So overnight I tried crunching 8 tasks simultaneously with my threadrippers and Radeon VII's. One pc keeps throwing up a driver hanging (20.11.2) and the wu's essential stop crunching (probably hardware issue?) The other threadripper with a Radeon VII didn't throw up a driver issue but it was clearly stuck. Each gpu wu had run for over 4hrs last night with no result. Restarting BOINC either ditches all that work and starts again from scratch, or starts again from a previous check point on those work units.

I didn't see this issue running 4 w/u's at a time so I'm going back to that.
This only seems to affect my pc's with Radeon VII's. My other pc's running multiple wu's seem ok.

FWIW, I have no such issues running either Milkyway or Einstein at home with multiple wu's running and far greater power draw.

I'll be stopping this testing shortly with the Pentathlon fast approaching.

My PC is a Ryzen 5 3600 with a Nvidea GT 1030. I consider the GPU to be in the low to average power range.
It was successfully running 1 OPNG at a time, mostly at 100% but with many drops to zero. The temperature of the GPU fluctuated, dropping from 50 deg C to 30 deg C every 20 to 30 minutes as each work unit completed. I decided to try running 2 OPNG at the same time. Yesterday there were only 7 such drops in 12 hours i.e. it was much more steady.
The PC averaged about 70k running OPN1 only. It was 700k running 12 OPN1 and 1 OPNG and increased to about 900k running 11 OPN1 and 2 OPNG. The effect on temperature is more important to me than the additional points but hopefully it does mean more work is achieved.

AutoDock for GPUs
Improved calculation routines using OpenCL and CUDA have been developed by the AutoDock Scripps research team.[18]
It results in observed speedups of up to 4x (quad-core CPU) and 56x (GPU) over the original serial AutoDock 4.2 (Solis-Wets) on CPU.
The CUDA version was developed in a collaboration between the Scripps research team and Nvidia. It is currently faster than the OpenCL version.[9][18]

https://en.wikipedia.org/wiki/AutoDock

Dear all!!

Could you please advise how to set up my Win10 client to run more than 1 GPU task?
I have BOINC client version 7.14.3 for windows_intelx86.

As far as I understand I have to put some manual corrections to cc_config.xml or something...

This should ease your pain. You create the app_config.xml using a text editor (Notepad is OK), save it as an ".xml" file (not a .txt file), and place it in your WCG project folder.

<app_config>
<app>
   <name>opng</name>
   <gpu_versions>
   <cpu_usage>1.0</cpu_usage>
   <gpu_usage>0.5</gpu_usage>
   </gpu_versions>
</app>
</app_config>

You then need to activate it, as by restarting BOINC.

This runs two GPU work units at once. To run three, just reduce the 0.5 to 0.33, etc.