more efficiently than others as we all well know. I perhaps wrote a bit to hastily in the manner I did.

My gripe is about the massive red "Attention" listed in this project (and others) - think of it as the "final" straw.

Over the last ten odd years I've participated in these projects, I've found that the WU's become less efficient, and significantly more bloated, and though I am WELL aware that someone is volunteering their time to make the WU's, I'm painfully aware that the majority of people here are NOT cutting edge with their machines. Arjant2 for instance has one i7, the rest is dual core.
So on this Genefer, he gets to process 4 WU's with 2 incompletes per dual core machine. Oh, not including he misses out on any ATI GPU work, and only two machines have nVidia. He sets the bar I'm referring to.

OpenCL has been out for YEARS so the whining from ATI users that too few WU's are present is as annoying as "we're working on it" - whilst the coders simultaneously code only for the newest gear.

The stats on Boinc projects a a whole show a dual core non double precision gpu is the NORM. So code for that. WU's remain smaller and less intensive.

I'm a designer, running Adobe apps and video editing, and tying up cpu and gpus for 25+ hours is so far off the mark it makes my arse hurt. It's not simply Prime that is off the nut. My machine is capable of working through the crunching... it's the fact that it's HOGGING resources for other projects which are hogging resources for all other projects.

In the end, it's the shear amount of time each project is demanding and they are all upping their crunch time exponentially - it's like all boinc projects are establishing a mutually assured destruction.

I have four cores running 24/7 and two GPU's - one double precision, one not.
I'm running pretty decent credits per day. I'm not whingeing that my machines aren't up to snuff, I'm getting frustrated that these things are getting simply too out of hand.

By the Genefer example, my GPU is out, people have to update their nVidia drivers (which can crap out other software unrelated to boinc) and at the end of the day all ATI people are left to calculate 25hour WU's - so for the three day challenge that means (in my case) 8 WU's and 4 incompletes before the time runs out.

But SOME nVidia cards will crank out 14 WU's in the same time with one incomplete (14.4)

So arjant2 and Pyrus, get a grip on your egos.
I have participated in this project because I want to. I allocate full resources because I want to. What I don't appreciate is having to constantly have coders raise a bar that is not reflected in the the GENERAL populace of the OTHER people participating.
Your own machines included. My other machines included.

I have three other machines chipping in their time now and then and they would take three days to do one Genefer WU. That's ludicrous.
That's why I'm complaining.

So you two can go upgrade your machines or sit and wait the 25 hours per cpu all you want. Hell, go drop another $500 on a double GPU nVidia, I couldn't care less what you want to do. It's whatever YOU want to do, have at it.

I'm doing what I want to do.
I'm voicing a long nagging complaint about the development path that these projects are taking. Planned obsolescence of the people volunteering their machines and constantly seeing their equipment take longer and longer to complete a WU - unless you upgrade.

If I was the conspiracy type I'd say nVidia and ATI planned this all along to increase sales and all the projects are a front for increasing revenue.

I'm being facetious by the way *smiles*

Arjant2Project donor
Send message
Joined: 14 Aug 10
Posts: 193
ID: 65556
Credit: 163,239,345
RAC: 326,050
321 LLR Gold: Earned 500,000 credits (858,704)Cullen LLR Silver: Earned 100,000 credits (261,684)ESP LLR Silver: Earned 100,000 credits (109,285)PPS LLR Ruby: Earned 2,000,000 credits (2,326,228)PSP LLR Silver: Earned 100,000 credits (312,735)SoB LLR Silver: Earned 100,000 credits (154,103)SR5 LLR Silver: Earned 100,000 credits (139,377)SGS LLR Silver: Earned 100,000 credits (334,971)TRP LLR Amethyst: Earned 1,000,000 credits (1,232,766)Woodall LLR Silver: Earned 100,000 credits (441,996)321 Sieve Bronze: Earned 10,000 credits (22,692)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,124,289)PPS Sieve Double Bronze: Earned 100,000,000 credits (143,073,904)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Bronze: Earned 10,000 credits (49,211)TRP Sieve (suspended) Silver: Earned 100,000 credits (371,632)AP 26/27 Ruby: Earned 2,000,000 credits (4,115,774)GFN Ruby: Earned 2,000,000 credits (4,330,983)PSA Gold: Earned 500,000 credits (979,013)
Message 55589 - Posted: 18 Jun 2012 | 17:03:34 UTC

Why do Pyrus and I have ego's?
Excuse me for saying, but who is the one with his big ego and his "2K" computer?

You are now complaining that your cpu can't crunch GFN in a reasonable time and Nvidia cards can?

You also refer to my dual cores. I won't be using the CPU's for GFN, why should I? There are other sub projects which benefit far more from cpu power.
My dual cores will keep on crunching other LLR's.
I still don't see your problem. There are many projects that are GPU only, are you complaining there too that you can't use your CPU?

You also say that WU's are becoming more inefficient. In fact, they are getting more efficient. Because of the new apps WU's are done far more efficient. And yes, for some of those new efficient apps you need newer hardware with new techniques. Is that a problem? I don't think so, it only helps the project.

So what has been said before, use your computer the way you want it, but don't 'blame' any volunteer here.

____________

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55592 - Posted: 18 Jun 2012 | 17:25:44 UTC

Before this gets out of hand and I have to start hiding posts and restricting posting privileges (both of which are things I really HATE to do, but will do if needed), everyone please remember that being civil to one another is a requirement and not merely a nice thing to do.

Keep the comments positive. Talk about how we can make things better. Feel free to talk about thing you don't like. But keep the criticism positive in nature. And never, ever, talk about other people in a negative manner.

If in doubt, hit the delete button rather than the "post reply" button. If there's any doubt, it's probably better to wait until tomorrow to post your reply. That gives you more time to think about the subject and formulate a message that will yield better results for you.

I can't begin to count the number of times I've spent 30 or 60 minutes writing something, only to realize that I'm better off saying nothing and simply erasing everything.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55593 - Posted: 18 Jun 2012 | 17:28:16 UTC
Last modified: 18 Jun 2012 | 17:31:49 UTC

Moved the discussion over to this thread...

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55595 - Posted: 18 Jun 2012 | 17:37:45 UTC - in response to Message 55589.

Arjant2,
I made a post reflecting the times to process the Genefer WU's quote by Prime itself at the start of this thread.
Their "requirements" have mimicked other posts about other challenges.

"here's a challenge - but you need this and this, as this won't work."

The majority of crunchers do not have that gear.
So since it's apparent the coders have time to code for things that are not mainstream, surely they could use their time more wisely to shore up the mainstream WU's to work more efficiently, over a smaller amount of time, so that all projects aren't taking days to complete a WU.

I don't know what YOU are going on about.

My machine is what it is.
You have NO IDEA what's under the hood to make it cost so much.
My price reference was to iterate that I'm not using "all in one" brandname computer I got on clearance sale at Best Buy.

My machines do just fine.
What I'm saying is that it's becoming overkill to constantly code for equipment not in use, rather than coding for equipment that IS in use. So yes, I'm miffed that I have to use noob cpu processing to participate, which upsets me bacause I have this big honking GPU just sitting there.

Your need to be a troll and be obtuse as to what I'm saying baffles me. I'm obviously not computer illiterate, and my "frustration" is so that the hundreds of thousands of participants who have computers more like YOURS can have better crunching results.

You can settle for using your older machines to do lesser tasks, or you can upgrade

OR

(and this is my sticking point) the coders could write more stuff to work with the mainstream scenarios of dual core ATI machines so that you don't HAVE to upgrade.

If you want to settle for less - and defend it - then start your own thread.

I've stated my "issue" three times.
Maybe you should try reading rather than flaming.

Anyway, I have more than amply voiced my concern at the WU's vs actual available computing power and really wish this hadn't gotten so lengthy.

As for Pyrus... You sir, I must apologize to.
I read your message initially after reading Arjant2's drivel and grossly mistook what (and how) you were talking about.

Upon rereading your note, I see that it was solely my mistake at being snotty in your direction and my "ego" comments directed to you were completely undeserved and for that I do apologize.

Because "editing" is not in my nature, I will leave my error in commenting so that people can mock me as they may :)

Scott BrownProject donor
Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Avatar
Send message
Joined: 17 Oct 05
Posts: 1923
ID: 1178
Credit: 6,441,245,511
RAC: 2,828,016
Discovered the World's First base 116 Generalized Cullen prime!!!Discovered 16 mega primesEliminated 7 conjecture "k"sDiscovered 1 Sophie Germain pairDiscovered 1 Fermat divisor2012 Tour de Primes highest prime count2012 Tour de Primes most Mountain Stage primes2015 Tour de Primes highest prime count2016 Tour de Primes highest prime countFound 23 primes in the 2018 Tour de PrimesFound 1 mega prime in the 2018 Tour de PrimesFound 2 primes in the 2018 Tour de Primes Mountain Stage2019 Tour de Primes highest prime countFound 22 primes in the 2019 Tour de Primes321 LLR Double Bronze: Earned 100,000,000 credits (100,829,118)Cullen LLR Double Bronze: Earned 100,000,000 credits (103,870,990)ESP LLR Double Bronze: Earned 100,000,000 credits (137,499,413)Generalized Cullen/Woodall LLR Double Bronze: Earned 100,000,000 credits (108,461,080)PPS LLR Double Silver: Earned 200,000,000 credits (384,802,005)PSP LLR Double Bronze: Earned 100,000,000 credits (108,003,110)SoB LLR Double Bronze: Earned 100,000,000 credits (135,747,083)SR5 LLR Double Silver: Earned 200,000,000 credits (201,224,339)SGS LLR Double Bronze: Earned 100,000,000 credits (161,503,178)TPS LLR (retired) Silver: Earned 100,000 credits (235,439)TRP LLR Double Bronze: Earned 100,000,000 credits (121,443,822)Woodall LLR Double Bronze: Earned 100,000,000 credits (101,447,725)321 Sieve Double Silver: Earned 200,000,000 credits (203,510,966)Cullen/Woodall Sieve (suspended) Emerald: Earned 50,000,000 credits (83,794,448)Generalized Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (285,139,652)PPS Sieve Double Ruby: Earned 2,000,000,000 credits (2,108,274,845)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Double Silver: Earned 200,000,000 credits (203,523,358)TRP Sieve (suspended) Double Silver: Earned 200,000,000 credits (201,489,157)AP 26/27 Double Bronze: Earned 100,000,000 credits (135,790,733)GFN Double Amethyst: Earned 1,000,000,000 credits (1,295,584,459)PSA Double Silver: Earned 200,000,000 credits (259,058,048)
Message 55596 - Posted: 18 Jun 2012 | 17:52:49 UTC

Hi Axiom,

I am sorry that you are feeling as you do. In truth, all long-standing participants in the BOINC world have had some level of frustration about things at some point. But I would like to clear up some issues that are contributing to your frustration, if I may.

First, the nature of large prime searching necessitates that the work will get longer. To put it simply, we are searching for ever increasingly large primes, and as such, the length of time it takes to complete prime calculations will also increase.

Second, regarding GFN work specifically, the work available on BOINC for the challenge is quite long out of some necessity. Without going into substantial technical detail, PG is quite maxed out on the number of sub-projects that can be handled in the BOINC environment (which, BTW, was not designed explicitly for a multiple sub-projects set-up) such that only the current two varieties of work are available. Smaller GFN work is available on PRPnet (Indeed, the GFN32768 work is comparable in length to PPS LLR, though it doesn't make the top 5000 list anymore), though it does not count for the challenge. Ultimately, it may be possible to move some of this work to BOINC as other sub-projects complete, but not just yet. Also, please note that this is a BOINC limitation more so than an issue with PG.

Third, suggesting that issues with GPU crunching lie overwhelmingly with individual projects is overstated. To be sure, there are projects (which will go nameless here) that have pushed most GPUs out due to running only high end work. That, however, is not the case here at PG. The PPS sieve app works with all CUDA capable NVidia GPUs and all OpenCL capable ATI/AMD GPUs. Though soon to be defunct, the CW sieve has equal application to NVida GPUs. The GFN project offers another option, but due to computational needs, requires a double-precision GPU (and as Mike noted, an OpenCL application is hopefully on the way...just not in time for this challenge). Over on PRPnet, there is now the addition of an OpenCL application (in very late Beta stages) for either NVidia or ATI/AMD for both the Wieferich and Wall-Sun-Sun projects. What is key here is that, while PG offers numerous projects, each sub-project does have particular computational needs. Indeed, even the first GPU application available at PG, the AP26 project app, required the atomic functions capability in CUDA 1.1, which excluded those with the very first generation NVidia CUDA-capable cards.

The bottom line to this sort of crunching is that things will progress. Here at PG, there is a great desire to be as inclusive as possible. This goal, however, is not always attainable given 1) limited volunteer (especially programming) resources and 2) the nature of particular prime search problems. Nevertheless, I think it is quite impressive that PG supports as wide a range of work as it does such that almost any personal computer from the last several years can still crunch something here...just not everything. Heck, I even crunched a CW sieve unit on my Android phone!

I hope that you will continue to crunch here at PG, but if you chose to move on, I wish you the best of fortunes on one of the many other worthwhile projects.


Arjant2Project donor
Send message
Joined: 14 Aug 10
Posts: 193
ID: 65556
Credit: 163,239,345
RAC: 326,050
321 LLR Gold: Earned 500,000 credits (858,704)Cullen LLR Silver: Earned 100,000 credits (261,684)ESP LLR Silver: Earned 100,000 credits (109,285)PPS LLR Ruby: Earned 2,000,000 credits (2,326,228)PSP LLR Silver: Earned 100,000 credits (312,735)SoB LLR Silver: Earned 100,000 credits (154,103)SR5 LLR Silver: Earned 100,000 credits (139,377)SGS LLR Silver: Earned 100,000 credits (334,971)TRP LLR Amethyst: Earned 1,000,000 credits (1,232,766)Woodall LLR Silver: Earned 100,000 credits (441,996)321 Sieve Bronze: Earned 10,000 credits (22,692)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,124,289)PPS Sieve Double Bronze: Earned 100,000,000 credits (143,073,904)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Bronze: Earned 10,000 credits (49,211)TRP Sieve (suspended) Silver: Earned 100,000 credits (371,632)AP 26/27 Ruby: Earned 2,000,000 credits (4,115,774)GFN Ruby: Earned 2,000,000 credits (4,330,983)PSA Gold: Earned 500,000 credits (979,013)
Message 55598 - Posted: 18 Jun 2012 | 18:13:23 UTC - in response to Message 55595.
Last modified: 18 Jun 2012 | 18:13:59 UTC



Your need to be a troll and be obtuse as to what I'm saying baffles me. I'm obviously not computer illiterate, and my "frustration" is so that the hundreds of thousands of participants who have computers more like YOURS can have better crunching results.
......

If you want to settle for less - and defend it - then start your own thread.

First of all, I am the troll? Did you read your first post again and the way you wrote it? I just reacted to your way of posting.
Second, why should I start my own thread? you posted in the challenge topic and I reacted to that. If you want people to say 'yes sir, we agree' and nothing else, you should start your own forum and be a dictator there.

I still stand by my earlier posts and when you tell me I have to get a grip on my ego, please look in the mirror. Not everyone has to agree with you.
____________

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55601 - Posted: 18 Jun 2012 | 18:28:33 UTC

My gripe is about the massive red "Attention" listed in this project (and others)


AH!

Let me tell you the origin of that message, by way of a little story. :)

A little while ago I asked some of the other people here if, perhaps, it made sense to extend this challenge to 13 days and run the long World Record WUs instead.

In the politest way possible, much the way a parent would explain to their young child why it's not practical for them to own a pony, it was explained to me why that would be a really, really bad idea.

There was a time when PrimeGrid did run long challenges in the summer. Bad things happened. Lots of people who only crunch PrimeGrid part time come and crunch for the challenge. People who never crunched PrimeGrid before come and crunch for the challenge.

Prime95, which has the same core library inside as does our LLR program, is often used as a torture test by overclockers because it pushes the CPU as hard as possible. LLR does the same thing.

PrimeGrid's programs, as well as many other programs related to Prime numbers, often make CPUs run hotter than any other software can.

When you combine summer's heat, people who have never crunched here before, a program that pushes a CPU harder than it's ever been pushed before, and the fact that many people don't even know you're supposed to blow the dust out of your computer every now and then (let alone the much worse effects that tobacco smoke has on computer cooling), they had a rather undesirable effect.

Hardware failures. Not computation errors, but damaged hardware.

Bad stuff.

So now we put those big, red, blinking (ok, html doesn't do blinking, but if it could, those warning be be blinking, and Klaxons would be sounding, too) warning messages adorn every challenge description. And we don't run long challenges in the summer.

Mike
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Ken_g6Project donor
Volunteer developer
Avatar
Send message
Joined: 4 Jul 06
Posts: 854
ID: 3110
Credit: 151,479,047
RAC: 93,727
Found 2 primes in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Gold: Earned 500,000 credits (987,927)Cullen LLR Amethyst: Earned 1,000,000 credits (1,125,935)ESP LLR Amethyst: Earned 1,000,000 credits (1,193,202)Generalized Cullen/Woodall LLR Gold: Earned 500,000 credits (767,102)PPS LLR Jade: Earned 10,000,000 credits (12,185,834)PSP LLR Amethyst: Earned 1,000,000 credits (1,088,102)SoB LLR Ruby: Earned 2,000,000 credits (3,860,053)SR5 LLR Amethyst: Earned 1,000,000 credits (1,145,068)SGS LLR Gold: Earned 500,000 credits (998,185)TPS LLR (retired) Bronze: Earned 10,000 credits (19,376)TRP LLR Amethyst: Earned 1,000,000 credits (1,834,402)Woodall LLR Amethyst: Earned 1,000,000 credits (1,156,424)321 Sieve Ruby: Earned 2,000,000 credits (2,546,090)Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (8,584,236)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,461,309)PPS Sieve Emerald: Earned 50,000,000 credits (79,291,322)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Silver: Earned 100,000 credits (352,129)TRP Sieve (suspended) Gold: Earned 500,000 credits (776,202)AP 26/27 Turquoise: Earned 5,000,000 credits (5,851,049)GFN Sapphire: Earned 20,000,000 credits (21,100,212)PSA Ruby: Earned 2,000,000 credits (4,154,617)
Message 55604 - Posted: 18 Jun 2012 | 18:48:06 UTC - in response to Message 55601.

Maybe you should add a more direct statement about that to the "Attention" section. Something like:

Systems without sufficient cooling can and have damaged themselves running PrimeGrid tasks!
____________

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55606 - Posted: 18 Jun 2012 | 19:19:16 UTC - in response to Message 55596.

Hi Scott.

Thank you for that. I had started off on the wrong foot, mainly, it's a monday, and secondly, well, I'm frustrated at inefficiencies, and lastly, sometimes I'm an a-hole. Since my machine is made out of unicorns and mogwai and it's extra special long time to me, I cringe when things slow to a crawl and time shoots up. It's why I upgrade/renew consistently.

Now recognize, I'm 4 core, 3.8ghz, liquid cooled, 16gig ram, 4 terabytes of sata 10k rpm hdds, an HD5850 and an HD5450, and while at full crunch of 90% resources in boinc settings I'm still whipping through non lag video editing though Premiere, or running Photoshop 64, Illustrator, Fireworks, Dreamweaver and all my interwebby stuff side by side lag free with all three monitors ablaze.

So my gripes about WU's are in their time to complete - not a lack of resources to process them whatever they may be (aside from specific hardware).

I don't think it's necessary to have the WU's programmed for the 1% of the machines that can handle them when it's the 99% actually doing the bulk of the work - at a snail's pace.

Of course I still crunch away (for days) at Seti@Home and I'm used to seeing times that take hours (and days) to complete. I'm frustrated and have said so to them too.

I just don't think these long completion times are necessary.

I have actually seen crunching on the android and though it's "impractical" it's pretty darned cool. As the multi-cores become more prolific it will be interesting to see mobile folding and such... battery life will blow though, and you'll have a nice hand warmer in the winter, but still, it's nifty.

But just because a thing CAN be done, doesn't mean it SHOULD be implemented system wide.

For instance, the current June 20 Genefer challenge is pretty much like Prime telling me "oh in order for you to participate you have to ditch your iPhone for an Android phone. Of sure, you can use the iPhone like the other 99% of the people on Prime and maybe complete one WU, but we didn't code for that, we coded for Android so you could complete like ten of them.."
Huh?

That said though, here's a case n point of what's usefull...
Milkyway@home occasionally releases WU's that are single units but actually get processed on all the cores available at once - rather than the default one WU per core which they normally release (not including GPU stuff) which take 5 times as long. Sure I did four WU's simultaneously, but at the price of an extra 5 hours. An all core WU was awesome to see.

THAT is the stuff I'm talking about.

Now I have two ATI/AMD HD5450's, they have some parallel processing but not double precision, so though they crunch better than the CPU, they are still slow and left out of some WU's. I have one HD5850 and it beats the snot out of the CPU and the other cards.

Milkyway runs on the CPu and the 5850, but not the 5450's- but other projects use all three as they don't "rely" on the double precision. So I "config" boinc and crunching works "okay" on all projects.

But then we come to other projects like Prime's Genefer 1.07 (or seti, or rosetta, other sieves, etc).
At 19 hours in I'm sitting at 89.5% for Genefer 1.07. I've got a PPS 1.38 here on my Ati completed in 44 minutes. It's like watching an oven vs a microwave.

Now sure, I could go buy an nVidia and do this in 2.5 hours or so, but the hardware doesn't help my regular computer needs so I'm left with ATI.
Nor do I have the budget to turn around in six months and buy the next gen card because the coders are bored with the "old gear" that people just managed to pick up now. I'm sick of it in the computer world and in the software world, it's impractical - planned obsolescence.

See a 64bit CPU is double precision native, so if all CPU WU projects could thread or split over all the cores at once, you would improve crunch time and get closer to the GPU capabilities of the more higher end participants... It's not perfect, but it's more efficient. This would allow a normal computer owner who is participating to have faster results - so long as they are 64bit.
So Genefer could use all cores at once and be done quickly...

I know it's preference, and people "set it and forget it" and that's how they treat most boinc participation, but I like to see progress, and the faster the better. So gearing my machine up, also gears up my project credits. Fun.
But then comes "ooohh a challenge" and upon reading, "oh, not for me then. Well Crap. *sigh*"

So I get that there should be "minimums", and I know there's issues coding for different platforms, and I know you cannot please everyone all of the time.

All I'm saying is that it would be nice to hear someone say, yeah we should look into the fact that "everyone" is NOT using what we're coding for, and we'd have better results faster if we changed our strategies to work with what we have available to us. No one wants to participate in the fun stuff when it's always out of reach... We try for a bit to catch up, but the bar keeps rising.

Participation only goes as far as accessibility... I'm an exception, short of a super computer, I have 90% of what's needed for excellent crunching so it's frustrating to see a bar get raised consistently, and little progress in efficiency for the crunchers you have.

My question is why would someone code for an android phone instead of coding for ATI?

There's a coding disconnect here, and it should be addressed.
That's all I'm trying to say.

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55607 - Posted: 18 Jun 2012 | 19:31:33 UTC - in response to Message 55601.

Hi Mike, you right about the heat.. hence the liquid cooling lol.

I picked up an Antec Coolermaster KUHLER H2O thingamabob, have to strip your mobo down and mount the pump to the regular cooler pegs, but the unit fits completely inside, and simply dual fans out the back... I'm at full tilt right now and only 42 C cpu (40 decibels, 32C case). Room temp is 28c (sun's beating in the window atm)

So I DO upgrade sometimes just for Boinc work....

Profile j.sheridanProject donor
Send message
Joined: 21 Mar 11
Posts: 437
ID: 91622
Credit: 679,433,675
RAC: 0
321 LLR Ruby: Earned 2,000,000 credits (3,002,360)Cullen LLR Ruby: Earned 2,000,000 credits (3,040,127)PPS LLR Ruby: Earned 2,000,000 credits (4,146,102)PSP LLR Turquoise: Earned 5,000,000 credits (5,432,836)SoB LLR Turquoise: Earned 5,000,000 credits (5,623,361)SR5 LLR Gold: Earned 500,000 credits (640,957)SGS LLR Ruby: Earned 2,000,000 credits (4,053,556)TRP LLR Ruby: Earned 2,000,000 credits (3,826,598)Woodall LLR Ruby: Earned 2,000,000 credits (3,123,627)321 Sieve Silver: Earned 100,000 credits (237,768)Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (265,102,350)PPS Sieve Double Silver: Earned 200,000,000 credits (292,642,933)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,801,812)GFN Emerald: Earned 50,000,000 credits (82,759,285)
Message 55608 - Posted: 18 Jun 2012 | 20:23:22 UTC - in response to Message 55606.

I had started off on the wrong foot

ain't that the truth


So my gripes about WU's are in their time to complete

If you can come up with a faster way to process such enormous numbers then I'm sure everyone would love to hear it.

Those of us who've been using boinc since the late 90s remember just how long it took to crunch through the most basic of workunits so what's being achieved now is nothing short of amazing.

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55609 - Posted: 18 Jun 2012 | 21:14:02 UTC - in response to Message 55606.

I don't think it's necessary to have the WU's programmed for the 1% of the machines that can handle them when it's the 99% actually doing the bulk of the work - at a snail's pace.


Ug. Please don't try to equate this to the 1% type of protests. This really isn't like that at all.

I just don't think these long completion times are necessary.


I disagree. Or, rather, if you have a practical solution as to how to improve the situation, please say so. Besides the obvious lack of an ATI version of Genefer (which would still require double precision), I think we're doing the best we can.

I have actually seen crunching on the android and though it's "impractical" it's pretty darned cool. As the multi-cores become more prolific it will be interesting to see mobile folding and such... battery life will blow though, and you'll have a nice hand warmer in the winter, but still, it's nifty.


I'm waiting for someone to complain that BOINC killed their phone. I'm guessing that anyone crunching on their phone pretty much must be doing so with the unit plugged into the wall for power. I can't imagine crunching on batteries, regardless of whether you're crunching on a phone, a laptop, or a desktop connected to a UPS.

My biggest concern is that mobile devices were definitely not designed with crunching in mind. It's doubtful that their cooling -- such as it is -- was designed to deal with the CPU running full blast 24/7. Then consider that many people like their phones enough to put them in a nice protective case -- which I'm sure, in many cases, acts as an excellent thermal insulator. Rather than "hand warmer", I'd be thinking more along the lines of the classic "Mission Impossible" line: "This cellphone will self-destruct in 5 seconds..."


For instance, the current June 20 Genefer challenge is pretty much like Prime telling me "oh in order for you to participate you have to ditch your iPhone for an Android phone.


So, what's the better alternative? Never run GFN challenges? Part of the purpose of a challenge is to get people excited about crunching (because we've all got a competitive streak in us), and to introduce people to new areas of mathematical computing. There's a LOT of projects here, with lots of different math concepts, and we do like to spread the challenges around to all the projects.


That said though, here's a case n point of what's usefull...
Milkyway@home occasionally releases WU's that are single units but actually get processed on all the cores available at once


That's actually not a good example. Running multi-core decreases the WU time, but does NOT increase the amount of work done.

It's theoretically possible to create a version of Genefer that runs on multiple CPU cores.

With such a program, on my Core2Quad, I could possible crunch one of these WUs in about 13 hours instead of 52 hours.

But I would be crunching one WU instead of four WUs. As things stand now, at the end of 52 hours, either way, I've completed 4 WUs.

Yes, there's absolutely an advantage to having a mutlicore program for a challenge, since, using the same numbers, in a 72 hour challenge you could complete 5 WUs (which would take 65 hours) vs the 4 WUs I could complete right now. But a new and more complex version of the program would need to be written -- and that program would really only be advantageous for this challenge. For normal, non-challenge computing the total throughput (described, for example, in credits/hour) should be close to identical with either version of the program.

There would be a lot of work to produce a product of limited usefulness.

Now I have two ATI/AMD HD5450's, they have some parallel processing but not double precision


This is the deal with double precision. It's possible, in theory, to use single precision GPUs.

Although I haven't tested it, ALL OTHER THINGS BEING EQUAL, a single precision version of Genefer should be four times slower than the double precision version.

More or less. That equation is actually pretty complex.

Now, with Nvidia, there wasn't a good reason to even think about a single precision application. ALL of Nvidia's newer GPUs are double precision, so creating a single precision version of the program would A) be four times slower and B) only run on older, slower GPUs -- which would further compound the slowness. Again, that doesn't seem like a particularly good place to spend your programming resources.

However, the recent Kepler GPUs -- the GTX 600 series -- seriously underperform at double precision, even though their single precision performance is awesome. (I originally wrote that last sentence with the word "SUCK" rather than "seriously underperform", but that was a bit of an exaggeration.) The GTX 680 is, in double precision, slower than the previous generation GTX 580, but the performance is still pretty good. But the crippled DP performance of the GTX 680, especially if it indicates a future trend for Nvidia of downplaying DP in favor of hardware that enchances game performance (that's what they're built for, after all), has gotten people thinking about whether a single precision program might, somewhere down the road, make sense.

That's all for Nvidia. For ATI, the situation is a little different because, as I understand it (I don't currently own an ATI GPU so I'm not current on the details of their product lines) some of their current GPUs support DP and some don't. That does complicate matters, so creating a single precision version of an ATI app might make sense.

But then we come to other projects like Prime's Genefer 1.07 (or seti, or rosetta, other sieves, etc).
At 19 hours in I'm sitting at 89.5% for Genefer 1.07. I've got a PPS 1.38 here on my Ati completed in 44 minutes. It's like watching an oven vs a microwave.


Yup. You can bake that potato for an hour or two in a conventional oven, or you can microwave it for 10 minutes. But if you don't own a microwave and you want a baked potato, you're going to have to wait.

Same thing here. If you want to do a primality test on a 2.9 million digit number (that's where the short WUs are now), and you don't have a microwave, it's going to take a long time to do.

Cooking a baked potato in a conventional over for 5 minutes and expecting it to be fully cooked isn't realistic. Likewise, with the size numbers we're crunching, you either need a fast GPU or you have to wait. (As Scott mentioned, there are smaller numbers available to crunch on PRPNet but it's not practical at this time to make those available on BOINC, unfortunately. )

Nor do I have the budget to turn around in six months and buy the next gen card because the coders are bored with the "old gear" that people just managed to pick up now. I'm sick of it in the computer world and in the software world, it's impractical - planned obsolescence.


I think you've got this part wrong. Genefer does NOT require latest generation hardware. It will run on many Nvidia GPUs that are three years old (the GTX 200 series), and almost any Nvidia GPU that's less than 2 years old (the GTX 400 series, or better). And those restrictions don't exist because we *want* to be exclusive. That's when Nvidia started manufacturing hardware that could run this test efficiently.

My question is why would someone code for an android phone instead of coding for ATI?


Not to win any challenges, that's for sure!

____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 55612 - Posted: 18 Jun 2012 | 21:28:38 UTC - in response to Message 55604.

Maybe you should add a more direct statement about that to the "Attention" section. Something like:

Systems without sufficient cooling can and have damaged themselves running PrimeGrid tasks!



That's a good idea. I'd rather run the risk of scaring off a few crunchers in order to emphasize how important it is to keep the dust cleared out of your computer, especially in the summer.

Done.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55613 - Posted: 18 Jun 2012 | 21:28:50 UTC - in response to Message 55608.

j.sheridan

yeah as I said, I've followed and crunched for quite some time too, and yes, it is fantastic what we can do today versus "yesteryear"

I'm not knocking the efforts of the people who are paid or donate their time implementing the WU's. But maybe I need another analogy to descibe what I'm saying.

The Hubble uses 486 processor, and it's "just" a big honking camera with control thrusters and some communication gear. They aren't writing new software to work with a P4 and trying to upload that to the hubble - they make what they need for the new gear to be lean and mean and four-eighty-sixy. The coder doesn't tell nasa to upgrade, it's not practical.

Win 7 isn't any "different" than windows 98, it's just more flash and pizzaz. It still connects to a network, the internet, a monitor, sound card, etc. Of course things have changed under the hood - 32bit vs 64bit, better use of ram, more options for ram, multitasking, etc.
The best retail computer made at the end of XP could barely run Vista / Win 7 because of all the special effects and other bloat. So we had to upgrade like everything just to turn win7 on.

So the "average" user is who we need to code for. 4 gigs of ram, a dual core and no double precision gpu. The average user doesn't have a clue about command line WU's. Or the skills to build a linux box just to run as an appliance (and still no proper nVidia linux support lol).

We have people coding for an android OS, but there's not a lot of practicality there, to the point where why couldn't they use that time to create code to use multiple cores simultaneously, or an ati gpu? That's a sticking point I'm really confused over.

Put the resources into where the majority of computing is. It's fine to have the occasional long WU, but all project Wu's are STILL taking too "long" and though we are "doing more complex WU's" we're not seeing a time gain - until we brute force throw the "newest" hardware at it - and you have to buy just the right hardware for just the right project - and the coders just keep upping the anty by coding for the best hardware - not the average.

Most WU's still take half a day, a day, two... on a regular machine. Some projects only code nVidia, some so all ATI people are out for that.

All the multiple cores seem to do is allow multiple WU's simultaneously, rather than quickly process one at a time spread over the whole processor.

It's still taking god awful long to complete a task on a normal generic computer - just like it did in yesteryear.

If you can come up with a faster way to process such enormous numbers then I'm sure everyone would love to hear it.


I've mentioned some projects like Milkyway@home already span a WU over all
cores at once... and it's FASTER per WU than (in my case) four separate Wu's over each core.
Add coding for ATI when it's missing so that the project can benefit from those WU's.

I don't have the answers or the skill set, but I have the common sense to make suggestions, and that's what I'm trying to do... that's why I'm attempting a dialogue. I'd love to hear from a coder on a project to address some of the questions to see what's possible - or have a FAQ somewhere as to why "something isn't".

I donate my skills all the time, but I don't want my work to be inefficient - or seem to be... And since I don't know everything, I welcome ideas and input to try to streamline my work so I have more time for other things... I'm sure a coder would feel the same way.

one would think project directors would love more units returned, at at the end of the day would you rather make WU's for all the tens of supercomputers out there? or for the millions of average users like these projects came about to utilize?

Scott BrownProject donor
Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Avatar
Send message
Joined: 17 Oct 05
Posts: 1923
ID: 1178
Credit: 6,441,245,511
RAC: 2,828,016
Discovered the World's First base 116 Generalized Cullen prime!!!Discovered 16 mega primesEliminated 7 conjecture "k"sDiscovered 1 Sophie Germain pairDiscovered 1 Fermat divisor2012 Tour de Primes highest prime count2012 Tour de Primes most Mountain Stage primes2015 Tour de Primes highest prime count2016 Tour de Primes highest prime countFound 23 primes in the 2018 Tour de PrimesFound 1 mega prime in the 2018 Tour de PrimesFound 2 primes in the 2018 Tour de Primes Mountain Stage2019 Tour de Primes highest prime countFound 22 primes in the 2019 Tour de Primes321 LLR Double Bronze: Earned 100,000,000 credits (100,829,118)Cullen LLR Double Bronze: Earned 100,000,000 credits (103,870,990)ESP LLR Double Bronze: Earned 100,000,000 credits (137,499,413)Generalized Cullen/Woodall LLR Double Bronze: Earned 100,000,000 credits (108,461,080)PPS LLR Double Silver: Earned 200,000,000 credits (384,802,005)PSP LLR Double Bronze: Earned 100,000,000 credits (108,003,110)SoB LLR Double Bronze: Earned 100,000,000 credits (135,747,083)SR5 LLR Double Silver: Earned 200,000,000 credits (201,224,339)SGS LLR Double Bronze: Earned 100,000,000 credits (161,503,178)TPS LLR (retired) Silver: Earned 100,000 credits (235,439)TRP LLR Double Bronze: Earned 100,000,000 credits (121,443,822)Woodall LLR Double Bronze: Earned 100,000,000 credits (101,447,725)321 Sieve Double Silver: Earned 200,000,000 credits (203,510,966)Cullen/Woodall Sieve (suspended) Emerald: Earned 50,000,000 credits (83,794,448)Generalized Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (285,139,652)PPS Sieve Double Ruby: Earned 2,000,000,000 credits (2,108,274,845)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Double Silver: Earned 200,000,000 credits (203,523,358)TRP Sieve (suspended) Double Silver: Earned 200,000,000 credits (201,489,157)AP 26/27 Double Bronze: Earned 100,000,000 credits (135,790,733)GFN Double Amethyst: Earned 1,000,000,000 credits (1,295,584,459)PSA Double Silver: Earned 200,000,000 credits (259,058,048)
Message 55614 - Posted: 18 Jun 2012 | 21:35:06 UTC


Mike provided a very thorough reply, so I won't repeat his responses.

On the final point regarding Android apps developed before ATI GPU apps, I should point out that a coder in Poland who works with Android apps used the open source code for the CW sieve (as well as similar open source project code from about 5 other projects) to develop an Android port of BOINC as well as Android versions of the apps. This was done (as far as I know) independently of any project.

I know that PG has always been eager to develop ATI GPU, but generally the volunteer programmers that have done the work have not been able to do so as readily as for CUDA. If anyone would like to give it a try, I am sure that their assistance would be most welcome! :)

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55616 - Posted: 18 Jun 2012 | 21:47:29 UTC - in response to Message 55609.

Thanks Mike,

that was a great run down!

I concur about nVidia having many more DP core options, and I'm only ATI due to some software I need to run is less cranky with ATI. Ati should step up it's game, but that's another rant...

However, the milkyway multicore spread on a single WU did outperform completed units compared to multi WU's on each core.

I'm not sure exactly why, but my performance tools indicated it was due to the bandwidth between the cores helped speed up the calculations of each and it used the cache much more efficiently... where the single "per core" WU's flooded the cache with r/w and didn't use the bandwidth between cores at all - except for the OS and other software which did and of course interrupted the WU's processes quite frequently.

In the end, the multicore WU strategy seemed to duplicate the GPU's parallel abilities more closely.

So yes, of course I'm happy I'm able to crunch what I can, I just have to always think it could be "better" but not necessarily "cost more", and wish I could trounce a few contests with all the tools I have - without having to buy more gear or stare at the slowly increasing complete percentage like the mystical joy we all have of watching the little squares during a hard drive defrag... lol

Profile j.sheridanProject donor
Send message
Joined: 21 Mar 11
Posts: 437
ID: 91622
Credit: 679,433,675
RAC: 0
321 LLR Ruby: Earned 2,000,000 credits (3,002,360)Cullen LLR Ruby: Earned 2,000,000 credits (3,040,127)PPS LLR Ruby: Earned 2,000,000 credits (4,146,102)PSP LLR Turquoise: Earned 5,000,000 credits (5,432,836)SoB LLR Turquoise: Earned 5,000,000 credits (5,623,361)SR5 LLR Gold: Earned 500,000 credits (640,957)SGS LLR Ruby: Earned 2,000,000 credits (4,053,556)TRP LLR Ruby: Earned 2,000,000 credits (3,826,598)Woodall LLR Ruby: Earned 2,000,000 credits (3,123,627)321 Sieve Silver: Earned 100,000 credits (237,768)Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (265,102,350)PPS Sieve Double Silver: Earned 200,000,000 credits (292,642,933)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,801,812)GFN Emerald: Earned 50,000,000 credits (82,759,285)
Message 55617 - Posted: 18 Jun 2012 | 22:00:10 UTC - in response to Message 55613.


they make what they need for the new gear to be lean and mean and four-eighty-sixy. The coder doesn't tell nasa to upgrade, it's not practical.

Primegrid code is already lean and mean - including some hand written assembler. I think you are failing to comprehend the scale of the numbers being looked for.

Besides, if you want fast results why don't you just run PPS ? Then you can have an update every 10-15 minutes.

Axiom
Send message
Joined: 7 Jan 12
Posts: 8
ID: 126261
Credit: 5,342,299
RAC: 0
SoB LLR Bronze: Earned 10,000 credits (36,495)Woodall LLR Bronze: Earned 10,000 credits (14,130)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (19,682)PPS Sieve Turquoise: Earned 5,000,000 credits (5,181,227)GFN Bronze: Earned 10,000 credits (53,052)
Message 55618 - Posted: 18 Jun 2012 | 22:08:47 UTC - in response to Message 55617.

Primegrid code is already lean and mean - including some hand written assembler. I think you are failing to comprehend the scale of the numbers being looked for.

Besides, if you want fast results why don't you just run PPS ? Then you can have an update every 10-15 minutes.


I get the scale. I can't hold the number in my head, but I get that they are mind bogglingly large... The big honking numbers are my computer's problem.

I'd just like to participate in all projects that I can is all, with all the tools available - as timely as possible - and allow other projects relatively equal time as well.

Profile j.sheridanProject donor
Send message
Joined: 21 Mar 11
Posts: 437
ID: 91622
Credit: 679,433,675
RAC: 0
321 LLR Ruby: Earned 2,000,000 credits (3,002,360)Cullen LLR Ruby: Earned 2,000,000 credits (3,040,127)PPS LLR Ruby: Earned 2,000,000 credits (4,146,102)PSP LLR Turquoise: Earned 5,000,000 credits (5,432,836)SoB LLR Turquoise: Earned 5,000,000 credits (5,623,361)SR5 LLR Gold: Earned 500,000 credits (640,957)SGS LLR Ruby: Earned 2,000,000 credits (4,053,556)TRP LLR Ruby: Earned 2,000,000 credits (3,826,598)Woodall LLR Ruby: Earned 2,000,000 credits (3,123,627)321 Sieve Silver: Earned 100,000 credits (237,768)Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (265,102,350)PPS Sieve Double Silver: Earned 200,000,000 credits (292,642,933)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,801,812)GFN Emerald: Earned 50,000,000 credits (82,759,285)
Message 55623 - Posted: 18 Jun 2012 | 22:22:13 UTC - in response to Message 55618.


I'd just like to participate in all projects that I can is all, with all the tools available - as timely as possible - and allow other projects relatively equal time as well.

except for GFN world record you can, it's just that some of those projects will take days or even weeks to run. Sure it's nice to see the credits ticking along with PPS but there's also something satisfying about successfully completing monster tasks like SoB.

Profile Doc NoProject donor
Avatar
Send message
Joined: 2 Jan 09
Posts: 132
ID: 33689
Credit: 613,412,735
RAC: 182,882
Discovered 3 mega primes321 LLR Ruby: Earned 2,000,000 credits (4,921,665)Cullen LLR Ruby: Earned 2,000,000 credits (2,917,119)ESP LLR Ruby: Earned 2,000,000 credits (2,517,704)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (4,717,201)PPS LLR Sapphire: Earned 20,000,000 credits (29,285,636)PSP LLR Jade: Earned 10,000,000 credits (10,362,000)SoB LLR Sapphire: Earned 20,000,000 credits (21,624,492)SR5 LLR Turquoise: Earned 5,000,000 credits (9,750,498)SGS LLR Ruby: Earned 2,000,000 credits (4,378,393)TRP LLR Ruby: Earned 2,000,000 credits (3,223,165)Woodall LLR Ruby: Earned 2,000,000 credits (4,254,369)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (360,397)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,801,060)PPS Sieve Emerald: Earned 50,000,000 credits (71,565,350)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Ruby: Earned 2,000,000 credits (4,364,469)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,845,234)AP 26/27 Turquoise: Earned 5,000,000 credits (5,486,351)GFN Double Silver: Earned 200,000,000 credits (422,785,843)PSA Silver: Earned 100,000 credits (246,526)
Message 62379 - Posted: 8 Feb 2013 | 19:11:51 UTC - in response to Message 55608.

Those of us who've been using boinc since the late 90s remember just how long it took to crunch through the most basic of workunits so what's being achieved now is nothing short of amazing.


No kidding. I wish I would have taken pictures of offices full of machines running the SETI at Home screensaver :)

Profile Doc NoProject donor
Avatar
Send message
Joined: 2 Jan 09
Posts: 132
ID: 33689
Credit: 613,412,735
RAC: 182,882
Discovered 3 mega primes321 LLR Ruby: Earned 2,000,000 credits (4,921,665)Cullen LLR Ruby: Earned 2,000,000 credits (2,917,119)ESP LLR Ruby: Earned 2,000,000 credits (2,517,704)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (4,717,201)PPS LLR Sapphire: Earned 20,000,000 credits (29,285,636)PSP LLR Jade: Earned 10,000,000 credits (10,362,000)SoB LLR Sapphire: Earned 20,000,000 credits (21,624,492)SR5 LLR Turquoise: Earned 5,000,000 credits (9,750,498)SGS LLR Ruby: Earned 2,000,000 credits (4,378,393)TRP LLR Ruby: Earned 2,000,000 credits (3,223,165)Woodall LLR Ruby: Earned 2,000,000 credits (4,254,369)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (360,397)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,801,060)PPS Sieve Emerald: Earned 50,000,000 credits (71,565,350)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Ruby: Earned 2,000,000 credits (4,364,469)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,845,234)AP 26/27 Turquoise: Earned 5,000,000 credits (5,486,351)GFN Double Silver: Earned 200,000,000 credits (422,785,843)PSA Silver: Earned 100,000 credits (246,526)
Message 62380 - Posted: 8 Feb 2013 | 19:17:57 UTC - in response to Message 55609.
Last modified: 8 Feb 2013 | 19:22:14 UTC

However, the recent Kepler GPUs -- the GTX 600 series -- seriously underperform at double precision, even though their single precision performance is awesome. (I originally wrote that last sentence with the word "SUCK" rather than "seriously underperform", but that was a bit of an exaggeration.) The GTX 680 is, in double precision, slower than the previous generation GTX 580, but the performance is still pretty good.


That's very interesting.

Is there any place where hardware and performance is talked about in a concise matter?

Things like i5 vs i7 vs AMD?
Hyperthreading or not?
GPU families for ATI and Nvidia?

Or is it just scattered in random threads here in the forums? Because I have a shell with an LGA1155 socket that I'd like to slap a CPU in, and possibly a GPU to add to my computers and I'm having a heck of a time finding a consolidated place for this info.

And if it doesn't exist, perhaps it's time to create one - is there a wiki page somewhere I can start, or should I just try to start a thread here in the forum to start collecting info like what I quoted from you...

Profile Doc NoProject donor
Avatar
Send message
Joined: 2 Jan 09
Posts: 132
ID: 33689
Credit: 613,412,735
RAC: 182,882
Discovered 3 mega primes321 LLR Ruby: Earned 2,000,000 credits (4,921,665)Cullen LLR Ruby: Earned 2,000,000 credits (2,917,119)ESP LLR Ruby: Earned 2,000,000 credits (2,517,704)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (4,717,201)PPS LLR Sapphire: Earned 20,000,000 credits (29,285,636)PSP LLR Jade: Earned 10,000,000 credits (10,362,000)SoB LLR Sapphire: Earned 20,000,000 credits (21,624,492)SR5 LLR Turquoise: Earned 5,000,000 credits (9,750,498)SGS LLR Ruby: Earned 2,000,000 credits (4,378,393)TRP LLR Ruby: Earned 2,000,000 credits (3,223,165)Woodall LLR Ruby: Earned 2,000,000 credits (4,254,369)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (360,397)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,801,060)PPS Sieve Emerald: Earned 50,000,000 credits (71,565,350)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Ruby: Earned 2,000,000 credits (4,364,469)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,845,234)AP 26/27 Turquoise: Earned 5,000,000 credits (5,486,351)GFN Double Silver: Earned 200,000,000 credits (422,785,843)PSA Silver: Earned 100,000 credits (246,526)
Message 62381 - Posted: 8 Feb 2013 | 19:20:57 UTC - in response to Message 55616.
Last modified: 8 Feb 2013 | 20:03:47 UTC

I'm not sure exactly why, but my performance tools indicated it was due to the bandwidth between the cores helped speed up the calculations of each and it used the cache much more efficiently... where the single "per core" WU's flooded the cache with r/w and didn't use the bandwidth between cores at all - except for the OS and other software which did and of course interrupted the WU's processes quite frequently.


Yup, from a cache coherency perspective, I would expect one work unit on *multiple cores* vs. multiple work units on *individual* cores to be FAR more efficient.

EDIT: I just realized what I originally wrote didn't necessarily make sense - corrected with the ** parts

And perhaps I just answered my i5/i7 question - i7's have 8MB of L3 cache to 6MB for i5's...

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12940
ID: 53948
Credit: 195,717,000
RAC: 154,723
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,085,897)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Turquoise: Earned 5,000,000 credits (6,785,665)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,221,148)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (65,601,523)PSA Jade: Earned 10,000,000 credits (12,404,447)
Message 62382 - Posted: 8 Feb 2013 | 19:50:33 UTC - in response to Message 62381.

I'm not sure exactly why, but my performance tools indicated it was due to the bandwidth between the cores helped speed up the calculations of each and it used the cache much more efficiently... where the single "per core" WU's flooded the cache with r/w and didn't use the bandwidth between cores at all - except for the OS and other software which did and of course interrupted the WU's processes quite frequently.


Yup, from a cache coherency perspective, I would expect one work unit on one core vs. multiple work units on multiple cores to be FAR more efficient.


I don't have concrete numbers, but when Yves wrote the new (and as yet unreleased) versions of the Genefer programs, he did write a multi-core version. He said it was significantly less efficient, but I don't have any details.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Doc NoProject donor
Avatar
Send message
Joined: 2 Jan 09
Posts: 132
ID: 33689
Credit: 613,412,735
RAC: 182,882
Discovered 3 mega primes321 LLR Ruby: Earned 2,000,000 credits (4,921,665)Cullen LLR Ruby: Earned 2,000,000 credits (2,917,119)ESP LLR Ruby: Earned 2,000,000 credits (2,517,704)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (4,717,201)PPS LLR Sapphire: Earned 20,000,000 credits (29,285,636)PSP LLR Jade: Earned 10,000,000 credits (10,362,000)SoB LLR Sapphire: Earned 20,000,000 credits (21,624,492)SR5 LLR Turquoise: Earned 5,000,000 credits (9,750,498)SGS LLR Ruby: Earned 2,000,000 credits (4,378,393)TRP LLR Ruby: Earned 2,000,000 credits (3,223,165)Woodall LLR Ruby: Earned 2,000,000 credits (4,254,369)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (360,397)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,801,060)PPS Sieve Emerald: Earned 50,000,000 credits (71,565,350)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Ruby: Earned 2,000,000 credits (4,364,469)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,845,234)AP 26/27 Turquoise: Earned 5,000,000 credits (5,486,351)GFN Double Silver: Earned 200,000,000 credits (422,785,843)PSA Silver: Earned 100,000 credits (246,526)
Message 62384 - Posted: 8 Feb 2013 | 20:05:31 UTC - in response to Message 62382.

I don't have concrete numbers, but when Yves wrote the new (and as yet unreleased) versions of the Genefer programs, he did write a multi-core version. He said it was significantly less efficient, but I don't have any details.


Apologies for my original posting that was obtuse and a little confusing - to clarify and restate you say he found that one work unit on multiple cores was less efficient? Interesting - perhaps cache isn't as big an issue then.

And I wonder if it's one work unit on the entire CPU/all cores or multiple work units across a fraction of cores on the CPU.

So many variations to think about!

Message boards : Number crunching : Are hardware requirements too high?

[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.40, 1.94, 1.90
Generated 21 Oct 2019 | 10:25:31 UTC