Benchmark Result Discussion

[GenericUsername1]

VaM continues to urge me to upgrade my system, so here I am with a new 5080 benchmark!

Previous benchmarks:
- Intel i7-9700k + Nvidia 3080Ti
- AMD 7700X + Nvidia 3080Ti
- AMD 7800X3D + Nvidia 3080Ti
- AMD 7800X3D + Nvidia 5080 <- you are here

I am very happy with the results and feel like this configuration will serve me well until VaM 2.0 gets released. Or until I develop a kink for Chewbacca, curse those expensive hair physics..

 

[ngagnu]

I was looking for an upgrade from my 3080 FE 10gb. Confused between 9070xt or 5070ti. Can someone tell which one is better in VAM.

 

[Seraphim]

I was looking for an upgrade from my 3080 FE 10gb. Confused between 9070xt or 5070ti. Can someone tell which one is better in VAM.

5070ti in VR

 

[Skaterr33]

Here are my values with my new PC.

The RAM runs with DDR5-6000, CL30-40-40-76

The processor has “PBO Curve Optimizer -20mv” in the mainboard settings.
This means it gets a little less power, but only gets a maximum of 80°C under full load.

I hope the values achieved here are okay for the hardware...

 

[henshin]

This is a resume of the previous 13 pages for VR:


CPU: AMD Ryzen 7 7800X3D
GPU: RTX 5090
RAM: 64GB
FPS: 102-300 (VirtualDesktop Only)
FPS: 70-130 (SteamVR)
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 5090
RAM: 64GB
FPS: 77-160
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 4090
RAM: 32GB
FPS: 76-150
----------------

CPU: AMD Ryzen 7 7800X3D
GPU: RTX 4090
RAM: 32GB
FPS: 72-161
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 5090
RAM: 64GB
FPS: 68-87
----------------

CPU: Intel i5-13600KF
GPU: RTX 5090
RAM: 64GB
FPS: 67-165
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 5080
RAM: 64GB
FPS: 58-137
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 5080
RAM: 32GB
FPS: 60-100
----------------

CPU: AMD Ryzen 7 7800X3D
GPU: RTX 4090
RAM: 64GB
FPS: 49-150
FPS: 56-136 (patched)
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 5090
RAM: 64GB
FPS: 45-98
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 4090
RAM: 64GB
FPS: 45-56
----------------

CPU: AMD Ryzen 9 7950X3D
GPU: RTX 4070ti
RAM: 64GB
FPS: 44-134
----------------

CPU: AMD Ryzen 9 7950X3D
GPU: RTX 4070ti
RAM: 32GB
FPS: 43-113
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RX 9070 XT
RAM: 64GB
FPS: 40
----------------

CPU: Intel i5-13600KF
GPU: RTX 4090
RAM: 64GB
FPS: 33-77
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 3090
RAM: 64GB
FPS: 30
----------------

CPU: Intel i5-9400F
GPU: RTX 2070
RAM: 48GB
FPS: 20-41
----------------

CPU: AMD Ryzen 7 9800X3D
GPU: RTX 3080
RAM: 64GB
FPS: 17-60
----------------

CPU: Intel i7-6700K
GPU: GTX 1660 SUPER
RAM: 16GB
FPS: 14-34
----------------


CONCLUSIONS (?)
AMD is better than Intel? and Ryzen 7 7800X3D is very similar to Ryzen 7 9800X3D
What about Ryzen 7 vs 9? ( X3D version is important I think)
I couldn't find too many tests with Intel in VR

5090 vs 4090 is +50%? (for some tests the boost is not so high, why?)
5080 similar to 4090 (4090 +10%?)

CPU patch is getting +10/15%?

VirtualDesktop Only (without steamVR) makes a big difference?

RAM 32vs64 no big difference


Is there any relation between Desktop and VR FPS? because there are no so many VR tests, maybe we can apply some factor to the Desktop test? is it 0.5*Desktop?

@MacGruber said: the Benchmark is not really suited anymore for current high-end PCs? why is that?
Are the scenes from VAM now much more demanding? what type of scene could be that (maybe there is some example in the hub to test that?)

 

[nnackler777]

CPU OC'd to ~ 5.2 mhz, RAM @ 5600 mhz, GPU stock (straight out of the box)... this 5090 Suprim is a monster. Will run again after I turn on nvidia auto-OC in their app, see if it makes much difference.

 

[nnackler777]

Also ran (same settings as last post) but using VaM settings where I typically have them... not too shabby.

 

[starmoon]

5060ti 16 GB not terrible but leaves a lot. The16 GB of vram is nice but load times seem longer this (PNY) is the only 5060ti 16GB actually at MSRP $430.

+40 fps -2ms physics with patch. No additional OC over whatever factory added.
Previous 3060ti

 

[MacGruber]

@MacGruber said: the Benchmark is not really suited anymore for current high-end PCs? why is that?
Are the scenes from VAM now much more demanding? what type of scene could be that (maybe there is some example in the hub to test that?)

At those super high FPS values some people get you just measure noise. Benchmark probably needs a version that runs with 120Hz physics (instead of 60) for an more even CPU load. Also scenes with more characters, since scenes with a single character do not create enough load. Of course the issue is that when you got multiple versions of the benchmark, results are not comparable anymore.

 

[nebka94]

 

[henshin]

These tests are with a new VAM installation, I have only downloaded Benchmark

WITH CPU PATCH (v12evo)

[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
applyMorphs=2
skinmeshPartMaxPerChar=4
applyMorphMaxPerChar=2
affinity=1,3,5,7,9,11,13,15

[profiler]
enabled=0

and in boot.config:
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

I'm not getting any boost with CPU patch, is that normal? am I missing something?

 

[EinTyp]

My PhysicsTime suck dick.

 

[Seraphim]

These tests are with a new VAM installation, I have only downloaded Benchmark

View attachment 522786

WITH CPU PATCH (v12evo)

[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
applyMorphs=2
skinmeshPartMaxPerChar=4
applyMorphMaxPerChar=2
affinity=1,3,5,7,9,11,13,15

[profiler]
enabled=0

and in boot.config:
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


View attachment 522791

I'm not getting any boost with CPU patch, is that normal? am I missing something?

23% boost and 14% faster cpu time. What do you mean?

 

[henshin]

23% boost and 14% faster cpu time. What do you mean?

I'm just seeing FPS avg, in both tests is about 270. Are the min and max the important ones?

 

[Seraphim]

I'm just seeing FPS avg, in both tests is about 270. Are the min and max the important ones?

Yeah minimum is the important one, and Baseline 3 for cpu.

 

[h00l1gan]

+CPU Patch
3060 Ti 1900 ghz

5070 3000 ghz

 

[redeyes]

I've recently upgrade to a 9950x3d from a 9700k which is quite the jump
This is with CPU patch + bare minimum asset folder to run benchmark
And as the gcard was the same for both and an overkill card at 1920x1080 it really shows how much the CPU is holding everything back

Old 9700k to compare

 

[lojepo4410crtsec]

Got a 5090 so here are my results. I moved from a 4080 super previously which was good, but left some to be desired and wasn't allowing me to run Voxta well with Vam. I've been able to max out everything, but FPS in scenes so far hasn't been mind blowing (needs more testing). Curious what others are seeing as my 5090 barely runs at 40% most of the time.

I tried a few things for each of them I did the OSW disable, turned off in Steam, and turned off VD ASW. confirmed i was seeing 130-150fps in a scene.

ALERT: Just an FYI as I was an idiot and you might be too. I ran 3dMark and got a depressing score originally. Looked online and someone mentioned doing a clean install of drivers even though I had latest. Figured wth and saw a 25% increase in score and was over average without any overclocking or anything. So definitely do this. was in the high 11k range and went up to almost 15k. 3dmark demo.

First, VirtualDesktop ran directly in VD no steam. not sure why it says rift, but on Quest 3.
View attachment 477816

Then I tried my oculus script -vrmode oculus which launched steam.
View attachment 477820

Then I ran the default -vrmode openvr which ran steam.
View attachment 477821

Also ran a desktop at 1440p.
View attachment 477823

Does it 40% the GPU in high res VR too? 7800X3D as the bottleneck would be interesting

 

[redeyes]

Does it 40% the GPU in high res VR too? 7800X3D as the bottleneck would be interesting

You see my result above on a 4090 - it wasn't going at 100% at 1920x1080 - though in fact neither was the CPU - not even on any core was it maxed out - I suspect the limitation at these fps speeds is infact memory transfer speeds from g-card to cpu

 

[WristRock3t]

Just putting this here for reference.

CPU: AMD Ryzen 5 5500
GPU: RTX 3060
RAM: 64GB

WITH CPU PATCH (v12evo)

SkinMeshPartDLL.ini

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

[threadsVR]
computeColliders=6
skinmeshPart=6
skinmeshPartMaxPerChar=6
applyMorphs=6
applyMorphMaxPerChar=6
affinity=1,3,5,7,9,11
[profiler]
enabled=0

boot.config

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

 

[redeyes]

9950x3d - with CPU patch pinning cores makes core parking pointless

Just for those interested, if you use the CPU patch and pin VAM to the 3d vcache cores correctly this basically has the same effect as core parking. But what's interesting is if you're running other programs, like watching a video in another monitor, running screen grabbing software or my case running htc wireless adapter that consume some amount of CPU processing - with parking it actually makes it worst as now those processes will get pushed onto the same chip as VAM is running on consuming clocks cycles and slowing it down.

Basically the 9950x3d is fantastic and with no compromise if you using pinning. The parking idea is does increase clock speeds by a small amount, but if you're running other processes that small gain gets consumed by having all processes running on the same chip.

So for me, I think pinning VAM / a game to the 3d vcache chip is better than parking.

Without core parking

With Parking