I have installed Folding@Home with FAHClient, FAHControl and FAHViewer. The folding works with CPU, but I am having trouble folding with GPU. The installed nvidia driver is 361.48 from nvidia opensuse leap repository.
Someone on Folding@Home forum mentioned that I perhaps was missing OpenCL, but I have installed nvidia-computeG04 which contains OpenCL. I could install libOpenCL, but that conflicts with nvidia-computeG04.
File /usr/lib64/libOpenCL.so.1.0.0
from install of
libOpenCL1-2.2.7-1.1.x86_64 (openSUSE-leap/42.1-Oss)
conflicts with file from package
nvidia-computeG04-361.42-21.1.x86_64 (@System)
I tried to install libOpenCL in spite of the conflict and still folding would not work using GPU.
Has anyone gotten folding to work using GPU on OpenSUSE?
https://foldingforum.org/viewtopic.php?f=61&t=28751&p=285006#p285006
The Log from FAHControl
13:03:12:WU00:FS00:Assigned to work server 140.163.4.235
13:03:12:WU00:FS00:Requesting new work unit for slot 00: READY gpu:0:GK106 [GeForce GTX 650 Ti] from 140.163.4.235
13:03:12:WU00:FS00:Connecting to 140.163.4.235:8080
13:03:12:WU00:FS00:Downloading 3.78MiB
13:03:16:WU00:FS00:Download complete
13:03:16:WU00:FS00:Received Unit: id:00 state:DOWNLOAD error:NO_ERROR project:10485 run:0 clone:53 gen:264 core:0x18 unit:0x0000014f538b3dbb54aeb39d80684dce
13:03:16:WU00:FS00:Starting
13:03:16:WU00:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/web.stanford.edu/~pande/Linux/AMD64/NVIDIA/Fermi/Core_18.fah/FahCore_18 -dir 00 -suffix 01 -version 704 -lifeline 6462 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
13:03:16:WU00:FS00:Started FahCore on PID 6510
13:03:16:WU00:FS00:Core PID:6514
13:03:16:WU00:FS00:FahCore 0x18 started
13:03:16:WU00:FS00:0x18:*********************** Log Started 2016-04-19T13:03:16Z ***********************
13:03:16:WU00:FS00:0x18:Project: 10485 (Run 0, Clone 53, Gen 264)
13:03:16:WU00:FS00:0x18:Unit: 0x0000014f538b3dbb54aeb39d80684dce
13:03:16:WU00:FS00:0x18:CPU: 0x00000000000000000000000000000000
13:03:16:WU00:FS00:0x18:Machine: 0
13:03:16:WU00:FS00:0x18:Reading tar file state.xml
13:03:16:WU00:FS00:0x18:Reading tar file system.xml
13:03:17:WU00:FS00:0x18:Reading tar file integrator.xml
13:03:17:WU00:FS00:0x18:Reading tar file core.xml
13:03:17:WU00:FS00:0x18:Digital signatures verified
13:03:17:WU00:FS00:0x18:Folding@home GPU core18
13:03:17:WU00:FS00:0x18:Version 0.0.4
13:03:17:WU00:FS00:0x18:ERROR:exception: Bad platformId size.
13:03:17:WU00:FS00:0x18:Saving result file logfile_01.txt
13:03:17:WU00:FS00:0x18:Saving result file log.txt
13:03:17:WU00:FS00:0x18:Folding@home Core Shutdown: BAD_WORK_UNIT
13:03:17:WARNING:WU00:FS00:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
13:03:17:WU00:FS00:Sending unit results: id:00 state:SEND error:FAULTY project:10485 run:0 clone:53 gen:264 core:0x18 unit:0x0000014f538b3dbb54aeb39d80684dce
13:03:17:WU00:FS00:Uploading 1.92KiB to 140.163.4.235
13:03:17:WU00:FS00:Connecting to 140.163.4.235:8080
13:03:17:WU01:FS00:Connecting to 171.67.108.45:80
13:03:18:WU00:FS00:Upload complete
13:03:18:WU00:FS00:Server responded WORK_ACK (400)
13:03:18:WU00:FS00:Cleaning up
I believe that openCL requires the NVIDIA cuda driver in addition to the screen diver
The package nvidia-computeG04 gives both Cuda and OpenCL.
Information for package nvidia-computeG04:
------------------------------------------
Repository: Nvidia
Name: nvidia-computeG04
Version: 361.42-21.1
Arch: x86_64
Vendor: obs://build.suse.de/home:sndirsch:drivers
Installed: Yes
Status: up-to-date
Installed Size: 51.3 MiB
Summary: NVIDIA driver for computing with GPGPU
Description:
NVIDIA driver for computing with GPGPUs using CUDA or OpenCL
Hi
I run seti@home on occasions, site is down for maintenance so can’t login… anyway with just the 340.96 driver and an 8800GT Card I see using the BOINC client;
....
19-Apr-2016 14:25:19 ---] CUDA: NVIDIA GPU 0: GeForce 8800 GT (driver version unknown, CUDA version 6.5, compute capability 1.1, 512MB, 200MB available, 504 GFLOPS peak)
19-Apr-2016 14:25:19 ---] OpenCL: NVIDIA GPU 0: GeForce 8800 GT (driver version 340.96, device version OpenCL 1.0 CUDA, 512MB, 200MB available, 504 GFLOPS peak)
19-Apr-2016 14:25:19 ---] Creating new client state file
19-Apr-2016 14:25:19 ---] Host name: big-bird.homelinux.org
19-Apr-2016 14:25:19 ---] Processor: 4 AuthenticAMD AMD Athlon(tm) II X4 635 Processor [Family 16 Model 5 Stepping 3]
....
Sure you need the compute one…? I only use the run file to install the driver (I find it way easier…)
I can’t remove nvidia-computeG04 it would then also remove the x11-video-nvidia package.
The following 2 packages are going to be REMOVED:
nvidia-computeG04 x11-video-nvidiaG04
You need all the GO4 packages
I have installed all the Nvidia G04 packages
nvidia-computeG04
nvidia-gfxG04-kmp-default
nvidia-glG04
x11-video-nvidiaG04
And you can’t remove one only since they are dependent
Still think you may need cuda. But the fold docs should tell you for sure what you need
I have CUDA. The package nvidia-computeG04 contains both OpenCL and CUDA.
That may not be sufficient for ALL programs
There is no other package that provides CUDA. And the libOpenCL conflicts with nvidia-computeG04.
Some explanation I found on Folding@Home forum from someone who had the same problem on OpenSUSE 13.1
The user fahclient who runs the service FAHClient and application FAHControl is not in the group video and thus does not get access to the GPU.
It sounded like a plausible explanation to the problem considering I have the correct Nvidia driver installed along with libOpenCL.
I have now confirmed that the user fahclient running FAHClient does not have access to the GPU. I was able to run the client with my own user and it could use the GPU.
I added the user fahclient to group video in the file /etc/group and thought that was enough, but it did not take affect, even if “id fahclient” did show it belonged to this group.
Did you reboot re re log in after change in groups??
And no cude is not in the repos need to get from NVIDA
Did not reboot after changing /etc/group. Is there a way I don’t have to reboot. When I changed my own user and added the group video all I needed was to log out and in. The user fahclient was not logged in anywhere.
I have installed all the packages from Nvidia OpenSUSE repository, including nvidia-computeG04 which contains CUDA and OpenCL.
I remember I had to add my user to the group video to solve a problem I had, I think it was with video hardware acceleration.
Wonder why there is no documentation about this on https://en.opensuse.org/SDB:NVIDIA.
Was my manually edit of /etc/group the equivalent of usermod --append -G video yourusername?
You’d have to ask NVIDIA why no docs. It is their repo.
Re logging will work
NVIDIA explains things about as well as MS
This has to do with OpenSUSE internals and nothing to do with Nvidia. That a user need to belong to group video.
It is also OpenSUSE documentation page, not written by Nvidia.
If re logging will work, then it must be something else that my own user has access to that this other service user does not. What does a user need to access the graphic cards GPU?
I just restarted my system after adding the user fahclient to the group video. It did not work, so there must to something else preventing this user to access the GPU. Anyone have any idea why my own user does have access?
/etc/passwd
djviking:x:1000:100:Viking:/home/djviking:/bin/bash
fahclient:x:478:100:Folding@home Client:/var/lib/fahclient:/sbin/nologin
/etc/group
video:x:33:djviking,fahclient
users:x:100:djviking
I just also tried to add fahclient to users
sudo usermod --append -G users fahclient
Then restarted the service FAHClient, but still no access to the GPU.
Some new information has come to light.
**mintaka:/home/djviking #** su - fahclient
fahclient@mintaka:~> glxinfo2
Error: unable to open display
Could this be a reason why fahclient cannot access the GPU?
fahclient@mintaka:~> /sbin/lspci -vnn | grep VGA -A 12
01:00.0 **VGA** compatible controller [0300]: NVIDIA Corporation GK106 [GeForce GTX 650 Ti] [10de:11c6] (rev a1) (prog-if 00 **VGA** controller])
Subsystem: ASUSTeK Computer Inc. Device [1043:842b]
Flags: bus master, fast devsel, latency 0, IRQ 38
Memory at fd000000 (32-bit, non-prefetchable) [size=16]
Memory at f0000000 (64-bit, prefetchable) [size=128]
Memory at fa000000 (64-bit, prefetchable) [size=32]
I/O ports at 9c00 [size=128]
[virtual] Expansion ROM at fe500000 [disabled] [size=512]
Capabilities: <access denied>
Kernel driver in use: nvidia
Kernel modules: nouveau, nvidia
01:00.1 Audio device [0403]: NVIDIA Corporation GK106 HDMI Audio Controller [10de:0e0b] (rev a1)
However I also get “Capabilities: <access denied>” on my own user[/size][/size][/size][/size][/size]