I've been messing with this for the past 4 hours. Here is the setup I tested it on:I just tried with my intel 12700H and 3070Ti Mobile, I have no difference. I think I set it wrong. I left 1,3,5,... because I didn't disable HT. I also left engine affinity. Can someone guide me?
Original :
View attachment 403310
Patched V13beta1
View attachment 403311
settings :
[threads]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=4
applyMorphs=4
applyMorphMaxPerChar=4
#affinity=1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16
affinity=1,3,5,7,9,11
#engineAffinity=1,3
[threadsVR]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=4
applyMorphs=4
applyMorphMaxPerChar=4
affinity=1,3,5,7,9,11
#engineAffinity=1,3
[profiler]
enabled=0
And the boot.config file :
gfx-enable-gfx-jobs=1
gfx-enable-native-gfx-jobs=1
gfx-disable-mt-rendering=0
wait-for-native-debugger=0
gc-max-time-slice=3
job-worker-count=12
single-instance=1
The command : Get-Process | select VaM.exe -expand Modules -ea 0 | where {$_.ModuleName -like 'skinmeshpartdll.dll'} :
result : 1136 SkinMeshPartDLL.dll
Gigabyte Aero 16XE5
i7-12700H
RTX 3070 Ti Laptop GPU
32GB DDR5 RAM (4800)
The i7-12700H has 6 Performance Cores and 8 Efficiency (total 14). The Performance cores are Hyperthreaded and I'm pretty sure it is not disabled due to the task manager showing 14 cores but 20 logical processors. According to all of what I've read in this thread the idea is to only use the P cores and skip every other thread. However, with about 2 hours of testing, the best setup I found was this. Anything less than using 12 resulted in less performance increase.
[threads]
computeColliders=12
skinmeshPart=12
skinmeshPartMaxPerChar=12
applyMorphs=12
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,9,10,11,12
affinity=1,2,3,4,5,6,7,8,9,10,11,12
engineAffinity=1,2,3,4
[threadsVR]
computeColliders=12
skinmeshPart=12
skinmeshPartMaxPerChar=12
applyMorphs=12
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,9,10,11,12
affinity=1,2,3,4,5,6,7,8,9,10,11,12
engineAffinity=1,2,3,4
[profiler]
enabled=0
I have minimal FPS increase however, the "wait" time is pegged at 0 and the FPS are much more stable. I think that is the best I can get from this at the moment but, an improvement is an improvement. I've not tried that rig's setup in VR yet though to see if there is any change. The scene I tested it on was a work in progress scene I have with a bunch of complex CUAs and 2 people. One of them wearing 4 pieces of SIM clothing. Closer I get to a figures the FPS drops to around 60fps. When I'm further away from a figure it is upwards of 100FPS. However, just reducing the hair physics bumped both near and far up by about 20FPS each. Physics update settings are set at "auto".
Additionally, I ran a test at 18 cores and using all of them. Task manager showed usage but, there was no improvement beyond what I had gotten with just using P cores. Give this setup a shot and let me know if it does anything for you.
***Update***
I removed (1) piece of SIM clothing, cut hair physics in half (not much of noticeable difference), and removed (2) of the CUAs (that were animated - there is some sort of render issues with them causing frames to drag). Consistent 120-130 FPS on desktop and 60-90 FPS in VR. I've never had 90 FPS in VR before on that laptop. For VR I was using a Quest 2 with virtual desktop and hard wired with link cable to laptop. @Makimax - try the settings I mentioned and let me know what it does for you.
Last edited: