FAH GPU Client: No work units

402blownstroker

[H]ard|DCer of the Month - Nov. 2012
Joined
Jan 5, 2006
Messages
3,242
I got a box setup for GPU folding. It is running Cent OS 6, a NVidia 580 card, and version 7 of the client. Also running 319.60 of the NVidia driver. After about 7 hours it has not pulled down a WU. Does anyone see any issues with the config file?

Code:
<config>
  <!-- Folding Core -->
  <checkpoint v='3'/>

  <!-- Folding Slot Configuration -->
  <power v='full'/>

  <!-- Network -->
  <proxy v=':8080'/>

  <!-- User Information -->
  <team v='33'/>
  <user v='402blownstroker'/>

  <!-- Folding Slots -->
  <slot id='1' type='GPU'>
    <cuda-index v='0'/>
    <gpu-index v='0'/>
    <opencl-index v='0'/>
  </slot>
</config>

Passkey has been removed....
 
Try client-type = advanced on the GPU slot; after restart the client you would have the chance to get WU from public beta and core 17; they might be unstable, but not often.
 
  • Click on the Configure icon in the v7 client
  • Select the Slots tab
  • Highlight the ID and click Edit
  • The Folding Slot window opens. Leave the increments at -1
  • In the Extra Slot Options at the bottom, click Add
  • In the Name field type client-type
  • In the Value field type advanced or beta
The part of the xml file that looks odd to me is that you have v='0' values hard-coded into the config.xml for cuda, gpu, and opencl index values. Both my single GPU systems and multi-GPU systems exclude specifying these in the config.xml file. Try editing the values per the fourth bullet point above and let the v7 client save the file. Stop and start the v7 client or restart after the "Saving" dissapears from the bottom of the v7 client window.
sudo /etc/init.d/FAHClient stop
sudo /etc/init.d/FAHClient start

Check this guide under the Requirements Section and the whitelist linked in that section.

Example config.xml single GPU
<config>
<!-- Folding Slot Configuration -->
<power v='full'/>

<!-- HTTP Server -->
<allow v='127.0.0.1'/>

<!-- Network -->
<proxy v=':8080'/>

<!-- Remote Command Server -->
<password v='0000'/>

<!-- Slot Control -->
<pause-on-start v='true'/>

<!-- User Information -->
<passkey v='0000'/>
<team v='33'/>
<user v='Mike_Shaffer'/>

<!-- Work Unit Control -->
<next-unit-percentage v='99.9'/>

<!-- Folding Slots -->
<slot id='0' type='GPU'>
<client-type v='advanced'/>
</slot>
</config>
 
Last edited:
Hmmm.... with adding in the client-type to the existing config file or using the above sample, the core is downloaded and then instantly dies:

Code:
00:01:40:WU01:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/www.stanford.edu/~pande/Linux/AMD64/NVIDIA/Fermi/Core_17.fah/FahCore_17 -dir 01 -suffix 01 -version 703 -lifeline 4037 -checkpoint 15 -gpu 0 -gpu-vendor nvidia
00:01:40:WU01:FS00:Started FahCore on PID 4069
00:01:40:WU01:FS00:Core PID:4073
00:01:40:WU01:FS00:FahCore 0x17 started
00:01:40:WARNING:WU01:FS00:FahCore returned: FAILED_2 (1 = 0x1)
 
In the "The Folding Slot window opens. Leave the increments at -1" make sure that the gpu-index is also -1 to let the v7 client find the right slot. In the single GPU the GPU is slot-id=0. In both multi-gpu the slot-id =0 for the first card and slot-id=2 for the second card. While the slot sequence matches the card placement in the MB, the one multi-gpu has the card in slot 0 and the second card 3 pci slots away. It is likely that your slot-id=1 was correct whereas my slot-id=0 for your system is not.
 
Well that did not seem to make a difference. As soon as the core engages, it stops with the 'FAILED_2 (1 = 0x1)'. At least the wine and cuda wrapper worked. Good job Stanford on the native GPU client.
 
Must use the OEM drivers from NV. Open source drivers don't work.
Must have client-type = advanced, already done I see.
 
<!-- Work Unit Control -->
<next-unit-percentage v='99.9'/>


Probably not the cause, as you may not have copied this verbatim from Mike, but whole numbers only, between 90-100.

Would also need to see more of the log... ***System*** section, and the part where fahcore 17 starts up through where it fails.


Did you set these 3 indexes manually?

<slot id='1' type='GPU'>
<cuda-index v='0'/>
<gpu-index v='0'/>
<opencl-index v='0'/>
</slot>

If so, try again without them...
 
Last edited:
Back
Top