Benchmark Result Discussion

[moki246]

5800X3d - 3600 64Gb - RTX4080

CPU Performance Patch (Up to 30% faster physics, up to 60% more FPS) - Plugins + Scripts -

As requested in this thread https://hub.virtamate.com/threads/benchmark-result-discussion.13131/page-37 here is a release of the cpu performance patch. FAQ at bottom Supports VaM 1.22.0.13! (Update via VaM Updater, THEN copy the performance...

hub.virtamate.com

I tested hot new patch and result :


But in real play, I can not feel more FPS. nevertherless, more faster than before about loading and low fps etc.


70 sec loading scene turn into 55 sec loading scene.



AND now, version 11 tested.
simpler, hair mirror down and low fps up 88 to 100.
-> that was incorrect test.

interesting....


-2024. 2. 11.
version 12. scene loading 55sec to 21sec. how could it be.....

but on my testing scene I could not feel changes.

nevertheless it is crazy... day by day improvement......


---
-2024. 2. 11.
on running mmd 4 person test.

I confirmed changes.

avg FPS 40 to 50 ..

But still freezing...that is zhfx`s problem....


bench result is true. there was about 20% imprvements.



before patch

after patch ver 10

Version 11

version 12

 

[jpegxl]

Update：5800X3D+4070 SUPER
With 5000+ Vars

Fresh install

Fresh install with CPU Performance Patch V11

 

[checkking]

I got very underperforming results for my rig, especially with Baseline 3. Anybody knows what I could try to improve?

Edit:
Results with patch v11 and avx1:

This is 77% increased average FPS with Baseline 3!!! I don't know why my results were so bad, considering that I am CPU bottlenecked and the 5800X performs better with all other benchmarks except mine, but maybe the patch is fixing it.

 

[tuxxxedo_mask]

Interesting, you get MORE FPS in Baseline3 benchmark now than in Simpler Physics. Both are the exactly same scene, but simpler physics has soft body turned off. You are getting MORE FPS with soft body turned ON lmao. Would you mind sharing a performance profile like discribed in the resource description?

After upgrading to version 12, I reran the benchmark test.
Interestingly, I obtained the same perplexing results: baseline 3 outperformed simpler physics.

My performance profile was set to the default configuration.

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

As per your instructions, I have generated the ThreadProfile.csv output.
Please change the prefix to "zip" and then extract the file.
I hope this will render it somewhat usable for your work.

 

[atani]

My results in desktop mode with a ultrawide screen (1440p):

CPU: AMD Ryzen 9 3900X 12 core (no OC)
GPU: NVIDIA GeForce 3070 Ti 8GB (no OC)
RAM: 32 GB 2400 MHz (no XMP profile set as it's unstable AF)
VaM folder of 111 GB with around 2150 VARs

Regular benchmark:

Running with the v12 patch:

Baseline 3 is close to a 100% increase on average ?

Spoiler: Skinmesh ini settings

No idea if this is optimal for my CPU, my knowledge of the concepts is low and information online is confusing AF
[threads]
computeColliders=6
skinmeshPart=4
applyMorphs=2
skinmeshPartMaxPerChar=2
applyMorphMaxPerChar=4
affinity=1,3,5,7,9,11

[threadsVR]
computeColliders=4
skinmeshPart=1
applyMorphs=2
skinmeshPartMaxPerChar=2
applyMorphMaxPerChar=4
affinity=1,3,5,7,9,11

[profiler]
enabled=0

 

[turtlebackgoofy]

My results in desktop mode with a ultrawide screen (1440p):

CPU: AMD Ryzen 9 3900X 12 core (no OC)
GPU: NVIDIA GeForce 3070 Ti 8GB (no OC)
RAM: 32 GB 2400 MHz (no XMP profile set as it's unstable AF)
VaM folder of 111 GB with around 2150 VARs

Regular benchmark:
View attachment 334066

Running with the v12 patch:
View attachment 334067
Baseline 3 is close to a 100% increase on average ?

Spoiler: Skinmesh ini settings

No idea if this is optimal for my CPU, my knowledge of the concepts is low and information online is confusing AF
[threads]
computeColliders=6
skinmeshPart=4
applyMorphs=2
skinmeshPartMaxPerChar=2
applyMorphMaxPerChar=4
affinity=1,3,5,7,9,11

[threadsVR]
computeColliders=4
skinmeshPart=1
applyMorphs=2
skinmeshPartMaxPerChar=2
applyMorphMaxPerChar=4
affinity=1,3,5,7,9,11

[profiler]
enabled=0

follow the updated example configs, just set all variables to your core count (6)

[threads]
computeColliders=6
skinmeshPart=6
applyMorphs=6
skinmeshPartMaxPerChar=6
applyMorphMaxPerChar=6
affinity=1,3,5,7,9,11

that should give you more FPS in all benchmark scenes and slightly lower in baseline3

 

[Jimmichoochoo]

I get pretty much ZERO difference with the patch.

Before patch:

After patch:

My Skinmeshpartdll.ini

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

[threadsVR]
computeColliders=4
skinmeshPart=1
affinity=1,3,5,7,9,11,13,15

[profiler]
enabled=0

 

[fear_the_potato]

Did a clean install of VaM. Performance was worse with performance patch 12 applied. Tried with profiler set to 0 and then 1, same results.

AMD Ryzen 7 5800X + RTX3080
[threads]
computeColliders=8
skinmeshPart=8
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

Before patch:

After patch:

 

[turtlebackgoofy]

Did a clean install of VaM. Performance was worse with performance patch 12 applied. Tried with profiler set to 0 and then 1, same results.

AMD Ryzen 7 5800X + RTX3080
[threads]
computeColliders=8
skinmeshPart=8
applyMorphs=8
skinmeshPartMaxPerChar=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15

Before patch:
View attachment 334331
After patch:
View attachment 334333

Your frame TotalTimes improved across the board but not your FPS, do you use some plugins like discord, nvidia overlay or steam overlay?

 

[turtlebackgoofy]

I get pretty much ZERO difference with the patch.

Before patch:

View attachment 334074

After patch:

View attachment 334075

My Skinmeshpartdll.ini

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

[threadsVR]
computeColliders=4
skinmeshPart=1
affinity=1,3,5,7,9,11,13,15

[profiler]
enabled=0

Are you sure you installed it correctly? There is absolutely no difference in any timings.

 

[fear_the_potato]

Your frame TotalTimes improved across the board but not your FPS, do you use some plugins like discord, nvidia overlay or steam overlay?

No - none of those. Although it wouldn't surprise me if there's something in the background causing a problem somewhere. Someone with the same processor as me posted earlier that they had no improvement until they installed an AVX1 config, but I didn't see that anymore to try.

 

[turtlebackgoofy]

No - none of those. Although it wouldn't surprise me if there's something in the background causing a problem somewhere. Someone with the same processor as me posted earlier that they had no improvement until they installed an AVX1 config, but I didn't see that anymore to try.

It should have nothing to do with AVX, the CPU and GPU finished the frames faster than before and yet the FPS counter is not displaying it correctly. Do you see more FPS in scenes other than the benchmark?

 

[fear_the_potato]

It should have nothing to do with AVX, the CPU and GPU finished the frames faster than before and yet the FPS counter is not displaying it correctly. Do you see more FPS in scenes other than the benchmark?

These benchmarks actually look really good...but I'm also did a fresh install to test. I think my main VaM installation gets much poorer results. It's a bit of a mess. Couldn't even run the benchmark on it because it got stuck trying to load the Janie character. I should work on starting clean and porting over just what I use.

 

[trety]

Full benchmark, using v12 patch

 

[turtlebackgoofy]

Full benchmark, using v12 patch
View attachment 334820

jesus christ. BTW the render limit is currently 300fps, I could uncap it so you get even higher FPS Baseline2 would be closer to 442fps afterwards

 

[starmoon]

Base

Patch 12

I have tried versions 10, 11, and 12 results are always the same, small change in physics no real change in fps. The 3060 ti is just at its limit. Overclocking only gives another 2-3 fps.
Ryzen 5 5600x3d 3060ti 32 GB 3200MHz
[threads]
computeColliders=6
skinmeshPart=6
applyMorphs=6
skinmeshPartMaxPerChar=6
applyMorphMaxPerChar=6
affinity=1,3,5,7,9,11

 

[Jimmichoochoo]

Base
View attachment 334861Patch 12
View attachment 334862

I have tried versions 10, 11, and 12 results are always the same, small change in physics no real change in fps. The 3060 ti is just at its limit. Overclocking only gives another 2-3 fps.
Ryzen 5 5600x3d 3060ti 32 GB 3200MHz
[threads]
computeColliders=6
skinmeshPart=6
applyMorphs=6
skinmeshPartMaxPerChar=6
applyMorphMaxPerChar=6
affinity=1,3,5,7,9,11


View attachment 334868

Yeah I think this is the same issue I'm having, as I had no change either after the patch.

 

[fear_the_potato]

Your frame TotalTimes improved across the board but not your FPS, do you use some plugins like discord, nvidia overlay or steam overlay?

Tried again - rebooted the pc. Killed as many background tasks as possible. Clean install of VaM. Reran the tests with and without patch 12. Patch 12 gives me a drop in performance.

 

[Stars&Garters]

Full benchmark, using v12 patch
View attachment 334820

Your results absolutely crush mine on a clean, v12-patched install despite having the same CPU-GPU combo, but I did note you appeared to have a higher spec MOBO and 64 GB RAM (despite older z690 config, mine is an ASUS PRIME Z790-A with 32GB DDR5 RAM)?

Anyways I could clearly use your help oh wise one ?. Behold my comparative shame:

 

[veijokyy]

Your results absolutely crush mine on a clean, v12-patched install despite having the same CPU-GPU combo, but I did note you appeared to have a higher spec MOBO and 64 GB RAM (despite older z690 config, mine is an ASUS PRIME Z790-A with 32GB DDR5 RAM)?

Anyways I could clearly use your help oh wise one ?. Behold my comparative shame:

View attachment 334967

You are running the benchmark with 4k resolution and the other one was done in 1080p.

 

[Stars&Garters]

You are running the benchmark with 4k resolution and the other one was done in 1080p.

...sometimes it's the embarassingly simple things, although even turtlebackgoofy didn't seem to notice that. Damn, I had even wondered it to myself but then forgot about comparing resolutions before opening my digital mouth >_>;

Still gonna' send them the threadprofile data tho, but now I'm curious about setting 1080p res and coming closer to those numbers.

1080p BENCHMARK EDIT:

trety's results are still 20 FPS better than mine on average, except for Baseline 3 where he actually did crush me. Maybe now it actually has something to do with the MOBO/RAM difference, but idk maybe not.

 

[fear_the_potato]

Tried again - rebooted the pc. Killed as many background tasks as possible. Clean install of VaM. Reran the tests with and without patch 12. Patch 12 gives me a drop in performance.

For what it's worth, I decided to do a very simple test. Just loaded up the default scene. Went in to user preferences to see the FPS counter. Moved the camera around violently and watched the FPS. Without patch 12 = ~210fps. With patch 12 = ~310fps.

 

[trety]

...sometimes it's the embarassingly simple things, although even turtlebackgoofy didn't seem to notice that. Damn, I had even wondered it to myself but then forgot about comparing resolutions before opening my digital mouth >_>;

Still gonna' send them the threadprofile data tho, but now I'm curious about setting 1080p res and coming closer to those numbers.

1080p BENCHMARK EDIT:
View attachment 334972

trety's results are still 20 FPS better than mine on average, except for Baseline 3 where he actually did crush me. Maybe now it actually has something to do with the MOBO/RAM difference, but idk maybe not.

What settings did you used with patch? 'Recommended' in the resource page, or mine from here?

Performance Patch (up to 30% more physics speed)

On 5800x3d [threads] computeColliders=6 skinmeshPart=1 affinity=1,3,5,7,9,11,13,15 Before patch After Patch FPS is about the same but physics time is a bit lower.

hub.virtamate.com

I'm running both watercooled CPU and GPU, and i have changed intel's mounting bracket to thermal grizzly one :

Even if my CPU wasn't really thermal throating, hitting at most 80ish degrees in simple physics test before, it's now at max 65 at this test, actually giving a few fps more too.

My RAM is 6400Mhz, but i'm using it at 6000Mhz XMP to reduce VaM crashes. VaM tends to crash very easily with XMP, even if any other games, or RAM benchmarks are just fine.

CPU is at stock settings with just disabled Hyper-threading in bios.

 

[Stars&Garters]

What settings did you used with patch? 'Recommended' in the resource page, or mine from here?

Performance Patch (up to 30% more physics speed)

On 5800x3d [threads] computeColliders=6 skinmeshPart=1 affinity=1,3,5,7,9,11,13,15 Before patch After Patch FPS is about the same but physics time is a bit lower.

hub.virtamate.com

I'm running both watercooled CPU and GPU, and i have changed intel's mounting bracket to thermal grizzly one :

Even if my CPU wasn't really thermal throating, hitting at most 80ish degrees in simple physics test before, it's now at max 65 at this test, actually giving a few fps more too.

My RAM is 6400Mhz, but i'm using it at 6000Mhz XMP to reduce VaM crashes. VaM tends to crash very easily with XMP, even if any other games, or RAM benchmarks are just fine.

CPU is at stock settings with just disabled Hyper-threading in bios.

The recommended (all 8s, but with HT enabled and 1,3,5,7,9,11,13,15 affinities), outside of VR at least where I ended up throwing progressively larger numbers (up to 8s) without really knowing what I was doing.

I've tried HT disabled/12345678 affinity a couple of times, without your exact settings admittedly, but it didn't do much. I only have a 3 fan 420mm AIO for CPU cooling and also the Liquid Suprim 4090 but both are obviously not proper custom liquid cooling, and my RAM seems just a tad slower than yours (although I'm also running 6000MHz on an XMP Tweaked profile).