GTX 780 Ti - Bad PlatformId Size SOLVED: RDP problem

[CritiCal]

Hello,

Yesterday, my new GTX 780 Ti arrived. I set it up to fold and got the expected ~200k PPD for a few hours, and went to bed expecting everything to be working. When I got up, however, it seemed as though the 780 had done no work at all. Looking at the log files, it seems to have completed a single unit, before it started to fail with ERROR:exception: Bad platformId size. I would upload the entire logs, but since I reinstalled the client they have been deleted, and all the work units were the same anyway. Here is one example:

Code: Select all

07:45:27:WU00:FS00:Connecting to 171.64.65.69:8080
07:45:27:WU01:FS00:Connecting to assign-GPU.stanford.edu:80
07:45:28:WU01:FS00:News: Welcome to Folding@Home
07:45:28:WU01:FS00:Assigned to work server 171.64.65.69
07:45:28:WU01:FS00:Requesting new work unit for slot 00: READY gpu:0:GK110 [GeForce GTX 780 Ti] from 171.64.65.69
07:45:28:WU01:FS00:Connecting to 171.64.65.69:8080
07:45:28:WU01:FS00:Downloading 4.17MiB
07:45:33:WU01:FS00:Download complete
07:45:33:WU01:FS00:Received Unit: id:01 state:DOWNLOAD error:NO_ERROR project:8900 run:76 clone:6 gen:72 core:0x17 unit:0x0000006e028c126651a6377714c6e7ba
07:45:33:WU01:FS00:Starting
07:45:33:WU01:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/admin/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17.exe -dir 01 -suffix 01 -version 703 -lifeline 2840 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
07:45:33:WU01:FS00:Started FahCore on PID 6088
07:45:34:WU01:FS00:Core PID:6052
07:45:34:WU01:FS00:FahCore 0x17 started
07:45:34:WU01:FS00:0x17:*********************** Log Started 2014-01-29T07:45:34Z ***********************
07:45:34:WU01:FS00:0x17:Project: 8900 (Run 76, Clone 6, Gen 72)
07:45:34:WU01:FS00:0x17:Unit: 0x0000006e028c126651a6377714c6e7ba
07:45:34:WU01:FS00:0x17:CPU: 0x00000000000000000000000000000000
07:45:34:WU01:FS00:0x17:Machine: 0
07:45:34:WU01:FS00:0x17:Reading tar file state.xml
07:45:35:WU01:FS00:0x17:Reading tar file system.xml
07:45:35:WU01:FS00:0x17:Reading tar file integrator.xml
07:45:35:WU01:FS00:0x17:Reading tar file core.xml
07:45:35:WU01:FS00:0x17:Digital signatures verified
07:45:35:WU01:FS00:0x17:Folding@home GPU core17
07:45:35:WU01:FS00:0x17:Version 0.0.52
07:45:35:WU01:FS00:0x17:ERROR:exception: Bad platformId size.
07:45:35:WU01:FS00:0x17:Saving result file logfile_01.txt
07:45:35:WU01:FS00:0x17:Saving result file log.txt
07:45:35:WU01:FS00:0x17:Folding@home Core Shutdown: BAD_WORK_UNIT
07:45:35:WARNING:WU01:FS00:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
07:45:35:WU01:FS00:Sending unit results: id:01 state:SEND error:FAULTY project:8900 run:76 clone:6 gen:72 core:0x17 unit:0x0000006e028c126651a6377714c6e7ba
07:45:35:WU01:FS00:Uploading 1.84KiB to 171.64.65.69
07:45:35:WU01:FS00:Connecting to 171.64.65.69:8080
07:45:36:WU01:FS00:Upload complete
07:45:36:WU01:FS00:Server responded WORK_ACK (400)
07:45:36:WU01:FS00:Cleaning up
07:45:36:WU02:FS00:Connecting to assign-GPU.stanford.edu:80
07:45:36:WU02:FS00:News: Welcome to Folding@Home
07:45:36:WU02:FS00:Assigned to work server 171.64.65.69
07:45:36:WU02:FS00:Requesting new work unit for slot 00: READY gpu:0:GK110 [GeForce GTX 780 Ti] from 171.64.65.69
07:45:36:WU02:FS00:Connecting to 171.64.65.69:8080
07:45:37:WU02:FS00:Downloading 4.18MiB
07:45:43:WU02:FS00:Download 94.27%
07:45:43:WU02:FS00:Download complete
07:45:43:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:NO_ERROR project:8900 run:762 clone:2 gen:57 core:0x17 unit:0x00000060028c126651a6d029b9a05ff2
07:45:43:WU02:FS00:Starting
07:45:43:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/admin/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17.exe -dir 02 -suffix 01 -version 703 -lifeline 2840 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
07:45:43:WU02:FS00:Started FahCore on PID 4448
07:45:43:WU02:FS00:Core PID:5216
07:45:43:WU02:FS00:FahCore 0x17 started
07:45:43:WU02:FS00:0x17:*********************** Log Started 2014-01-29T07:45:43Z ***********************
07:45:43:WU02:FS00:0x17:Project: 8900 (Run 762, Clone 2, Gen 57)
07:45:43:WU02:FS00:0x17:Unit: 0x00000060028c126651a6d029b9a05ff2
07:45:43:WU02:FS00:0x17:CPU: 0x00000000000000000000000000000000
07:45:43:WU02:FS00:0x17:Machine: 0
07:45:43:WU02:FS00:0x17:Reading tar file state.xml
07:45:44:WU02:FS00:0x17:Reading tar file system.xml
07:45:44:WU02:FS00:0x17:Reading tar file integrator.xml
07:45:44:WU02:FS00:0x17:Reading tar file core.xml
07:45:44:WU02:FS00:0x17:Digital signatures verified
07:45:44:WU02:FS00:0x17:Folding@home GPU core17
07:45:45:WU02:FS00:0x17:Version 0.0.52
07:45:45:WU02:FS00:0x17:ERROR:exception: Bad platformId size.
07:45:45:WU02:FS00:0x17:Saving result file logfile_01.txt
07:45:45:WU02:FS00:0x17:Saving result file log.txt
07:45:45:WU02:FS00:0x17:Folding@home Core Shutdown: BAD_WORK_UNIT
07:45:45:WARNING:WU02:FS00:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
07:45:45:WU02:FS00:Sending unit results: id:02 state:SEND error:FAULTY project:8900 run:762 clone:2 gen:57 core:0x17 unit:0x00000060028c126651a6d029b9a05ff2
07:45:45:WU02:FS00:Uploading 1.83KiB to 171.64.65.69
07:45:45:WU02:FS00:Connecting to 171.64.65.69:8080
07:45:46:WU02:FS00:Upload complete
07:45:46:WU02:FS00:Server responded WORK_ACK (400)
07:45:46:WU02:FS00:Cleaning up
07:45:48:WARNING:WU00:FS00:WorkServer connection failed on port 8080 trying 80
07:45:48:WU00:FS00:Connecting to 171.64.65.69:80
07:45:48:WU00:FS00:Upload 100.00%
07:45:48:WU00:FS00:Upload complete
07:45:48:WU00:FS00:Server responded WORK_ACK (400)
07:45:48:WU00:FS00:Cleaning up

All work units seem to get this exact problem. I would have thought it was a faulty card, but the catch is that the exact same was happening to my old GTX 580 which I just removed from my system. I only recently got back into folding, and had never before used the v7 client, so I thought maybe the GTX 5xx series didn't work with v7 or something like that, and since the 780 was already on its way I didn't bother to fix it (thinking the new GPU would fix the problem). I did try to clock the card down from its stock 1020/1085 MHz core to 915/980 (the lowest possible with OC GURU, it's a card of the type gv-n78toc-3gdrev_10 from Gigabyte), but it did not help at all. While it was folding, GPU-z never reported a TDP of more than 24%, so the card hardly seemed to be under load at all (which I thought was strange). The GTX 660 Ti I have in another computer seems to be doing OK, at least it's not getting this exact error. The computer is running Windows 8 with NVIDIA Driver 332.21. More specs coming soon.

EDIT: It appears that since Remote Desktop doesn't use the actual GPU of the computer, Folding@Home can't use it either if started over RDP. Therefore, the solution was simply to use an alternative to RDP, and the graphics card is now folding at 192k PPD. Sorry to bother you with this stupid problem.

[EXT64]

Thanks for posting back so quickly with the solution - I'm sure it will help others who run into the same problem

[7im]

Remote Desktop is known to cause problems with GPU folding for that exact reason. Thanks for reporting the error information. Hadn't seen that in win 8 yet.

[PantherX]

Welcome to the F@H Forum CritiCal,

Please note that your GTX 660 Ti and GTX 580 will preform best for folding when using WHQL Driver 327.23 version. The reason is that the latest WHQL Drivers from Nvidia will perform poorly on non- GK110 GPUs. You can read about it here (viewtopic.php?f=80&t=25530) and here (viewtopic.php?f=80&t=25282).

[n_w95482]

EDIT: It appears that since Remote Desktop doesn't use the actual GPU of the computer, Folding@Home can't use it either if started over RDP. Therefore, the solution was simply to use an alternative to RDP, and the graphics card is now folding at 192k PPD. Sorry to bother you with this stupid problem.

Yep, I've encountered the same problem as well. I've noticed that if I RDP in while it's in the middle of a WU, it'll sometimes finish it successfully. In every case that it does, it bombs out when starting the new one. Sometimes it dies within a few minutes of me connecting via RDP.

I've since switched to using TeamViewer to remotely administer my GPU folding machines, zero issues with it. VNC and Splashtop work as well, among others.

[codysluder]

Yes, RDT is known to disable whatever FAH needs from the GPU and several alternatives work fine, including VNC and TeamViewer. (Thanks, Micro$oft ) I find that remote monitoring in the Advanced Control program works great for what I want to do but the alternative connections do let you do other things like manage non-F@H files.