G80+ GPUs with Fermi/Kepler WUs

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

Moderators: just brew it!, farmpuma

G80+ GPUs with Fermi/Kepler WUs

Postposted on Sat Dec 08, 2012 12:05 pm

Starting last night (Friday 07 Dec) my GPU folder (Nvidia 9600GSO) began downloading work units from only the 171.64.65.105 server. The WUs from this server are OPENMMGPU units more commonly known as Fermi/Kepler work units. My GPU2 client (6.23) complained of a missing dll file. Research revealed the missing file was part of the GPU3 (6.41) client normally used for this type of work unit. Both of the clients are failing with the same looping error.

The issue is being discussed in this thread - http://foldingforum.org/viewtopic.php?f ... op#p231190 - on the folding forum.

Edit: Changed thread title from "FahCore_15 outdated core loop."
Last edited by farmpuma on Sun Dec 09, 2012 2:43 am, edited 1 time in total.
Reason: title change
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Re: FahCore_15 outdated core loop / update

Postposted on Sat Dec 08, 2012 9:01 pm

The CORE_OUTDATED looping error for G80 GPUs trying to fold Fermi/Kepler WUs was resolved this afternoon with the availability of FahCore_15 v2.25. Unfortunately a new problem was revealed for many of the G80s in the form of the GPU_MEMTEST_ERROR. This new problem is being discussed in this thread - http://foldingforum.org/viewtopic.php?f ... 93#p231393

Note: The title of this thread will most likely change.
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

GPU_MEMTEST_ERROR

Postposted on Sun Dec 09, 2012 3:24 am

Sifting through the posts in the two previously linked threads it seems apparent to me that the problem lies in FahCore_15. It was the last and only thing that changed. If it were an issue with the WUs then all the G80+ video cards would probably be failing. However, even some 8800GT cards are happily crunching WUs which are failing on my G92 9600GSO.

I suspect the GPU_MEMTEST is an arbitrary check and subsequent rejection of any card with less than one GB of memory. Were it an actual check of the ratio of memory to GPU cores the one GB 9800GTX would fail worse than my 384MB to 96core 9600GSO. And it is my understanding (I could be wrong) that the video card memory is only used as a buffer to shuffle data to and from the processing cores rather than hold the entire work unit data set. Again, I could be wrong, but it would seem to be the logical way to handle the data.

Note: A slightly modified version will be posted in the folding forum thread.
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: GPU_MEMTEST_ERROR

Postposted on Sun Dec 09, 2012 9:42 pm

From the third page of the folding forum thread linked in the above second post:

Well, judging by the posts between my post at the bottom of page one and now, it appears I was indeed wrong on several counts. Probably the best examples are the 8800GTs with 512MBs of memory which are happily crunching Fermi/Kepler work units. We with under memory powered G92 video cards will need to take solace in a job well done to this point and release our cards to light gaming pasture. The future of Nvidia video card folding is Fermi/Kepler (for the time being) and the future in now.

Thank you to Dr. Pande and crew for all your hard work and patience. And thank you for the temporary batch of FahCore_11 work units to help ease our transition.
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

sad epilogue

Postposted on Tue Dec 11, 2012 4:13 am

While reading the V7 installation guide I noticed a note mentioning the folding EoL for my still new in the box AMD 4830 listed as September 2011 and the actual end of supported WUs as June 2012. I really, really need to get a grip on this procrastination thing. Yeah, I'll get around to that one of these days.
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: G80+ GPUs with Fermi/Kepler WUs

Postposted on Tue Dec 11, 2012 9:45 am

I can procrastinate with the best of 'em! So far in my case, it has led to me never running GPU WUs at all. I may still give it a shot when they support GPU natively on Linux...
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37503
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: G80+ GPUs with Fermi/Kepler WUs

Postposted on Thu Dec 20, 2012 6:35 pm

I am the high emperor of procrastination, and all others are like knaves! I still have 15 year old games I haven't played, so take that! :P :roll:

Anyway, I like 'em GPU work units! They keep my toes warm at night. I don't know if that's really true, it just sounded like a fun thing to say.

But I am debating whether or not to slap my new 7970 into my Q6600 system for a few days before I order my new motherboard, CPU, and memory to see if I can make this old system howl a little more before I turn it into an HTPC or something. I'd probably run these under Frankie just for benchmarking purposes. It will depend on whether or not:

A) the card fits (it's a MONSTER, so appropriate for Frankie).
B) there are no hard conflicts with the Q6600 and P5KC motherboard.

The Fermi issue (with folding) was ultimately about 40% of my decision to go with an AMD card. The tipping point and other 60% was due to no Geforce cards supporting more than 4 monitors concurrently. Or else, I may have switched to that GTX 690. Or two. Who needs to eat, anyway...I'm fat enough as it is!
BIF
Gerbil Jedi
Gold subscriber
 
 
Posts: 1543
Joined: Tue May 25, 2004 7:41 pm


Return to TR Distributed Computing Effort

Who is online

Users browsing this forum: Exabot [Bot] and 4 guests