AlanTheBeast wrote:@PantherX : it's actually showing "cpu:11" rather than 12 that it had before...
While that is correct on the interface, FahCore_a7 will automatically change that to 10 IIRC and the line in the log file just before it starts a WU would be:
Calling: mdrun -s frame88.tpr -o frame88.trr -x frame88.xtc -cpt 30 -nt 10
Reason for the change of 11 to 10 is that prime numbers don't mix well with folding. It tends to throw a domain decomposition error. You can search for details here in the forum but in short, the WU can't be successfully split across X number of CPUs.
AlanTheBeast wrote:...I'm on my iMac VNC-ing to the PC with the GPU. It's quite sluggish - which it wasn't yesterday - I guess the GPU load is part of that since VNC is reporting the screen to the Mac...
That would be the hardware acceleration which incorrect assumed that your GPU is exclusively for it's own use. See if you can disable hardware acceleration from VNC:
https://sourceforge.net/p/vnc-tight/mai ... e/7904850/
AlanTheBeast wrote:...If I force the cpu count to 10, do you think it will, overall, be better for the GPU staying loaded? I don't care that VNC is sluggish - that's off most of the time anyway.
Note that the PC is doing nothing but folding with 1 project on CPU and another on GPU.
I just set the CPU to "10", but all 12 are near saturated. Maybe needs to get to the next "PF" to downshift or finish the project to change...
The log posted above showed you started with 12 CPUs (value after -nt):
14:24:55:WU00:FS00:0xa7:Calling: mdrun -s frame1.tpr -o frame1.trr -cpt 15 -nt 12
Unless I can see a fresh copy of your log file, can't say what is happening.
Since you have a dedicated Linux box, the set-up would be:
CPU Slot - Uses 10 CPUs
GPU Slot - Uses 1 CPU
Free CPU is 1 which can be used by the OS for any overhead stuff. That would be the ideal setup to ensure that neither your CPU/GPU have to "fight" over CPU cycles.