Other CPU Performance Patch (Up to 30% faster physics, up to 60% more FPS)

[hijku]

So cool, thank you for sharing these!

Intel 13900K/13800K/13700K
(we are using the performance cores only)




Using 13900k, just to confirm, that Process Lasso and this performance patch are reporting i9 performance cores with different numbers: 0,2,4,6,8,10,12,14 vs. 1,3,5,7,9,11,13,15.

So I guess that's ok then, right?

That's a good question actually.
@turtlebackgoofy I assume lasso uses 0-based indexing and your patch starts with 1?

 

[thinoreos]

Make sure you're not GPU limited. I noticed that VD encoding is pretty GPU demanding, even for 4090.
The easiest way to check that is to lower the render resolution inside VAM (from 1.0 to 0.75) and see if you're getting more fps. If you're then 4090 is too weak.
That's why I switched VD from Godlike preset to the lower one.

Just tried this, lowered VD to Ultra preset, lowered res and all.
With the patch in VR I get about +3fps more than without it at lower res. Are we capped out with our 4090s at this? There's no more performance gains to be had?

 

[turtlebackgoofy]

That's a good question actually.
@turtlebackgoofy I assume lasso uses 0-based indexing and your patch starts with 1?

yes, I just assumed for my config that it would be easier to understand for non-tech people. Usualy in computer stuff every index starts with 0

 

[huiyu914]

i have already used this plug, but this plug only boost the speed at the beginning, i need load scene/person/cloth... speed up ,not load game speed.

I use this: https://github.com/sFisherE/vam_plugin_release/tree/main

 

[Kawum]

could be motion smoothing. If you drop from 91fps to 89fps rendering fps, the displayed fps drops from 90 to 45 fps.

Hi. Looks like HT was the issue. Forgot to switch it off again in bios after mucking about. Only noticed as vanilla install did not had the issue. But affinity=1,2,3,4,5,6 would caused HT to kick in I assume only using 3 cores with HT?
All goot after switiching HT off again. Leaving 2 performance cores for whatever VD/Steam/OS have to do.
Don't have Lasso but could manually set affinity for VD and Steam via PowerShell if needed.
Noticed however that FPS max out at around 90 for vanilla and patched. Which is fine. Will try more complex scene next.
Thats my settings btw for VR:
[threadsVR]
computeColliders=6
skinmeshPart=6
applyMorphs=6
skinmeshPartMaxPerChar=6
applyMorphMaxPerChar=6
affinity=1,2,3,4,5,6
Boot.config
wait-for-native-debugger=0
gfx-enable-gfx-jobs=1
gfx-enable-native-gfx-jobs=1

 

[Kawum]

Just tried this, lowered VD to Ultra preset, lowered res and all.
With the patch in VR I get about +3fps more than without it at lower res. Are we capped out with our 4090s at this? There's no more performance gains to be had?

Hi. Noticed the same. The patch definately increases desktop performance for me but did not affect VR so far compared to vanilla install . That is using VD btw and affinity to 6 P-cores on 13900k. More tweaking and monitoring to be done.

 

[huiyu914]

I use this: https://github.com/sFisherE/vam_plugin_release/tree/main

i have already used this plug, but i don't need speed up at begeinning load game, i need speed up at load scene/cloth/skin...

 

[stylesi]

settings are fine, but I would suggest to set the same non-vr settings in vr and use processlasso to put the non-vam processes onto the efficiency cores (steamvr, oculus, whatever you use)

I used the Process lasso affinity presets tweak today. However, I put VD to use performance cores with vam.exe too. Then it works fine with VD Ultra preset and I am getting almost double fps in VR. 4090 and 13900K watercooled.

 

[VeeRifter]

@turtlebackgoofy
Unfortunately your patch introduces a bug with the built-in Clothing Creator. When you attempt to import any .duf file from Daz it fails, throwing the following error:

Exception during DAZ import: System.NullReferenceException: Object reference not set to an instance of an object
at DAZMorph.set_morphValue (Single value) [0x00000] in <filename unknown>:0
at MeshVR.DAZImport+<ImportDufCo>c__Iterator1.MoveNext () [0x00000] in <filename unknown>:0

Reverting back to the original Assembly-CSharp.dll fixes things and .duf import will work correctly once again.

Hope you can fix this since, apart from the bug, performance in VAM has improved significantly.

 

[turtlebackgoofy]

@turtlebackgoofy
Unfortunately your patch introduces a bug with the built-in Clothing Creator. When you attempt to import any .duf file from Daz it fails, throwing the following error:

Exception during DAZ import: System.NullReferenceException: Object reference not set to an instance of an object
at DAZMorph.set_morphValue (Single value) [0x00000] in <filename unknown>:0
at MeshVR.DAZImport+<ImportDufCo>c__Iterator1.MoveNext () [0x00000] in <filename unknown>:0

Reverting back to the original Assembly-CSharp.dll fixes things and .duf import will work correctly once again.

Hope you can fix this since, apart from the bug, performance in VAM has improved significantly.

thanks for reporting, will look into it

 

[thinoreos]

I used the Process lasso affinity presets tweak today. However, I put VD to use performance cores with vam.exe too. Then it works fine with VD Ultra preset and I am getting almost double fps in VR. 4090 and 13900K watercooled.

Send screenshots of fps before and after using the patch in VR, double fps seems a bit sus

 

[Krauser]

What settings should I use for this build?

I tried something out and fps actually went down after the patch lol

 

[Eye9]

I wish unity had a marker to tell the GC "this is something that will never get freed, put it on a different heap and let me GC collect it manualy"

Well there is GCHandle.Alloc (from .net) which will mark something as do-not-collect, but I don't know if it will put it onto a different heap.

 

[VamMoose]

Defiantly improved load times significantly on my 5900x and RTX 3080! Well done OP!

FPS is about the same. I did use the recommended configuration in the .ini file, Anyone else able to recommend settings that might help with FPS? I am mostly hovering around 45 FPS on my Valve Index set to 90hz which has a little bit of reprojection. Enjoyable sure, but would be nice to get higher FPS.

 

[turtlebackgoofy]

Well there is GCHandle.Alloc (from .net) which will mark something as do-not-collect, but I don't know if it will put it onto a different heap.

it is the native heap itself, but implementing that into the json deserializer would be a mess

 

[kretos]

I'm using following settings for i5-10400f + 3060ti:

SkinMeshPartDLL.ini:

affinity=1,3,5,7,9,11
computeColliders=6
skinmeshPart=6
applyMorphs=6
skinmeshPartMaxPerChar=12
applyMorphMaxPerChar=12

boot.config:

gfx-enable-gfx-jobs=1
gfx-enable-native-gfx-jobs=1
gfx-disable-mt-rendering=0
wait-for-native-debugger=0
vr-enabled=0
hdr-display-enabled=0
gc-max-time-slice=3
job-worker-count=12
single-instance=1

VAM (Desktop Mode).bat:

added after -vrmode -nolog -nostereorendering

I have no fps gain but this gives me best loading times. Not sure if some of those switches actually works - founded on internet.
Used together with var browser and caching on ssd.

If only textures wouldn't need caching it would a pleasure to use VAM...

 

[Seraphim]

Send screenshots of fps before and after using the patch in VR, double fps seems a bit sus

I got 27% more fps in a plugin morph heavy test scene. Dont rememer how much in the baseline 3 benchmark I think in VR it depends on how heavy the cpu physics (morphs) are in the scene for how much of a difference you'll see with the patch. I believe turtlebackgoofy has said so himself. Plus idk about intel.

 

[thinoreos]

I got 27% more fps in a plugin morph heavy test scene. Dont rememer how much in the baseline 3 benchmark I think in VR it depends on how heavy the cpu physics (morphs) are in the scene for how much of a difference you'll see with the patch. I believe turtlebackgoofy has said so himself. Plus idk about intel.

Yup getting the feeling this might not work as well for us intel users

 

[Wxy_332]

I used the Process lasso affinity presets tweak today. However, I put VD to use performance cores with vam.exe too. Then it works fine with VD Ultra preset and I am getting almost double fps in VR. 4090 and 13900K watercooled.

I also tried as your mentioned, Process lasso CPU affinity setting VD and VAM.exe to only performance cores, but no improve at all. Wried.

13700KF and 4090

 

[rpms]

So this is officially my favorite thread on the hub! Curious for any pointers for my 13900K/3080FTW/128GB machine though....

On desktop, and at normal benchmark settings, I used a fresh install of vam, with the latest patch (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

[threads]
computeColliders=8
skinmeshPart=1
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

I could only see +- 2 frames of difference though.

SO, I wondered if my gpu was bottlenecking, and ran benchmarks below with lower quality settings.
Most results are still similar, other than for baseline 3, which did show ~19% faster:

Spoiler: Clean:

Spoiler: Patched (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

I'm jealously looking at Trety's massive numbers across the board and wondering if I'm missing anything... or if it's just my lesser gpu holding my numbers back?

 

[fsdcx]

So this is officially my favorite thread on the hub! Curious for any pointers for my 13900K/3080FTW/128GB machine though....

On desktop, and at normal benchmark settings, I used a fresh install of vam, with the latest patch (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

[threads]
computeColliders=8
skinmeshPart=1
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

I could only see +- 2 frames of difference though.

SO, I wondered if my gpu was bottlenecking, and ran benchmarks below with lower quality settings.
Most results are still similar, other than for baseline 3, which did show ~19% faster:

Spoiler: Clean:

View attachment 337249

Spoiler: Patched (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

View attachment 337250

I'm jealously looking at Trety's massive numbers across the board and wondering if I'm missing anything... or if it's just my lesser gpu holding my numbers back?

Why not use default setting to play benchmark?

 

[Kawum]

So this is officially my favorite thread on the hub! Curious for any pointers for my 13900K/3080FTW/128GB machine though....

On desktop, and at normal benchmark settings, I used a fresh install of vam, with the latest patch (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

[threads]
computeColliders=8
skinmeshPart=1
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

I could only see +- 2 frames of difference though.

SO, I wondered if my gpu was bottlenecking, and ran benchmarks below with lower quality settings.
Most results are still similar, other than for baseline 3, which did show ~19% faster:

Spoiler: Clean:

View attachment 337249

Spoiler: Patched (trety's settings) + bootconfig + processlasso (set to all perf cores + HT off).

View attachment 337250

I'm jealously looking at Trety's massive numbers across the board and wondering if I'm missing anything... or if it's just my lesser gpu holding my numbers back?

Hi, Looks kike it;s your 3080.
Compare like for like vanilla install HT on in both cases as I did not find much difference with HT off, except of CPU running a bit cooler.
Don't use lasso as it will introduce another variable.
Stock:

With patch:

Again this is just a benchmark. I can see gains in more complex scenes in desktop mode, so more of a 'if it feels right it is' scenario. VR is a bit of a mixed bag as you add another couple of layers of more or less efficiently developed software with settings, not even metioning the ongoing battle to keep windows lean.
Laters

 

[turtlebackgoofy]

Hi, Looks kike it;s your 3080.
Compare like for like vanilla install HT on in both cases as I did not find much difference with HT off, except of CPU running a bit cooler.
Don't use lasso as it will introduce another variable.
Stock:
View attachment 337301
With patch:
View attachment 337302
Again this is just a benchmark. I can see gains in more complex scenes in desktop mode, so more of a 'if it feels right it is' scenario. VR is a bit of a mixed bag as you add another couple of layers of more or less efficiently developed software with settings, not even metioning the ongoing battle to keep windows lean.
Laters

there is still something going on with GPUs, it just doesnt feel right that a 3080 is struggling with THAT level of graphics. But maybe its just the old unity engine not rendering fast on current gen GPUs.

 

[Kawum]

there is still something going on with GPUs, it just doesnt feel right that a 3080 is struggling with THAT level of graphics. But maybe its just the old unity engine not rendering fast on current gen GPUs.

Indeed. Comparing performance differences between 3080 and 4090 are not that dramatic in VAM when comparing with newer titles. I took the 1080, 2080ti, 3080ti, 4090 path and from memory the biggest jump was between 1080 and 2080ti. After that I could have spend the surplus money on something else. However FSX does look pretty. The main performace impact with VAM was always adding additional resources. E.g. 8 character scenes for heating the house in the winter, who needs a heat pump if you have VAM .
One way to try it out could be to compare same scene between current and VAM2. Or compare Unity 2022.3 with older Unity version. Question though is cost vs benefit. So this patch together with var browser helped the game a lot.

 

[fsdcx]

there is still something going on with GPUs, it just doesnt feel right that a 3080 is struggling with THAT level of graphics. But maybe its just the old unity engine not rendering fast on current gen GPUs.

Can you improve hair performance? I think this is such a big thing