tip for anyone here looking to match settings to CPU, ChatGPT can now suggest settings tailored to your CPU. Here's what I asked it:
in skinmeshpart.dll for Virt-A-Mate, I have the following settings:
[threadsVR]
computeColliders=8s
kinmeshPart=8
skinmeshPartMaxPerChar=6
applyMorphs=8
applyMorphMaxPerChar=6
#affinity=0,2,4,6,8,10,12,14
Which settings are most appropriate for an Intel i-9 9900k?
GhatGPT:
Here’s the config I’d recommend:
[threadsVR]
computeColliders=8
skinmeshPart=8
skinmeshPartMaxPerChar=3
applyMorphs=8
applyMorphMaxPerChar=3
#affinity=0,2,4,6,8,10,12,14
Why this is the “heavy 2-char safe” setup
- computeColliders=8 → full use of your 8 physical cores for collision physics, which is usually the biggest CPU load in VaM.
- skinmeshPart=8 / applyMorphs=8 → plenty of worker threads in the pool, so both characters can run in parallel.
- MaxPerChar=3→ keeps each character capped at 3 threads for morphs + 3 for skinmesh.
- 2 characters × 3 = 6 threads each → 12 threads total across morph + mesh.
- This stays comfortably within your 8 physical cores, with some spillover into SMT but not overload.
- Affinity left commented → Windows can spread workload as needed. Only enable if you notice inconsistent thread usage or cache thrash.
Tried it and I do seem to be getting better performance! Who would have thought ChatGPT would know about vam