Crunching on GPU.

[adrianxw]

I used to have the program running at "Medium" where it used the CPU and GPU. I stopped GPU work when I started noticing arrays of spots appearing on the display, I guess some stuck bits in the video memory. I set the priority down to "Light" to suspend GPU working. The graphics card has been swapped and i set the control back to "Medium" but it has not started GPU computation again. What do I need to do to get it going again?

[JimboPalmer]

Do you know what card you used to have and what card you have now?

What OS are you using?

I always suggest the latest driver from the vendor (AMD or Nvidia)

[adrianxw]

The card is identical to the one I removed, ASUS HD7850. The other, almost identical, machine is shut down awaiting a new PSU. Both systems are Windows 8.1 x64 for the next few months.

[JimboPalmer]

https://www.amd.com/en/support/graphics ... on-hd-7850

There has not been a new driver since 2017, it is an old architecture, version 1.0 of GCN, introduced in 2012.

https://www.techpowerup.com/gpu-specs/r ... 7850.c1055

[Joe_H]

The spots on the original card may indicate localized overheating of components such as video RAM on the card. Possibly fixable by separating the cooling assembly and replacing the thermal transfer pads and then reassembling.

However depending on the timing, your card not getting work is probably due to reclassification of AMD GPUs done this past week. That changed the groups various cards were in, assignment settings on the servers for various GPU projects still are in the process of being updated to reflect the changes.

[adrianxw]

The pattern of spots on the failing card is not random. It presents as arrays of a four spot basically rhomboidal in form over about a third of the screen, which third is variable, and can be moved by, for example, picking up an object and moving it, a small window for example. The spots are not always the same colour, ie. are not necessarily dark, or light or monochrome. It may resolve once the modifications to the categories is complete then?

[adrianxw]

No, I didn't think so either.

[toTOW]

If you are still using v7.5.1 client, it might be wise to update it ...

[adrianxw]

I don't know which version it is using, I expect it will tell me if I look in the right place...

[Joe_H]

The version info for the client is posted in the first part of the log file, and also displayed in the System Info tab of FAHControl. For example, here is the beginning section from the log on my Mac system:

Code: Select all

*********************** Log Started 2022-10-26T15:29:38Z ***********************
15:29:38:******************************* libFAH ********************************
15:29:38:       Date: Oct 20 2020
15:29:38:       Time: 13:36:42
15:29:38:   Revision: 5ca109d295a6245e2a2f590b3d0085ad5e567aeb
15:29:38:     Branch: master
15:29:38:   Compiler: GNU 4.2.1 Compatible Apple LLVM 11.0.0 (clang-1100.0.33.8)
15:29:38:    Options: -std=c++11 -fsigned-char -O3 -funroll-loops
15:29:38:             -mmacosx-version-min=10.7 -Wno-unused-local-typedefs -stdlib=libc++
15:29:38:   Platform: darwin 19.2.0
15:29:38:       Bits: 64
15:29:38:       Mode: Release
15:29:38:****************************** FAHClient ******************************
15:29:38:    Version: 7.6.21
15:29:38:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
15:29:38:  Copyright: 2020 foldingathome.org
15:29:38:   Homepage: https://foldingathome.org/
15:29:38:       Date: Oct 20 2020
15:29:38:       Time: 13:39:56
15:29:38:   Revision: 6efbf0e138e22d3963e6a291f78dcb9c6422a278
15:29:38:     Branch: master
15:29:38:   Compiler: GNU 4.2.1 Compatible Apple LLVM 11.0.0 (clang-1100.0.33.8)
15:29:38:    Options: -std=c++11 -fsigned-char -O3 -funroll-loops
15:29:38:             -mmacosx-version-min=10.7 -Wno-unused-local-typedefs -stdlib=libc++
15:29:38:   Platform: darwin 19.2.0
15:29:38:       Bits: 64
15:29:38:       Mode: Release
15:29:38:     Config: /Library/Application Support/FAHClient/config.xml
15:29:38:******************************** CBang ********************************
15:29:38:       Date: Oct 20 2020
15:29:38:       Time: 11:39:07
15:29:38:   Revision: 7e4ce85225d7eaeb775e87c31740181ca603de60
15:29:38:     Branch: master
15:29:38:   Compiler: GNU 4.2.1 Compatible Apple LLVM 11.0.0 (clang-1100.0.33.8)
15:29:38:    Options: -std=c++11 -fsigned-char -O3 -funroll-loops
15:29:38:             -mmacosx-version-min=10.7 -Wno-unused-local-typedefs -stdlib=libc++
15:29:38:             -fPIC
15:29:38:   Platform: darwin 19.2.0
15:29:38:       Bits: 64
15:29:38:       Mode: Release
15:29:38:******************************* System ********************************
15:29:38:        CPU: Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz
15:29:38:     CPU ID: GenuineIntel Family 6 Model 58 Stepping 9
15:29:38:       CPUs: 4
15:29:38:     Memory: 16.00GiB
15:29:38:Free Memory: 11.66GiB
15:29:38:    Threads: POSIX_THREADS
15:29:38: OS Version: 10.14
15:29:38:Has Battery: true
15:29:38: On Battery: false
15:29:38: UTC Offset: -4
15:29:38:        PID: 7581
15:29:38:        CWD: /Library/Application Support/FAHClient
15:29:38:         OS: Darwin 18.7.0 x86_64
15:29:38:    OS Arch: AMD64
15:29:38:       GPUs: 0
15:29:38:       CUDA: Not detected: Failed to open dynamic library 'libcuda.dylib':
15:29:38:             dlopen(libcuda.dylib, 1): image not found
15:29:38:     OpenCL: Not detected: Failed to open dynamic library 'libOpenCL.dylib':
15:29:38:             dlopen(libOpenCL.dylib, 1): image not found
15:29:38:***********************************************************************
15:29:38:<config>

[JimboPalmer]

In case you wish to post a log file, to get better help, this link has hints:

viewtopic.php?t=26036

I recommend at least the first 100 lines of the log, that tells how it is configured.

[adrianxw]

I believe it says 7.6.20 there is a long list of versions showing changes. That is the top of the list.