Page 1 of 1

Project: 5753 (Run 10, Clone 66, Gen 11)

Posted: Thu Dec 04, 2008 12:04 pm
by Alphax
Right on cue :( This one hasn't uploaded yet ("Server does not have record of this unit. Will try again later.")...

Code: Select all

[09:33:52] 
[09:33:52] *------------------------------*
[09:33:52] Folding@Home GPU Core - Beta
[09:33:52] Version 1.19 (Mon Nov 3 09:34:13 PST 2008)
[09:33:52] 
[09:33:52] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[09:33:52] Build host: amoeba
[09:33:52] Board Type: Nvidia
[09:33:52] Core      : 
[09:33:52] Preparing to commence simulation
[09:33:52] - Looking at optimizations...
[09:33:52] - Files status OK
[09:33:52] - Expanded 96604 -> 489240 (decompressed 506.4 percent)
[09:33:52] Called DecompressByteArray: compressed_data_size=96604 data_size=489240, decompressed_data_size=489240 diff=0
[09:33:52] - Digital signature verified
[09:33:52] 
[09:33:52] Project: 5753 (Run 10, Clone 66, Gen 11)
[09:33:52] 
[09:33:53] Assembly optimizations on if available.
[09:33:53] Entering M.D.
[09:34:00] Will resume from checkpoint file
[09:34:06] Working on Protein
[09:34:11] Client config found, loading data.
[09:34:12] Starting GUI Server
[09:34:12] Resuming from checkpoint
[09:34:12] Verified work/wudata_09.log
[09:34:12] Verified work/wudata_09.edr
[09:34:12] Verified work/wudata_09.xtc
[09:34:12] Completed 9%
[09:54:30] Completed 10%
[10:14:14] Completed 11%
[10:33:57] Completed 12%
[10:53:07] Run: exception thrown during GuardedRun
[10:53:07] Run: exception thrown in GuardedRun -- Gromacs cannot continue further.
[10:53:07] Going to send back what have done -- stepsTotalG=10000000
[10:53:07] Work fraction=0.1297 steps=10000000.
[10:53:11] logfile size=12299 infoLength=12299 edr=0 trr=23
[10:53:11] - Writing 12835 bytes of core data to disk...
[10:53:11] Done: 12323 -> 4196 (compressed to 34.0 percent)
[10:53:11]   ... Done.
[10:53:11] 
[10:53:11] Folding@home Core Shutdown: EARLY_UNIT_END
[10:53:15] CoreStatus = 72 (114)
[10:53:15] Sending work to server
[10:53:15] Project: 5753 (Run 10, Clone 66, Gen 11)
Edit: uploaded it, finally. Let's see if someone with more capable hardware can get this back on track...

Re: Project: 5753 (Run 10, Clone 66, Gen 11)

Posted: Fri Dec 05, 2008 3:21 am
by bruce
This WU has been returned by you and one other person. Both had EUEs but at different percent-complete.

The guarded-run exceptions are a known problem which is getting a lot of attention by the development team. There's no way to know when a software update will be available to improve or fix this sort of situation.

Thank you for the report.

Re: Project: 5753 (Run 10, Clone 66, Gen 11)

Posted: Fri Dec 05, 2008 4:08 am
by Alphax
Aha. See, if the client collected & sent hardware info when it returned a not-quite-normal WU result, we'd have a much better idea if it's hardware limitations (as it probably is in my case - I suspect that my GPU simply doesn't have enough shaders or enough memory bandwidth) or a software bug (as seemed to be case when the 1.13 & 1.15 cores came out) :(

(And I was just expecting toTOW to give his usual "someone else completed it"...)

Re: Project: 5753 (Run 10, Clone 66, Gen 11)

Posted: Fri Dec 05, 2008 8:06 am
by bruce
The number of shaders will not determine which WUs will run and which will not. It determines speed. It's most easily understood if the shaders are actually creating an image on the screen rather than folding, but the same concept applies. If your screen resolution is X by Y, then there are X(times)Y individual pixels that must be evaluated. If one is passed to each shader it won't finish the screen so a second batch of pixels waits behind the first batch and the process continues until every pixel has been processed by a shader. Upgrade your card to more shaders, and each shader has to processs fewer pixels. Similarly, more memory bandwidth may make the frame rate of your games higher but it's still only a speed difference, not a difference that causes WUs to fail.

I vote for a software bug, but it could also be a hardware problem. If new software fixes it, then we'll know for sure (whenever that happens).

Re: Project: 5753 (Run 10, Clone 66, Gen 11)

Posted: Fri Dec 05, 2008 11:58 am
by toTOW
After a few hours, a third person was able to complete it successfully ;)