AMD/ATI HD5870 gpu not being used

[ader]

I'm using an AMD/ATI HD5870 gpu, but the FAH software does not appear to use it, the GPU load stays < 1%:


Is there a way to make FAH actually use this GPU? around 90% would be nice.

Thanks,
ader

Log:

Code: Select all

*********************** Log Started 2017-06-28T22:39:53Z ***********************
22:39:53:************************* Folding@home Client *************************
22:39:53:      Website: http://folding.stanford.edu/
22:39:53:    Copyright: (c) 2009-2014 Stanford University
22:39:53:       Author: Joseph Coffland <joseph@cauldrondevelopment.com>
22:39:53:         Args: --open-web-control
22:39:53:       Config: C:/Users/ader/AppData/Roaming/FAHClient/config.xml
22:39:53:******************************** Build ********************************
22:39:53:      Version: 7.4.4
22:39:53:         Date: Mar 4 2014
22:39:53:         Time: 20:26:54
22:39:53:      SVN Rev: 4130
22:39:53:       Branch: fah/trunk/client
22:39:53:     Compiler: Intel(R) C++ MSVC 1500 mode 1200
22:39:53:      Options: /TP /nologo /EHa /Qdiag-disable:4297,4103,1786,279 /Ox -arch:SSE
22:39:53:               /QaxSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 /Qopenmp /Qrestrict /MT /Qmkl
22:39:53:     Platform: win32 XP
22:39:53:         Bits: 32
22:39:53:         Mode: Release
22:39:53:******************************* System ********************************
22:39:53:          CPU: Intel(R) Core(TM) i7 CPU X 980 @ 3.33GHz
22:39:53:       CPU ID: GenuineIntel Family 6 Model 44 Stepping 2
22:39:53:         CPUs: 12
22:39:53:       Memory: 16.00GiB
22:39:53:  Free Memory: 11.57GiB
22:39:53:      Threads: WINDOWS_THREADS
22:39:53:   OS Version: 6.1
22:39:53:  Has Battery: false
22:39:53:   On Battery: false
22:39:53:   UTC Offset: -7
22:39:53:          PID: 8948
22:39:53:          CWD: C:/Users/ader/AppData/Roaming/FAHClient
22:39:53:           OS: Windows 7 Professional
22:39:53:      OS Arch: AMD64
22:39:53:         GPUs: 1
22:39:53:        GPU 0: ATI:4 Cypress [Radeon HD 5800/6800]
22:39:53:         CUDA: Not detected
22:39:53:Win32 Service: false
22:39:53:***********************************************************************
22:39:53:<config>
22:39:53:  <!-- Network -->
22:39:53:  <proxy v=':8080'/>
22:39:53:
22:39:53:  <!-- Slot Control -->
22:39:53:  <power v='full'/>
22:39:53:
22:39:53:  <!-- User Information -->
22:39:53:  <passkey v='********************************'/>
22:39:53:  <team v='224497'/>
22:39:53:  <user v='ader'/>
22:39:53:
22:39:53:  <!-- Folding Slots -->
22:39:53:  <slot id='0' type='CPU'>
22:39:53:    <cpus v='10'/>
22:39:53:    <paused v='true'/>
22:39:53:  </slot>
22:39:53:  <slot id='1' type='GPU'>
22:39:53:    <paused v='true'/>
22:39:53:  </slot>
22:39:53:</config>
22:39:53:Trying to access database...
22:39:53:Successfully acquired database lock
22:39:53:Enabled folding slot 00: PAUSED cpu:10 (by user)
22:39:53:Enabled folding slot 01: PAUSED gpu:0:Cypress [Radeon HD 5800/6800] (by user)
22:39:57:19:127.0.0.1:New Web connection
22:42:10:Removing old file 'configs/config-20170624-032506.xml'
22:42:10:Saving configuration to config.xml
22:42:10:<config>
22:42:10:  <!-- Folding Core -->
22:42:10:  <checkpoint v='10'/>
22:42:10:
22:42:10:  <!-- Network -->
22:42:10:  <proxy v=':8080'/>
22:42:10:
22:42:10:  <!-- Slot Control -->
22:42:10:  <power v='full'/>
22:42:10:
22:42:10:  <!-- User Information -->
22:42:10:  <passkey v='********************************'/>
22:42:10:  <team v='224497'/>
22:42:10:  <user v='ader'/>
22:42:10:
22:42:10:  <!-- Folding Slots -->
22:42:10:  <slot id='0' type='CPU'>
22:42:10:    <cpus v='10'/>
22:42:10:    <paused v='true'/>
22:42:10:  </slot>
22:42:10:  <slot id='1' type='GPU'>
22:42:10:    <paused v='true'/>
22:42:10:  </slot>
22:42:10:</config>
22:42:16:FS00:Unpaused
22:42:16:FS01:Unpaused
22:42:16:WU00:FS00:Starting
22:42:16:WU00:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/ader/AppData/Roaming/FAHClient/cores/fahwebx.stanford.edu/cores/Win32/AMD64/Core_a7.fah/FahCore_a7.exe -dir 00 -suffix 01 -version 704 -lifeline 8948 -checkpoint 10 -np 10
22:42:16:WU00:FS00:Started FahCore on PID 10028
22:42:16:WU00:FS00:Core PID:8256
22:42:16:WU00:FS00:FahCore 0xa7 started
22:42:17:WU01:FS01:Starting
22:42:17:WU01:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/ader/AppData/Roaming/FAHClient/cores/fahwebx.stanford.edu/cores/Win32/AMD64/ATI/R600/Core_21.fah/FahCore_21.exe -dir 01 -suffix 01 -version 704 -lifeline 8948 -checkpoint 10 -gpu 0 -gpu-vendor ati
22:42:17:WU01:FS01:Started FahCore on PID 15744
22:42:17:WU01:FS01:Core PID:14088
22:42:17:WU01:FS01:FahCore 0x21 started
22:42:17:WU00:FS00:0xa7:*********************** Log Started 2017-06-28T22:42:16Z ***********************
22:42:17:WU00:FS00:0xa7:************************** Gromacs Folding@home Core ***************************
22:42:17:WU00:FS00:0xa7:       Type: 0xa7
22:42:17:WU00:FS00:0xa7:       Core: Gromacs
22:42:17:WU00:FS00:0xa7:    Website: http://folding.stanford.edu/
22:42:17:WU00:FS00:0xa7:  Copyright: (c) 2009-2016 Stanford University
22:42:17:WU00:FS00:0xa7:     Author: Joseph Coffland <joseph@cauldrondevelopment.com>
22:42:17:WU00:FS00:0xa7:       Args: -dir 00 -suffix 01 -version 704 -lifeline 10028 -checkpoint 10 -np
22:42:17:WU00:FS00:0xa7:             10
22:42:17:WU00:FS00:0xa7:     Config: <none>
22:42:17:WU00:FS00:0xa7:************************************ Build *************************************
22:42:17:WU00:FS00:0xa7:    Version: 0.0.11
22:42:17:WU00:FS00:0xa7:       Date: Sep 21 2016
22:42:17:WU00:FS00:0xa7:       Time: 01:53:37
22:42:17:WU00:FS00:0xa7: Repository: Git
22:42:17:WU00:FS00:0xa7:   Revision: 957bd90e68d95ddcf1594dc15ff6c64cc4555146
22:42:17:WU00:FS00:0xa7:     Branch: master
22:42:17:WU00:FS00:0xa7:   Compiler: GNU 4.2.1 Compatible Clang 3.9.0 (trunk 274080)
22:42:17:WU00:FS00:0xa7:    Options: -std=gnu++98 -O3 -funroll-loops -ffast-math -mfpmath=sse
22:42:17:WU00:FS00:0xa7:             -fno-unsafe-math-optimizations -msse2 -I/mingw64/include
22:42:17:WU00:FS00:0xa7:             -Wno-inconsistent-dllimport -Wno-parentheses-equality
22:42:17:WU00:FS00:0xa7:             -Wno-deprecated-register -Wno-unused-local-typedef
22:42:17:WU00:FS00:0xa7:   Platform: linux2 4.6.0-1-amd64
22:42:17:WU00:FS00:0xa7:       Bits: 64
22:42:17:WU00:FS00:0xa7:       Mode: Release
22:42:17:WU00:FS00:0xa7:       SIMD: sse2
22:42:17:WU00:FS00:0xa7:************************************ System ************************************
22:42:17:WU00:FS00:0xa7:        CPU: Intel(R) Core(TM) i7 CPU X 980 @ 3.33GHz
22:42:17:WU00:FS00:0xa7:     CPU ID: GenuineIntel Family 6 Model 44 Stepping 2
22:42:17:WU00:FS00:0xa7:       CPUs: 12
22:42:17:WU00:FS00:0xa7:     Memory: 16.00GiB
22:42:17:WU00:FS00:0xa7:Free Memory: 11.53GiB
22:42:17:WU00:FS00:0xa7:    Threads: WINDOWS_THREADS
22:42:17:WU00:FS00:0xa7: OS Version: 6.1
22:42:17:WU00:FS00:0xa7:Has Battery: false
22:42:17:WU00:FS00:0xa7: On Battery: false
22:42:17:WU00:FS00:0xa7: UTC Offset: -7
22:42:17:WU00:FS00:0xa7:        PID: 8256
22:42:17:WU00:FS00:0xa7:        CWD: C:\Users\ader\AppData\Roaming\FAHClient\work
22:42:17:WU00:FS00:0xa7:         OS: Windows 7 Professional Service Pack 1
22:42:17:WU00:FS00:0xa7:    OS Arch: AMD64
22:42:17:WU00:FS00:0xa7:********************************************************************************
22:42:17:WU00:FS00:0xa7:Project: 13804 (Run 0, Clone 4123, Gen 34)
22:42:17:WU00:FS00:0xa7:Unit: 0x0000002980fccb04591a0338e125323b
22:42:17:WU00:FS00:0xa7:Digital signatures verified
22:42:17:WU00:FS00:0xa7:Calling: mdrun -s frame34.tpr -o frame34.trr -x frame34.xtc -cpi state.cpt -cpt 10 -nt 10
22:42:17:WU00:FS00:0xa7:Steps: first=8500000 total=250000
22:42:17:WU01:FS01:0x21:*********************** Log Started 2017-06-28T22:42:17Z ***********************
22:42:17:WU01:FS01:0x21:Project: 10496 (Run 88, Clone 60, Gen 23)
22:42:17:WU01:FS01:0x21:Unit: 0x000000218ca304f558896e3d47fb5684
22:42:17:WU01:FS01:0x21:CPU: 0x00000000000000000000000000000000
22:42:17:WU01:FS01:0x21:Machine: 1
22:42:17:WU01:FS01:0x21:Digital signatures verified
22:42:17:WU01:FS01:0x21:Folding@home GPU Core21 Folding@home Core
22:42:17:WU01:FS01:0x21:Version 0.0.18
22:42:17:WU01:FS01:0x21:  Found a checkpoint file
22:42:21:WU00:FS00:0xa7:Completed 99615 out of 250000 steps (39%)
22:42:56:Removing old file 'configs/config-20170624-043515.xml'
22:42:56:Saving configuration to config.xml
22:42:56:<config>
22:42:56:  <!-- Folding Core -->
22:42:56:  <checkpoint v='10'/>
22:42:56:
22:42:56:  <!-- Network -->
22:42:56:  <proxy v=':8080'/>
22:42:56:
22:42:56:  <!-- Slot Control -->
22:42:56:  <power v='full'/>
22:42:56:
22:42:56:  <!-- User Information -->
22:42:56:  <passkey v='********************************'/>
22:42:56:  <team v='224497'/>
22:42:56:  <user v='ader'/>
22:42:56:
22:42:56:  <!-- Folding Slots -->
22:42:56:  <slot id='0' type='CPU'>
22:42:56:    <cpus v='10'/>
22:42:56:  </slot>
22:42:56:  <slot id='1' type='GPU'/>
22:42:56:</config>
22:42:56:WU01:FS01:0x21:Completed 125000 out of 2000000 steps (6%)
22:42:56:WU01:FS01:0x21:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
22:43:24:WU00:FS00:0xa7:Completed 100000 out of 250000 steps (40%)
22:55:17:WU00:FS00:0xa7:Completed 102500 out of 250000 steps (41%)
22:59:35:WARNING:WU00:FS00:Detected clock skew (1 mins 12 secs), adjusting time estimates
22:59:35:WARNING:WU01:FS01:Detected clock skew (1 mins 12 secs), adjusting time estimates
23:08:34:WU00:FS00:0xa7:Completed 105000 out of 250000 steps (42%)
23:20:38:WU00:FS00:0xa7:Completed 107500 out of 250000 steps (43%)

[Joe_H]

Welcome to the folding support forum.

At first glance there does not appear to be a problem with your setup. What version of the video drivers is installed on your system?

Beyond that, the GPU core does take a while to start up which is done on the CPU. Though half an hour seems too long. After that the processing is done mostly on the GPU, though from reports the current Core is not particularly fast on older Terascale based cards. Core_21 does better on the GCN architecture.

[ader]

Hi Joe_H, I've monitored the GPU usage, and it remains the same < 1% all the time. Also, I've installed the latest driver from AMD (which supports Crimson Software for legacy GPUs), it comes with Catalyst 15.8 and Crimson 16.2.1; Driver 15.301.1901-160226a-299900E.

The work unit shows FahCore 0x21, is that the same as Core_21?

Thank you for your help

[Joe_H]

Hopefully this problem looks familiar to someone else, I have run out of ideas on what the problem might be. Used to have a 5870 and Windows system to test on, but that is not usable anymore. I was able to test the previous GPU cores, 16, 17 and 18 on that.

Oh, and yes those are two different ways the core can be identified. If you watch the active core process in Task Manager it will be identified as FahCore_21.

[JimboPalmer]

I found nothing that says Core 21 needs GCN, but in looking I found this viewtopic.php?f=74&t=28043

This hints it may be too slow once you get it working viewtopic.php?f=96&t=29598

viewtopic.php?f=81&t=28336&start=15#p281538

[bruce]

Welcome to foldingforum.org, ader

Possible problematic lines from your log:

Code: Select all

22:42:56:WU01:FS01:0x21:Completed 125000 out of 2000000 steps (6%)
22:42:56:WU01:FS01:0x21:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900

22:59:35:WARNING:WU00:FS00:Detected clock skew (1 mins 12 secs), adjusting time estimates
22:59:35:WARNING:WU01:FS01:Detected clock skew (1 mins 12 secs), adjusting time estimates

This WU has progressed to 5^, so at one time, you had it running. What's interesting is that you're repeatedly getting reports of clock skew after the WU starts. (I don't know why.) Can your Power Supply handle the 12v load? Is your CMOS battery in need of replacement.

We've gotten mixed reports regarding the Catalyst and Crimson drivers. Have you tried the legacy catalyst drivers? (Last I heard, Crimson was still considered beta)

According to the information I have, the HD5870 is capable of producing 2720 single precision GFLOPS (plus some DP flops) so it's fast enough to fold.

[ader]

Oh, and yes those are two different ways the core can be identified. If you watch the active core process in Task Manager it will be identified as FahCore_21.

ok, Thanks Joe

[ader]

but in looking I found this ...

Thanks Jimbo, I'm reviewing those threads, I'll give that fah benchmark a try tomorrow.

[ader]

Possible problematic lines from your log:
22:42:56:WU01:FS01:0x21:Completed 125000 out of 2000000 steps (6%)
22:42:56:WU01:FS01:0x21:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900

22:59:35:WARNING:WU00:FS00:Detected clock skew (1 mins 12 secs), adjusting time estimates
22:59:35:WARNING:WU01:FS01:Detected clock skew (1 mins 12 secs), adjusting time estimates

Yeah, I don't know why the log entry says 'Nvidia GPU', never had an nvividia GPU or driver installed on this machine, only AMD variants. Maybe FAH is misinterpreting some error it may be getting ?

This WU has progressed to 5^, so at one time, you had it running. What's interesting is that you're repeatedly getting reports of clock skew after the WU starts. (I don't know why.) Can your Power Supply handle the 12v load? Is your CMOS battery in need of replacement.

I'm not sure what the clock skew means, it might be the battery like you suggested. In anycase, the HD5870 seems to work well, ran the 3DMark benchmarking sortware on it, had low framerate but it was functional. Also ran the FAH on the same machine against an RX580, that GPU worked fine too and WUs completed in hours instead of days ( like 14+ days).
And, yes power supply is 1100watts and this GPU is the only peripheral in addition to standard PC components (HD/SDD/DVDROM); so plenty of power.

We've gotten mixed reports regarding the Catalyst and Crimson drivers. Have you tried the legacy catalyst drivers? (Last I heard, Crimson was still considered beta)

I had the older driver, but the recent versions of Catalyst like to update the driver automatically, so at some point it got updated. Will try to dig up an old version of catalyst/driver and install it to see if it works.

According to the information I have, the HD5870 is capable of producing 2720 single precision GFLOPS (plus some DP flops) so it's fast enough to fold.

Thanks for info and your help Bruce.
l8r

[bruce]

Possible problematic lines from your log:
22:42:56:WU01:FS01:0x21:Completed 125000 out of 2000000 steps (6%)
22:42:56:WU01:FS01:0x21:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900

Yeah, I don't know why the log entry says 'Nvidia GPU', never had an nvidia GPU or driver installed on this machine, only AMD variants. Maybe FAH is misinterpreting some error it may be getting ?

The log says you have a Cypress [Radeon HD 5800/6800.

That part of the log is explaining why FAHClient is not managing the temperature of your GPU.
(FAH only attempts to control the temperature if you have a single NVidia GPU and it's not particularly effective anyway. Temperature control really should always be the responsibility of the drivers and the fan.)

[ader]

That part of the log is explaining why FAHClient is not managing the temperature of your GPU.

ok, got it

[ader]

but in looking I found this ...

Hi Jimbo, I gave that FAHBench (from one of the threads you suggested) a try, seems like it didn't recognize the GPU, says CUDA error:



Thx

[ader]

FAHBench crashed when running the CPU test

[JimboPalmer]

It seems to be choking on OpenGL support. I suggest the Catalyst driver here http://support.amd.com/en-us/download/d ... 7%20-%2064

[bruce]

I agree with JimboPalmer.

We've gotten mixed reports regarding the Catalyst and Crimson drivers. Have you tried the legacy catalyst drivers? (Last I heard, Crimson was still considered beta)

I had the older driver, but the recent versions of Catalyst like to update the driver automatically, so at some point it got updated. Will try to dig up an old version of catalyst/driver and install it to see if it works.

If you were running Windows10, that might be a problem. You have Windows 7 Professional which doesn't force you to update.

I have to ask about why the driver is being updated. Presumably the last version of Legacy Catalyst drivers still works and there no particular reason to update. I suspect the first 8% of that WU was completed before you installed Crimson.

The message from FAHBench saying you have a CUDA error means that it's trying to run drivers for NVidia which is wrong. It needs to be running OpenCL for AMD.