Folding Forum

Posted: **Wed Mar 24, 2010 4:37 pm**

Hello

Last night I set up a GPU client on my SBS2008 box. My GPU is a GTX260 running at stock speeds.

When I first setup my client I could not get any work for most of the evening. Today when I look at my client it tells me I have exceeded my EUE limit. Sure enough, my GPU client has been trying, and failing, to do Project 5786 (Run 10, Clone 54, Gen 28). Figuring I had a problem with my set up I turned off my client, rebooted the server and started the client. I successfully completed a GPU WU and I assumed I was in the clear. Unfortunately, my client has just downloaded the problem Project 5786 again, and again it crashes with a NANs detected error:

Code: Select all

###############################################################################
###############################################################################

Launch directory: C:\Users\pspain\FaH\GPU0
Executable: C:\Users\pspain\FaH\GPU0\gpu0.exe
Arguments: -gpu 0 -verbosity 9

[16:28:00] - Ask before connecting: No
[16:28:00] - User name: SiriusB_[OcUK] (Team 10)
[16:28:00] - User ID: C34613D10C4C717
[16:28:00] - Machine ID: 2
[16:28:00]
[16:28:00] Loaded queue successfully.
[16:28:00]
[16:28:00] - Autosending finished units... [March 24 16:28:00 UTC]
[16:28:00] Trying to send all finished work units
[16:28:00] + No unsent completed units remaining.
[16:28:00] - Autosend completed
[16:28:00] + Processing work unit
[16:28:00] Core required: FahCore_11.exe
[16:28:00] Core found.
[16:28:00] Working on queue slot 03 [March 24 16:28:00 UTC]
[16:28:00] + Working ...
[16:28:00] - Calling '.\FahCore_11.exe -dir work/ -suffix 03 -checkpoint 15 -ver
bose -lifeline 6708 -version 623'

[16:28:00]
[16:28:00] *------------------------------*
[16:28:00] Folding@Home GPU Core
[16:28:00] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:28:00]
[16:28:00] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:28:00] Build host: amoeba
[16:28:00] Board Type: Nvidia
[16:28:00] Core      :
[16:28:00] Preparing to commence simulation
[16:28:00] - Looking at optimizations...
[16:28:00] - Files status OK
[16:28:00] - Expanded 64655 -> 341507 (decompressed 528.1 percent)
[16:28:00] Called DecompressByteArray: compressed_data_size=64655 data_size=3415
07, decompressed_data_size=341507 diff=0
[16:28:00] - Digital signature verified
[16:28:00]
[16:28:00] Project: 5786 (Run 10, Clone 54, Gen 28)
[16:28:00]
[16:28:00] Assembly optimizations on if available.
[16:28:00] Entering M.D.
[16:28:06] Tpr hash work/wudata_03.tpr:  4041654140 1181850360 4156308722 355751
3678 3262153356
[16:28:06]
[16:28:06] Calling fah_main args: 14 usage=100
[16:28:06]
[16:28:07] Working on GRoups of Organic Molecules in ACtion for Science
[16:28:09] Client config found, loading data.
[16:28:09] mdrun_gpu returned
[16:28:09] NANs detected on GPU
[16:28:09]
[16:28:09] Folding@home Core Shutdown: UNSTABLE_MACHINE
[16:28:09] Starting GUI Server
[16:28:13] CoreStatus = 7A (122)
[16:28:13] Sending work to server
[16:28:13] Project: 5786 (Run 10, Clone 54, Gen 28)
[16:28:13] - Read packet limit of 540015616... Set to 524286976.
[16:28:13] - Error: Could not get length of results file work/wuresults_03.dat
[16:28:13] - Error: Could not read unit 03 file. Removing from queue.
[16:28:13] Trying to send all finished work units
[16:28:13] + No unsent completed units remaining.
[16:28:13] - Preparing to get new work unit...
[16:28:13] + Attempting to get work packet
[16:28:13] - Will indicate memory of 4094 MB
[16:28:13] - Detect CPU. Vendor: GenuineIntel, Family: 6, Model: 15, Stepping: 1
1
[16:28:13] - Connecting to assignment server
[16:28:13] Connecting to http://assign-GPU.stanford.edu:8080/
[16:28:14] Posted data.
[16:28:14] Initial: 43AB; - Successful: assigned to (171.67.108.21).
[16:28:14] + News From Folding@Home: Welcome to Folding@Home
[16:28:14] Loaded queue successfully.
[16:28:14] Connecting to http://171.67.108.21:8080/
[16:28:14] Posted data.
[16:28:14] Initial: 0000; - Receiving payload (expected size: 65167)
[16:28:16] - Downloaded at ~31 kB/s
[16:28:16] - Averaged speed for that direction ~56 kB/s
[16:28:16] + Received work.
[16:28:16] Trying to send all finished work units
[16:28:16] + No unsent completed units remaining.
[16:28:16] + Closed connections
[16:28:21]
[16:28:21] + Processing work unit
[16:28:21] Core required: FahCore_11.exe
[16:28:21] Core found.
[16:28:21] Working on queue slot 04 [March 24 16:28:21 UTC]
[16:28:21] + Working ...
[16:28:21] - Calling '.\FahCore_11.exe -dir work/ -suffix 04 -checkpoint 15 -ver
bose -lifeline 6708 -version 623'

[16:28:21]
[16:28:21] *------------------------------*
[16:28:21] Folding@Home GPU Core
[16:28:21] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:28:21]
[16:28:21] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:28:21] Build host: amoeba
[16:28:21] Board Type: Nvidia
[16:28:21] Core      :
[16:28:21] Preparing to commence simulation
[16:28:21] - Looking at optimizations...
[16:28:21] DeleteFrameFiles: successfully deleted file=work/wudata_04.ckp
[16:28:21] - Created dyn
[16:28:21] - Files status OK
[16:28:21] - Expanded 64655 -> 341507 (decompressed 528.1 percent)
[16:28:21] Called DecompressByteArray: compressed_data_size=64655 data_size=3415
07, decompressed_data_size=341507 diff=0
[16:28:21] - Digital signature verified

Anybody know if this WU is broken, or is my GPU incredibly picky? Either way, can I force my client to stop downloading that WU?

I would like to add that I have run this GPU for many weeks on the same system running the W7 RC with no errors of any kind, which is why I am dubious my GPU is to blame.

Thanks.

Posted: **Wed Mar 24, 2010 6:20 pm**

Weelcome to the foldingforum.org Etzeitet

There is no report for that WU from SirusB.... There are a couple of other clients who have reported problems with it but it has also been completed successfully by quite a few others.

This is unusual. I suspect that others (like you) may have had unreported problems, too. I see that it has been reassigned a lot more frequently than I ever expect to see. My only guess is that it's a picky WU which has found a number of picky GPUs.

Posted: **Wed Mar 24, 2010 7:17 pm**

Hi bruce,

Thank you for your welcome and reply.

I wish I had more WUs done on this system to reaffirm my GPU is OK. Unfortunately my record is 1 good 1 not so good!

Is there anything I can do to grab another WU? I don't want to have to wait several days before my client or the assignment server gets the message! :p I have tried deleting the queue and unit info files with no luck.

Posted: **Wed Mar 24, 2010 8:54 pm**

Quick update:

After nuking the entire GPU client folder and setting up from scratch, I managed to get a different WU [5765] and it is sailing through it with no issues.

5768 is awfully picky it seems! Here's hoping I can get a few WUs done before ending up with that one again!

Posted: **Tue Apr 06, 2010 7:56 pm**

Arrrrrgh!

Following on from my earlier thread: viewtopic.php?f=19&t=13936

My GPUs are still crapping out on certain work units. I haven't seen any 5768's recently but 5784 and 5786 crash in precisely the same way as in my previous thread. If it was just one GPU I would suspect a hardware problem, but both my GTX260s crash with these WUs.

I am struggling to believe it is a hardware issue as the two GTX260s are from two different manufacturers [Palit and Sparkle] and both GPUs run the 10xx and 60xx projects with no problems. The same machine also runs an SMP client with zero problems on the same machine. I run SBS 2008 and have used several different NVidia drivers with the same results. Currently running 197.13.

Specs:

Q6600 @ stock
4GB Kingston something or other RAM
Asus P5B Deluxe Motherboard
Palit GTX260 @ stock
Sparkle GTX260 @ stock
SBS 2008
197.13 NVIDIA drivers.

I wouldn't mind but frustratingly the client is quite happy to attempt the same bloody WU 5 times before shutting down for 24 hours, losing a large amount of folding time. Even after the 24 hours it can go and get the same WU! The only way I have found to get a new non-broken WU is to delete everything in the client folder except the config, dlls and exe and change the machineid in the config. If I am lucky I get a WU I can fold after the first attempt. Just now I had to delete everything, change ID and rerun the client 4 or 5 times!

Is there anything that can be done to selectively avoid troublesome WUs? I suspect not as no doubt it could be exploited, but you can imagine how irksome it can be. It is horribly ineffecient too as if I didn't watch my clients like a hawk they can go for days at a time before they did any work [which they did this weekend while I was away camping!]

Posted: **Sat Apr 10, 2010 12:36 am**

I am probably going to have to give up folding on my GPUs. I can't stand babysitting my clients this much, especially when the majority of WUs out there seem to be the troublesome 5784s and 5786s. This is not a hardware or software issue as far as I can see. Every other WU runs fine! These WUs are failing before they even get going. It is not instability, it is nothing to do with my GPUs being pushed more by these WUs since every other WU uses the same core without issue.

Really annoyed there is no system in place for the clients to automatically reject bad WUs, given I have a history spanning the last two weeks of EUEs on the same WUs. Check the server logs, I have probably been assigned twice as many WUs as I have managed to complete. Is there no way for someone to look at the differences between the above projects and the likes of 5781 etc that DO work without a flaw and work out if something is amiss. You have a thread that is 18 pages long with the same WUs cropping up with the same errors on hardware that is perfectly able to run other WUs.

Sorry for the rant, but I try hard to contribute to the project but lately running the GPU client is liking pulling teeth!

Posted: **Sat Apr 10, 2010 2:33 am**

Those WUs are more highly optimized than other projects. If that causes problems with your GPU overheating or drawing to much power or becoming unstable, I suggest you decrease the percentage of time that your GPU is folding using the environment variable FAH_GPU_IDLE. I would start with 20 and if that fixes it, try 10.

Posted: **Sat Apr 10, 2010 4:25 pm**

I appreciate your reply and have implemented the variable you suggested. However I can tell you both GPU clients picked up a 5786 and both EUEd. Luckily for me they both picked up new WUs they could fold.

I really doubt it is a power draw issue as the PSU is new and more than capable of supporting both GPUs. For reference it is a Corsair HX650. I have watched my voltages and they do not budge. It is a rock solid PSU. Also, since the WUs crash before they even start crunching overheating is not an issue.

If my WUs crashed mid-way through then I would accept a power or heat issue. Since the WU crashes at the start, the problem is elsewhere. Dozens of successful WUs done on the same system would suggest the problem isn't my system, unless you know of an issue with certain WUs, the GPU client and SBS2008 or certain versions of NVidia drivers.

Log from first GPU:

Code: Select all

# Windows GPU Console Edition #################################################
###############################################################################

                       Folding@Home Client Version 6.23

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: C:\Users\pspain\FaH\GPU0
Executable: C:\Users\pspain\FaH\GPU0\gpu0.exe
Arguments: -gpu 0 -verbosity 9

[16:05:58] - Ask before connecting: No
[16:05:58] - User name: SiriusB_[OcUK] (Team 10)
[16:05:58] - User ID: C34613D10C4C717
[16:05:58] - Machine ID: 15
[16:05:58]
[16:05:58] Loaded queue successfully.
[16:05:58]
[16:05:58] + Processing work unit
[16:05:58] - Autosending finished units... [April 10 16:05:58 UTC]
[16:05:58] Core required: FahCore_11.exe
[16:05:58] Trying to send all finished work units
[16:05:58] Core found.
[16:05:58] + No unsent completed units remaining.
[16:05:58] Working on queue slot 07 [April 10 16:05:58 UTC]
[16:05:58] - Autosend completed
[16:05:58] + Working ...
[16:05:58] - Calling '.\FahCore_11.exe -dir work/ -suffix 07 -checkpoint 15 -ver
bose -lifeline 9604 -version 623'

[16:05:58]
[16:05:58] *------------------------------*
[16:05:58] Folding@Home GPU Core
[16:05:58] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:05:58]
[16:05:58] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:05:58] Build host: amoeba
[16:05:58] Board Type: Nvidia
[16:05:58] Core      :
[16:05:58] Preparing to commence simulation
[16:05:58] - Looking at optimizations...
[16:05:58] - Files status OK
[16:05:58] - Expanded 64681 -> 341507 (decompressed 527.9 percent)
[16:05:58] Called DecompressByteArray: compressed_data_size=64681 data_size=3415
07, decompressed_data_size=341507 diff=0
[16:05:58] - Digital signature verified
[16:05:58]
[16:05:58] Project: 5786 (Run 5, Clone 10, Gen 80)
[16:05:58]
[16:05:58] Assembly optimizations on if available.
[16:05:58] Entering M.D.
[16:06:04] Tpr hash work/wudata_07.tpr:  1454933175 1927711105 1281420434 278658
3095 1540034925
[16:06:04]
[16:06:04] Calling fah_main args: 14 usage=90
[16:06:04]
[16:06:05] Working on GRoups of Organic Molecules in ACtion for Science
[16:06:06] Client config found, loading data.
[16:06:07] mdrun_gpu returned
[16:06:07] NANs detected on GPU
[16:06:07]
[16:06:07] Folding@home Core Shutdown: UNSTABLE_MACHINE
[16:06:10] CoreStatus = 7A (122)
[16:06:10] Sending work to server
[16:06:10] Project: 5786 (Run 5, Clone 10, Gen 80)
[16:06:10] - Read packet limit of 540015616... Set to 524286976.
[16:06:10] - Error: Could not get length of results file work/wuresults_07.dat
[16:06:10] - Error: Could not read unit 07 file. Removing from queue.
[16:06:10] Trying to send all finished work units
[16:06:10] + No unsent completed units remaining.
[16:06:10] - Preparing to get new work unit...
[16:06:10] + Attempting to get work packet
[16:06:10] - Will indicate memory of 4094 MB
[16:06:10] - Detect CPU. Vendor: GenuineIntel, Family: 6, Model: 15, Stepping: 1
1
[16:06:10] - Connecting to assignment server
[16:06:10] Connecting to http://assign-GPU.stanford.edu:8080/
[16:06:11] Posted data.
[16:06:11] Initial: 43AB; - Successful: assigned to (171.67.108.11).
[16:06:11] + News From Folding@Home: Welcome to Folding@Home
[16:06:11] Loaded queue successfully.
[16:06:11] Connecting to http://171.67.108.11:8080/
[16:06:12] Posted data.
[16:06:12] Initial: 0000; - Receiving payload (expected size: 47249)
[16:06:13] - Downloaded at ~46 kB/s
[16:06:13] - Averaged speed for that direction ~60 kB/s
[16:06:13] + Received work.
[16:06:13] Trying to send all finished work units
[16:06:13] + No unsent completed units remaining.
[16:06:13] + Closed connections
[16:06:18]
[16:06:18] + Processing work unit
[16:06:18] Core required: FahCore_11.exe
[16:06:18] Core found.
[16:06:18] Working on queue slot 08 [April 10 16:06:18 UTC]
[16:06:18] + Working ...
[16:06:18] - Calling '.\FahCore_11.exe -dir work/ -suffix 08 -checkpoint 15 -ver
bose -lifeline 9604 -version 623'

[16:06:18]
[16:06:18] *------------------------------*
[16:06:18] Folding@Home GPU Core
[16:06:18] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:06:18]
[16:06:18] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:06:18] Build host: amoeba
[16:06:18] Board Type: Nvidia
[16:06:18] Core      :
[16:06:18] Preparing to commence simulation
[16:06:18] - Looking at optimizations...
[16:06:18] DeleteFrameFiles: successfully deleted file=work/wudata_08.ckp
[16:06:18] - Created dyn
[16:06:18] - Files status OK
[16:06:18] - Expanded 46737 -> 252912 (decompressed 541.1 percent)
[16:06:18] Called DecompressByteArray: compressed_data_size=46737 data_size=2529
12, decompressed_data_size=252912 diff=0
[16:06:18] - Digital signature verified
[16:06:18]
[16:06:18] Project: 5765 (Run 12, Clone 244, Gen 105)
[16:06:18]
[16:06:18] Assembly optimizations on if available.
[16:06:18] Entering M.D.
[16:06:24] Tpr hash work/wudata_08.tpr:  4190163122 3436051368 3726458250 223932
2703 3824773853
[16:06:24]
[16:06:24] Calling fah_main args: 14 usage=90
[16:06:24]
[16:06:25] Working on Protein
[16:06:25] Client config found, loading data.
[16:06:25] Starting GUI Server
[16:07:03] Completed 1%
[16:07:41] Completed 2%
[16:08:19] Completed 3%
[16:08:56] Completed 4%
[16:09:34] Completed 5%
[16:10:12] Completed 6%
[16:10:50] Completed 7%
[16:11:28] Completed 8%
[16:12:06] Completed 9%
[16:12:44] Completed 10%
[16:13:27] Completed 11%
[16:14:05] Completed 12%
[16:14:43] Completed 13%
[16:15:28] Completed 14%

Log from second GPU:

Code: Select all

# Windows GPU Console Edition #################################################
###############################################################################

                       Folding@Home Client Version 6.23

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: C:\Users\pspain\FaH\GPU1
Executable: C:\Users\pspain\FaH\GPU1\gpu1.exe
Arguments: -gpu 1 -verbosity 9

[16:06:29] - Ask before connecting: No
[16:06:29] - User name: SiriusB_[OcUK] (Team 10)
[16:06:29] - User ID: C34613D10C4C717
[16:06:29] - Machine ID: 11
[16:06:29]
[16:06:29] Loaded queue successfully.
[16:06:29]
[16:06:29] + Processing work unit
[16:06:29] Core required: FahCore_11.exe
[16:06:29] - Autosending finished units... [April 10 16:06:29 UTC]
[16:06:29] Core found.
[16:06:29] Trying to send all finished work units
[16:06:29] Working on queue slot 01 [April 10 16:06:29 UTC]
[16:06:29] + Working ...
[16:06:29] + No unsent completed units remaining.
[16:06:29] - Calling '.\FahCore_11.exe -dir work/ -suffix 01 -checkpoint 15 -ver
bose -lifeline 8564 -version 623'

[16:06:29] - Autosend completed
[16:06:29]
[16:06:29] *------------------------------*
[16:06:29] Folding@Home GPU Core
[16:06:29] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:06:29]
[16:06:29] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:06:29] Build host: amoeba
[16:06:29] Board Type: Nvidia
[16:06:29] Core      :
[16:06:29] Preparing to commence simulation
[16:06:29] - Looking at optimizations...
[16:06:29] - Files status OK
[16:06:29] - Expanded 64622 -> 341507 (decompressed 528.4 percent)
[16:06:29] Called DecompressByteArray: compressed_data_size=64622 data_size=3415
07, decompressed_data_size=341507 diff=0
[16:06:29] - Digital signature verified
[16:06:29]
[16:06:29] Project: 5786 (Run 6, Clone 0, Gen 39)
[16:06:29]
[16:06:29] Assembly optimizations on if available.
[16:06:29] Entering M.D.
[16:06:35] Tpr hash work/wudata_01.tpr:  2390296719 949216375 630254473 32763587
84 135935992
[16:06:35]
[16:06:36] Calling fah_main args: 14 usage=90
[16:06:36]
[16:06:36] Working on GRoups of Organic Molecules in ACtion for Science
[16:06:38] Client config found, loading data.
[16:06:38] Starting GUI Server
[16:06:38] mdrun_gpu returned
[16:06:38] NANs detected on GPU
[16:06:38]
[16:06:38] Folding@home Core Shutdown: UNSTABLE_MACHINE
[16:06:42] CoreStatus = 7A (122)
[16:06:42] Sending work to server
[16:06:42] Project: 5786 (Run 6, Clone 0, Gen 39)
[16:06:42] - Error: Could not get length of results file work/wuresults_01.dat
[16:06:42] - Error: Could not read unit 01 file. Removing from queue.
[16:06:42] Trying to send all finished work units
[16:06:42] + No unsent completed units remaining.
[16:06:42] - Preparing to get new work unit...
[16:06:42] + Attempting to get work packet
[16:06:42] - Will indicate memory of 4094 MB
[16:06:42] - Detect CPU. Vendor: GenuineIntel, Family: 6, Model: 15, Stepping: 1
1
[16:06:42] - Connecting to assignment server
[16:06:42] Connecting to http://assign-GPU.stanford.edu:8080/
[16:06:43] Posted data.
[16:06:43] Initial: 40AB; - Successful: assigned to (171.64.65.61).
[16:06:43] + News From Folding@Home: Welcome to Folding@Home
[16:06:43] Loaded queue successfully.
[16:06:43] Connecting to http://171.64.65.61:8080/
[16:06:44] Posted data.
[16:06:44] Initial: 0000; - Receiving payload (expected size: 74336)
[16:06:45] - Downloaded at ~72 kB/s
[16:06:45] - Averaged speed for that direction ~68 kB/s
[16:06:45] + Received work.
[16:06:45] Trying to send all finished work units
[16:06:45] + No unsent completed units remaining.
[16:06:45] + Closed connections
[16:06:50]
[16:06:50] + Processing work unit
[16:06:50] Core required: FahCore_11.exe
[16:06:50] Core found.
[16:06:50] Working on queue slot 02 [April 10 16:06:50 UTC]
[16:06:50] + Working ...
[16:06:50] - Calling '.\FahCore_11.exe -dir work/ -suffix 02 -checkpoint 15 -ver
bose -lifeline 8564 -version 623'

[16:06:52]
[16:06:52] *------------------------------*
[16:06:52] Folding@Home GPU Core
[16:06:52] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[16:06:52]
[16:06:52] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14
.00.50727.762 for 80x86
[16:06:52] Build host: amoeba
[16:06:52] Board Type: Nvidia
[16:06:52] Core      :
[16:06:52] Preparing to commence simulation
[16:06:52] - Looking at optimizations...
[16:06:52] DeleteFrameFiles: successfully deleted file=work/wudata_02.ckp
[16:06:52] - Created dyn
[16:06:52] - Files status OK
[16:06:52] - Expanded 73824 -> 383588 (decompressed 519.5 percent)
[16:06:52] Called DecompressByteArray: compressed_data_size=73824 data_size=3835
88, decompressed_data_size=383588 diff=0
[16:06:52] - Digital signature verified
[16:06:52]
[16:06:52] Project: 6601 (Run 8, Clone 838, Gen 6)
[16:06:52]
[16:06:52] Assembly optimizations on if available.
[16:06:52] Entering M.D.
[16:06:58] Tpr hash work/wudata_02.tpr:  1573423998 3212315442 3651801022 318698
4192 669694147
[16:06:58]
[16:06:58] Calling fah_main args: 14 usage=90
[16:06:59]
[16:06:59] Working on Protein
[16:07:01] Client config found, loading data.
[16:07:01] Starting GUI Server
[16:07:58] Completed 1%
[16:08:54] Completed 2%
[16:09:52] Completed 3%
[16:10:50] Completed 4%
[16:11:48] Completed 5%
[16:12:45] Completed 6%
[16:13:43] Completed 7%
[16:14:40] Completed 8%
[16:15:38] Completed 9%
[16:16:35] Completed 10%
[16:17:33] Completed 11%

Both the clients were restarted several hours after they stopped due to EUEs so zero possibility they were already hot. Also, you can see from the first log my GPU can crunch 5785 just fine. Are 5784 and 5786 highly optimised and 5785 isn't?

Folding Forum

Projects 5786 and 5784 - NANs Detected

Projects 5786 and 5784 - NANs Detected

Re: Project 5786 Run 10, Clone 54, Gen 28 - NANs detected on GPU

Re: Project 5786 Run 10, Clone 54, Gen 28 - NANs detected on GPU

Re: Project 5786 Run 10, Clone 54, Gen 28 - NANs detected on GPU

Projects 5786 and 5784 - NANs Detected

Re: Projects 5786 and 5784 - NANs Detected

Re: Projects 5786 and 5784 - NANs Detected

Re: Projects 5786 and 5784 - NANs Detected