New 'easy' version of the V7 F@h client

Come join the... uh... er... fold.

Moderators: just brew it!, farmpuma

Re: New 'easy' version of the V7 F@h client

Postposted on Sat Apr 05, 2014 9:55 pm

Well the WUs from my CPU are registering...
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Sun Apr 06, 2014 2:07 pm

It wouldn't be the first time one of their GPU servers went down with a backlog of WUs to credit stuck on it. I'd still give it a few days then before posting in the F@H forums for them to check it.
Kougar
Gerbil Team Leader
 
Posts: 245
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 6:11 am

Well I'll keep checking but until I see it come though I'll hold off on doing any more folding. :-?
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 8:27 am

drfish wrote:... I'm thinking about building a basement GPU folding array ...

From what I've read, with the advent of the core 17 work units the Radeon GCN video cards now achieve folding PPD almost equal to Nvidia Fermi and Kepler cards of equal gaming performance. In general the Nvidia advantage is now a smallish percentage instead of the multiple amounts of PPD of the previous work unit cores.

The Kepler GPUs use considerably less electrical power at full load than equivalent Radeons and while I would expect the Maxwell GPUs to maintain a similar advantage in PPD with their even more eco friendly performance I haven't had time to search the folding forum for data to support my theory.

Motherboards with more than two PCI-e x16 slots tend to be quite a bit more expensive and long term three or four smaller core density GPU cards should use more electricity for non-folding circuitry than two higher core density cards for the same PPD. Although the higher current voltage regulators and increased quantities of memory on the higher core density cards will negate some of the difference. You might consider waiting for Maxwell GPUs with higher core densities than the 750 series.
Image Image
.* * M-51 * *. .The Whirlpool Galaxy.
farmpuma
Minister of Gerbil Affairs
Silver subscriber
 
 
Posts: 2311
Joined: Mon Mar 22, 2004 12:33 am
Location: Soybean field, IN, USA, Earth .. just a bit south of John .. err .... Fart Wayne, Indiana

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 8:52 am

farmpuma wrote:You might consider waiting for Maxwell GPUs with higher core densities than the 750 series.


Yeah, waiting for the so called "big Maxwell" seems to be the right call. Now to see if I ever get credited for the 24 hours of work my 780 put into those big WUs...
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 2:04 pm

drfish, if you're that antsy ya might as well post in their forum!

farmpuma wrote:From what I've read, with the advent of the core 17 work units the Radeon GCN video cards now achieve folding PPD almost equal to Nvidia Fermi and Kepler cards of equal gaming performance. In general the Nvidia advantage is now a smallish percentage instead of the multiple amounts of PPD of the previous work unit cores.

The Kepler GPUs use considerably less electrical power at full load than equivalent Radeons and while I would expect the Maxwell GPUs to maintain a similar advantage in PPD with their even more eco friendly performance I haven't had time to search the folding forum for data to support my theory.


Don't forget to look at the CPU overhead involved too! Core 17 projects suck up almost a core's worth of CPU overhead on NVIDIA cards... reminds me of when F@H first began GPU folding around 2008, maxing the GPU meant maxing out a full core just for driver overhead. Except Core 15 projects don't even use enough to register 1% by comparison, so I don't really know why Core 17 needs so much again. It is a big issue with AMD cards before GCN, but I don't know how much CPU overhead Core 17 still requires on GCN parts?
Kougar
Gerbil Team Leader
 
Posts: 245
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 2:15 pm

Kougar wrote:drfish, if you're that antsy ya might as well post in their forum!

farmpuma wrote:From what I've read, with the advent of the core 17 work units the Radeon GCN video cards now achieve folding PPD almost equal to Nvidia Fermi and Kepler cards of equal gaming performance. In general the Nvidia advantage is now a smallish percentage instead of the multiple amounts of PPD of the previous work unit cores.

The Kepler GPUs use considerably less electrical power at full load than equivalent Radeons and while I would expect the Maxwell GPUs to maintain a similar advantage in PPD with their even more eco friendly performance I haven't had time to search the folding forum for data to support my theory.


Don't forget to look at the CPU overhead involved too! Core 17 projects suck up almost a core's worth of CPU overhead on NVIDIA cards... reminds me of when F@H first began GPU folding around 2008, maxing the GPU meant maxing out a full core just for driver overhead. Except Core 15 projects don't even use enough to register 1% by comparison, so I don't really know why Core 17 needs so much again. It is a big issue with AMD cards before GCN, but I don't know how much CPU overhead Core 17 still requires on GCN parts?

I thought the Nvidia CPU hogging issue is with the older Core x15 and not the latest x17? I could be wrong though.
Image
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
Flying Fox
Gerbil God
 
Posts: 24291
Joined: Mon May 24, 2004 2:19 am

Re: New 'easy' version of the V7 F@h client

Postposted on Mon Apr 07, 2014 7:44 pm

Kougar wrote:drfish, if you're that antsy ya might as well post in their forum!


Maybe... But I went out bitching about how broken the SMP client was when they thought it was the greatest thing since sliced bread - not sure if it's worth getting all worked up over as soon as I put my toe back in the pool... I'll give it a few more days at least...
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 6:05 am

drfish wrote:So I've done two of the big ~70k WUs on my GPU now and the latest one finished last night but they aren't on my EOC stats yet? Is that normal?

Where are you seeing a GPU work unit with a ~70k points value? The big core 17 work units are listed on the Projects Summary as "OPENMM..." and the largest one I see is a little over 16K points if my Kfactor early return bonus assumption is correct. As far as I know the only work units with a points value in the 70K and higher range are the big adv SMP "GRO_A5" work units with the crazy 47 to 50 Kfactors.
farmpuma
Minister of Gerbil Affairs
Silver subscriber
 
 
Posts: 2311
Joined: Mon Mar 22, 2004 12:33 am
Location: Soybean field, IN, USA, Earth .. just a bit south of John .. err .... Fart Wayne, Indiana

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 6:37 am

I'll have to start the client again to get the details but I triple checked the number of digits at the time. I can see from the list that nothing like what I'm describing exists apparently so I guess I'm confused but I'm sure it said ~70k each and ~130k PPD (they were taking almost exactly 12hrs to complete) - my CPU was doing about 10K PPD on the units it was getting and my EOC stats for the couple days I was participating just don't line up... :-? Maybe I'm completely missing something in the new interface...? Can you detail what exactly the Kfactor is? It's related to the early completion bonus?
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 7:44 am

From the "Quick Return Bonus" FAQ page - http://folding.stanford.edu/home/faq/faq-points/

"The k-factor, a coefficient in awarding bonus points, is currently set to a baseline value of 0.75, but may vary depending on the scientific value of a project."

"The Folding@home software on your computer calculates Total Points as follows:

final_points = base_points * max(1, sqrt( k * deadline_length / elapsed_time))"

Not quite the multiplying factor I had assumed, but relatively close.

Note: You must be using your personal passkey to receive bonus points.

Edit: Using the p7613 work unit I guesstimated earlier and plugging and chugging with your 12 hour completion time I get a maximum points value of ~23,771. I apologize for my overly enthusiastic reaction to your return to folding with a GTX 780.

Edit 2: Going by what you recall seeing, it appears the new client may have an error in its PPD calculations.
farmpuma
Minister of Gerbil Affairs
Silver subscriber
 
 
Posts: 2311
Joined: Mon Mar 22, 2004 12:33 am
Location: Soybean field, IN, USA, Earth .. just a bit south of John .. err .... Fart Wayne, Indiana

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 11:01 am

I'll try to test again tonight and post some screen shots to confirm. Hopefully I snag another of the same WUs (not all of the ones my GPU processed were that many points - another reason I don't think I'm crazy).
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 11:23 am

drfish wrote:Image

25k PPD for my entire farm of dozens of computers... Crazy...[/i]


... not taking the bait :-)
leor
Maximum Gerbil
Gold subscriber
 
 
Posts: 4739
Joined: Wed Dec 11, 2002 6:34 pm
Location: NYC

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 11:36 am

Oh gawd no! The electrical consumption of Ellie's accessories doesn't leave much in the power budget for folding (especially in the summer when we need to use even more AC to fight the heat those things kick out)... :o
TR BBQ ~XI~ has come and gone but you can relive it by checking out this gallery!
drfish
Maximum Gerbil
Gold subscriber
 
 
Posts: 4498
Joined: Wed Jan 22, 2003 7:53 pm
Location: Zeeland, MI

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 11:50 am

Flying Fox wrote:I thought the Nvidia CPU hogging issue is with the older Core x15 and not the latest x17? I could be wrong though.


Nah, Core 15 projects don't register any CPU overhead, at least on nvidia stuff. If I get a Core 17 project then it sometimes maxes out an entire core though.
Kougar
Gerbil Team Leader
 
Posts: 245
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: New 'easy' version of the V7 F@h client

Postposted on Tue Apr 08, 2014 12:59 pm

drfish wrote:Oh gawd no! The electrical consumption of Ellie's accessories doesn't leave much in the power budget for folding (especially in the summer when we need to use even more AC to fight the heat those things kick out)... :o


I don't pay for electricity where I currently live, so it wouldn't really be a fair fight anymore, but nothing can ever or will ever compare to how damned close that contest was. I actually stopped folding avidly when they added the really silly point totals that made all the effort we put in with our huge farms seem pointless.

Not that that's what I was in it for, but it took the competitive fun out of it for me.

That's some serious equipment Ellie's got there.
leor
Maximum Gerbil
Gold subscriber
 
 
Posts: 4739
Joined: Wed Dec 11, 2002 6:34 pm
Location: NYC

Previous

Return to TR Distributed Computing Effort

Who is online

Users browsing this forum: No registered users and 2 guests