Benchmark Result Discussion

[gokusanmar]

Since I am using a laptop, I am interested in checking out the difference between running VAM on a laptop screen vs monitor ( supposedly games run ~10% better when you have a monitor plugged if your laptop does not have MUX switch) and then 1080p vs 1440p.

 

[BooMoon]

here's the result of my simple settings ? , the resolution is custom my native res is 1366x768 .

does anyone has an rtx3060 to have an ideia how it runs VAM in 1080p ? i was thinking on get one but idk if will be a good upgrade .

 

[Jiraiya]

I have a 3060 and it's not bad, but honestly, look at your CPU first!!!! Your Physics time on baseline3 was 321 !!!! Compared to 29.95 for simpler physics.
I had a 3600K and it wasn't great, I upgraded to a 3060+5600X and get passable results.
My physics time is 21ish dropping to 10. Compared to your 321.
If you look at your render time you will see it's reasonable for most scenes and it's the physics time taking up the majority of your framerate. I would upgrade the CPU before looking at a new GPU, especially with prices now.

 

[MacGruber]

the resolution is custom my native res is 1366x768 .

Note that your native resolution seems to be 1920x1080. The Benchmark shows 1920x1051 probably because there is the window bar at the top and/or the Windows taskbar at the botttom.

 

[HolySchmidt]

2560x1440 Desktop


@gokusanmar
Are you referencing to this video?

Interesting, wouldn't have thought this is even possible as it's the same hardware.

@Jiraiya
Good comment, never looked at those numbers. Could you make a benchmark with 1920x1080? So then we would have an idea of how the 3060 performance is in this resolution, would be perfect for BooMoon.

@BooMoon
Simply install MSI Afterburner + RivaStatisticsTuner (one package, freeware) and then you'll see where your bottleneck is in what situation. Tutorials on this can be found on youtube.
Here's a comparison of both GPUs: https://gpu.userbenchmark.com/Compare/Nvidia-RTX-3060-vs-Nvidia-GTX-770/4105vs2174

At the end it will be probably both: GPU and CPU. Cause if you get a better GPU your actual CPU will be probably a bottleneck. The lower your resolution is, the more you benefit from a good CPU. The higher the resolution, the more your GPU will come into play.
The problem with VaM is that your CPU is important for all those physics.

 

[Jiraiya]

It's certainly a CPU bottleneck atm. Although that term doesn't describe the real problem. A bottleneck implies the GPU is doing a worse job because the CPU is slowing it down, or vice/versa. That isn't really the case here.
What you need to do is work out how much of your per frame draw time is taken with each step and where money is best spent to improve that.
For BooMoon his GPU render time is less than a tenth of the Phsyics/Script time which is CPU. A better GPU would take a tiny % off that number where a better CPU would slash that figure dramatically.
Think of it as a journey to work.
If you take 3 hours walking to a train station then spend 5 minutes going fast on the train, what part of your journey needs improvement? The slowest part. Without a doubt a CPU upgrade would get him a FAR better performance boost than GPU upgrade. Considering the prices of GPUs right now too, it would be silly to waste money on a 3060 that won't get him much when he can spend a lot less on a new CPU that will really improve his overall performance.

Here are my desktop stats. What I am noticing is interesting is the total ms per frame Vs the FPS. My baseline1 is ~4.99ms per frame with a 0.14ms wait time (averages). This gives 499ish ms for the 96.78fps I get. What the hell is my PC doing for the other half a second?
I am sure there are 1000ms in each second, am I being insane with my math or is there something I am missing here? Where is the missing time going?

 

[Voithe]

Sorry... this time i'll pass. I don't want my poor (but still GLORIOUS) GTX 1080 to melt down...

(or... maybe i can at least try it in desktop mode... )

 

[taichi]

Baseline 3 benchmark result is quite highly dependent on how bloated your VAM install is. I get avg 6.03ms PhysicsTime with clean VAM install with nothing extra just the benchmark and its dependencies. But on my main VAM install (200GB+ of crap) I get avg 10.57ms PhysicsTime.

Clean VAM:

Bloated VAM:

 

[keycode]

my new benchmark 3 after upgrading to windows 11. The second image is with benchmark 2 (windows 11) while the image 3 is my old result (with windows 10).
Setting vam in windowed mode was giving some fps plus, of course... but I can the same consider the fact that EVEN with windows 11 (EDIT: and the NEW hot complexity of baseline 3 scene!!! ) vam works still good.

 

[keycode]

and yes... upgrading to windows 11 without a clean install (keeping my old files was a strict priority) was giving this (hope temporary) issue: even reinstalling nvidia new drivers, hotfix etc. there is no way for getting detected the windows 11 upgrade in my nvidia control panel.

 

[MacGruber]

@keycode Keep in mind that version 2 and 3 are not producing comparable results. There are more props in the scenes now, light changed and of course two entirely new scenes where added that contribute to the average result.

 

[MacGruber]

Baseline 3 benchmark result is quite highly dependent on how bloated your VAM install is. I get avg 6.03ms PhysicsTime with clean VAM install with nothing extra just the benchmark and its dependencies. But on my main VAM install (200GB+ of crap) I get avg 10.57ms PhysicsTime.

That is curious. Can you please try to reproduce the results (= run again)? Might heat be a problem? Like the second one was slower because the machine was already warm? (The warm-up scene is supposed to take care of that, but it only helps so much)

 

[taichi]

A lot of it is preloading morphs from Vars it seems. I disabled preloading from like 150 addonpackages (I clicked through all of them might have missed some) and got quite a bit of performance back. But still missing a lot. Think I only had like 20 morphs in custom folder still. Don't think I'll gain back the rest deleting those. There is still some other overhead created from having tons of content not just preloaded morphs.

And I am 99.9% it is not heat. Even before your benchmark I knew about this problem comparing FPS in same scene in different VAM installations. But the scenes were static for easy FPS comparison, and not very physics heavy. So it was like 150 fps vs 130 fps when comparing clean VAM and ton of content VAM install. But your Baseline 3 benchmark I lose almost half of my performance having a bloated VAM.

Wish I could run Baseline 3 only as it is the only one I care about really. Would make troubleshooting this issue a lot faster if I could test faster if something I did helped

Ran on clean VAM install one more time to double check it was not a fluke. Still getting way better Baseline 3 performance here:

 

[RandomVAMUser]

V3 results (driver v21.10.2_minimal)

Spoiler: side experiment = failing to run custom benchmark (with my usual settings)

Yesterday after official pass decided to ran custom one with my usual settings and frame limiter (RTSS standalone, no afterburner).
For some reason after initial warm-up benchmark fps counter was like "Peace out!".That run was useless and totalTime was obviously inaccurate.
I did restart in between tests, just in case...

Today i ran official pass again (with limiter) as expected totalTime is up across the board.
That didn't tell me much of anything (it's same run, just limited).

right now i did 2 custom runs back2back (no restart)
first with update cap 2 (in hope to invite fps counter back) second one with cap 1 (to see if yesterdays run was error or something).

fps counter has left the chat...

 

[fbradcdsc]

In case you wanted to laugh at my setup

 

[honda78]

There seems to be something horribly wrong with physics on my computer: "Baseline 3" with recommended settings runs at about 3 FPS ... lowering "physics update cap" to 1 has a more expected result of ~60 FPS.
There shouldn't be something particularly bad with my rig - the most uncommon thing are 2x2080Ti in SLI mode which shouldn't interfere as VaM Physics is running on the CPU anyway (BTW: Did a test with Firestrike Extreme which returns 15000 points as a result).


I made some rudimentary tests myself by visualizing a model of a crystal lattice with sphere atoms: a grid of 10x10x10 atoms works well, 15x15x15 takes a very long time to load and 20x20x20 atoms never finish. Additionally - if i enable physics, the scene loads with a weird "wall"/lineup of the spheres in the scene: you can test below with the scene files attached if you are curious.

...looking forward to see VaM 2.x in action. I hope, the rework of the internals and physics engine will pay off.

 

[BooMoon]

It's certainly a CPU bottleneck atm. Although that term doesn't describe the real problem. A bottleneck implies the GPU is doing a worse job because the CPU is slowing it down, or vice/versa. That isn't really the case here.
What you need to do is work out how much of your per frame draw time is taken with each step and where money is best spent to improve that.
For BooMoon his GPU render time is less than a tenth of the Phsyics/Script time which is CPU. A better GPU would take a tiny % off that number where a better CPU would slash that figure dramatically.
Think of it as a journey to work.
If you take 3 hours walking to a train station then spend 5 minutes going fast on the train, what part of your journey needs improvement? The slowest part. Without a doubt a CPU upgrade would get him a FAR better performance boost than GPU upgrade. Considering the prices of GPUs right now too, it would be silly to waste money on a 3060 that won't get him much when he can spend a lot less on a new CPU that will really improve his overall performance.

Here are my desktop stats. What I am noticing is interesting is the total ms per frame Vs the FPS. My baseline1 is ~4.99ms per frame with a 0.14ms wait time (averages). This gives 499ish ms for the 96.78fps I get. What the hell is my PC doing for the other half a second?
I am sure there are 1000ms in each second, am I being insane with my math or is there something I am missing here? Where is the missing time going?

thats really interesting thanks for the info! , into a cost x benefit the cpu upgrade is way lower than a new gpu , i think it will be next upgrade then .

 

[keycode]

V3 results (driver v21.10.2_minimal)

Spoiler: side experiment = failing to run custom benchmark (with my usual settings)

Yesterday after official pass decided to ran custom one with my usual settings and frame limiter (RTSS standalone, no afterburner).
For some reason after initial warm-up benchmark fps counter was like "Peace out!".That run was useless and totalTime was obviously inaccurate.
I did restart in between tests, just in case...

Today i ran official pass again (with limiter) as expected totalTime is up across the board.
That didn't tell me much of anything (it's same run, just limited).

right now i did 2 custom runs back2back (no restart)
first with update cap 2 (in hope to invite fps counter back) second one with cap 1 (to see if yesterdays run was error or something).


fps counter has left the chat...

maybe it's banal to ask: did you ever consider testing Vam without SLI "just" with one hotty 2080ti? (besides the still very good 2080ti... also that AMD cpu positively looks like a beast to my noob opinion, even considering that vam 1 physics is not optimized for multi-core).

 

[taichi]

There seems to be something horribly wrong with physics on my computer: "Baseline 3" with recommended settings runs at about 3 FPS ...

Have you tried with a clean VAM install? My guess is you might have like thousands of morphs in Custom\Atom\Person\Morphs . Those really tank physics performance for me at least.

 

[danood]

Got curious after I posted the VR one and saw everyones desktop benchmarks. Actually confused how non-vr is worse lol. Interesting......

 

[Jiraiya]

Your physics time went up, which is CPU so you were doing something else on the desktop to slow the machine down?
Also, in the VR test your rendertime is 0.00 in all the boxes.
Looking at the claimed FPS, I do not believe that was a correct and accurate test. Something went wrong in my opinion and that 159fps is a false result.

 

[honda78]

Have you tried with a clean VAM install? My guess is you might have like thousands of morphs in Custom\Atom\Person\Morphs . Those really tank physics performance for me at least.

Thanks taichi, the test above was already done on a fresh download of VaM.
I found some never AMD chipset drivers for Windows 11, reverted from SLI to single GPU, used fixed CPU Clock of 4Ghz, moved the directory from SSD to NVMe, tried fullsceen and with a Rifts S, still same results.
It looks like i am the only one who tried the benchmark with a Threadripper, so maybe i'm simply out of luck with my specific hardware.

 

[kokmaker]

It seems like with my setup, the results are fairly par for the course? I'm looking to see if anyone has specifically upgraded just the cpu or gpu with similar specs as me. (It seems like most in this thread are in the budget to midrange boat or have an absolute beast of a pc, lol.) I'm not entirely sure which would benefit me more, as it seems like it could be a little of both. Though I would prefer to get a new graphics card since VAM is the only software I use that would see the most use out of better cpu it seems. Planning to do a vr test when I'm not a hole of infinite laziness. I also suspect all the files I've hoarded until now might be taking a toll, and it's time for a fresh install. Unfortunately, I just like to hoard things.

 

[Jiraiya]

Looking at your spec, I would say don't upgrade at the moment. Graphics cards are stupid hard to get and expensive and that CPU isn't that bad.
I personally would wait, there will be 40xx Nvidia cards announced in not long and new CPUs are always coming out. Unless you are desperate I would hold off.

 

[kokmaker]

Looking at your spec, I would say don't upgrade at the moment. Graphics cards are stupid hard to get and expensive and that CPU isn't that bad.
I personally would wait, there will be 40xx Nvidia cards announced in not long and new CPUs are always coming out. Unless you are desperate I would hold off.

Yeah, just a tad impatient since it looks like 2.x is still way further than I thought, (take their time ofc). I'm just really bad at delaying gratification. But it seems like I'm running into more physics issues as of late, like colliders triggers not worker after a while for no apparent reason.