Hi everyone, i hope im writing in proper section,

i have just downloaded the latest night build and i was happy with log info related to cpu limitations. I have dual xeon 2698 v4 (80 threads in total) and i was hoping i will able to use all of the power of both cpus. So i did few quick tests and here is what i have found:

- flip solver (batch/cmd) was working great used all 80 threads
- vdb, clouds oparations was also using 100% of cpus

but when i tried to render these clouds mantra used only 1 cpu (40 threads) or just a group0 (cpus)

Im going to attach the images so you can see everything, once again great work sidefx team

best regards

From the MPlay window, it definitely looks like Mantra has 80 render threads. Testing just now, (with forced processor groups), it looks like somehow, all of the render threads get boxed into processor group 1, even though some threads are being assigned to processor group 0 and some are assigned to processor group 1. I'll try to figure it out.

Houdini 16.0.537, which should be available in a few hours, hopefully has this fixed. Something that's supposedly unused apparently has to be explicitly set to zero, else the processor groups don't get assigned correctly. I'll remove the debug printouts about setting the processor groups once we have confirmation that it works for people.

Hi This is old thread but
Have you tried adjusting boot parm groupaware on and/or groupsize ?

https://docs.microsoft.com/en-us/windows-hardware/drivers/devtest/bcdedit--set
Edit by ndickson: Changed link to fix two dashes being replaced with en dash.

creepmaster
Have you tried adjusting boot parm groupaware on and/or groupsize ?

Yep. That's how I was able to test it on a machine with only 12 (logical) cores. However, on a machine with more than 64 cores, you can't make a processor group larger than 64 with it, unfortunately.