• Hi Guest!

    Please be aware that we have released another critical security patch for VaM. We strongly recommend updating to version 1.22.0.12 using the VaM_Updater found in your installation folder.

    Details about the security patch can be found here.

Benchmark Result Discussion

I'd like to thank @HolySchmidt for his comments about turning off hyperthreading. For the life of me I couldn't figure out what was causing micro stutters in physics heavy scenes and the CPU patch didn't seem to show any marked improvement for me on fps. Until turning off HT.

This wasn't exactly a fair test as the original was done with a fresh VaM install and the latest has about 100gb in the addonpackages folder. I honestly just wanted to be rid of the micro stutters, so I didn't expect a decent bump in performance as well.

Cheers HolySchmidt
Benchmark-20250127-003719.png
Benchmark-20250225-040423.png
 

Attachments

  • Benchmark-20250127-003719.png
    Benchmark-20250127-003719.png
    851.2 KB · Views: 0
Follow up to my Results: I have a 5090 with the 168 ROP problem, so it is slower than it should be. Trying to get it swapped at the moment.
 
Anyone who'd like to comment on the (VaM-)merits of the 7900XTX relative to Nvidia 50xx? Headset will likely be a Quest3S, Desktop is WQHD.

* Would 24Gb Vram make a difference in VaM, compared to, say 20Gb (or 16)
* I could get an action price of ~900€ right now - pretty attractive relative to the current "moon-prices" of Nvidia's 50xx cards.
* For VaM, only Rasterization performance seems important - and the XTX looks pretty competitive in that regard?

Thoughts?

(I'm budget-limited to 1.900€ - 2.000€ for the whole system upgrade. Reusing my current nvme and PSU, I could just about manage a Zen 7 9800X3D +7900XTX combo )
 
Anyone who'd like to comment on the (VaM-)merits of the 7900XTX relative to Nvidia 50xx? Headset will likely be a Quest3S, Desktop is WQHD.

* Would 24Gb Vram make a difference in VaM, compared to, say 20Gb (or 16)
* I could get an action price of ~900€ right now - pretty attractive relative to the current "moon-prices" of Nvidia's 50xx cards.
* For VaM, only Rasterization performance seems important - and the XTX looks pretty competitive in that regard?

Thoughts?

(I'm budget-limited to 1.900€ - 2.000€ for the whole system upgrade. Reusing my current nvme and PSU, I could just about manage a Zen 7 9800X3D +7900XTX combo )
Its a bit complicated question, but my short recommendation would be go for it.
Longer explanation:
pros:
- hell of a lot performance
- reliable trustworthy pcie 8pin connectors (its a joke that in 2025 it is a pro, but it is..)
- 24gb is plenty, yor dont have to afraid throwing anything on your scene, plus holds up will even with memory leak..
- price!
- with some finicking usually many AI stuff can be handled by them already

cons:
- vam benchmarks seems to show a 15-20% more performance on similar grade nv cards
- it seems amd favors udna architecture so rdna cards will likely get less and less of the new features..
- less efficient power handling
- AI stuff is less supported on amd cards
(- worse VR support then NV)

I have a 4090 now, but I used a 6800XT before. In my experience you can set your scene to your visual performance and the limiting factor will most likely be your cpu (I use a 5800x3d). The rdna2 was fine between fhd and qhd pixel count. For VR I am using a Q3 via virtual desktop. AMD handled that just fine. With the same settings in VR the 4090 is cpu limited. I just needed to adjust my settings and upped the resolution a bit (vam has even a built in slider for it).
The 4090 overall stomps the 6800XT, but eats significantly more also. It think the 7900XTX would be somewhere in the middle. You will get great performance at a reasonable price.
 
Its a bit complicated question, but my short recommendation would be go for it.
Longer explanation:
pros:
- hell of a lot performance
- reliable trustworthy pcie 8pin connectors (its a joke that in 2025 it is a pro, but it is..)
- 24gb is plenty, yor dont have to afraid throwing anything on your scene, plus holds up will even with memory leak..
- price!
- with some finicking usually many AI stuff can be handled by them already

cons:
- vam benchmarks seems to show a 15-20% more performance on similar grade nv cards
- it seems amd favors udna architecture so rdna cards will likely get less and less of the new features..
- less efficient power handling
- AI stuff is less supported on amd cards
(- worse VR support then NV)

I have a 4090 now, but I used a 6800XT before. In my experience you can set your scene to your visual performance and the limiting factor will most likely be your cpu (I use a 5800x3d). The rdna2 was fine between fhd and qhd pixel count. For VR I am using a Q3 via virtual desktop. AMD handled that just fine. With the same settings in VR the 4090 is cpu limited. I just needed to adjust my settings and upped the resolution a bit (vam has even a built in slider for it).
The 4090 overall stomps the 6800XT, but eats significantly more also. It think the 7900XTX would be somewhere in the middle. You will get great performance at a reasonable price.

Thanks for the super-detailed overview. You've somewhat alleviated my biggest bellyache: "worse VR support than Nvidia".

I'm currently running a Quest 1 with a Radeon 5700XT and the Q1 and the Navi1 combo locked me out of nearly all performance-enhancing (VR) technologies that made newer card/goggle combos even more performant - so I want to make sure I'm not making the same mistake again

Anything you think is important to know wrt FSR3.1 vs DLSS 3/4 in VR?
 
Last edited:
Thanks for the super-detailed overview. You've somewhat alleviated my biggest bellyache: "worse VR support than Nvidia".

I'm currently running a Quest 1 with a Radeon 5700XT and the Q1 and the Navi1 combo locked me out of nearly all performance-enhancing (VR) technologies that made newer card/goggle combos even more performant - so I want to make sure I'm not making the same mistake again

Anything you think is important to know wrt FSR3.1 vs DLSS 3/4 in VR?
Vrperfkt has forks that work with radeon! At least I used with rdna2. I dont know regarding rdna1.
On the other hand since you asked here I assumed you ask only vam relevant comparison. Regarding other games it depends on titles. You can check in reviews how that plays on given cards, but the 7900 xtx will still be the king of amd lineup after rdna4 comes out.
DLSS is the best upscaler. period. But you cant use that in vam and fsr is also usable. Multiframe generating is also an option on amd, more so you can use Lossless Scaling from steam in any title. Its not a native app so you need to fiddle it, but if you interested a great option.
To be honest I am still not convinced that I need the 4090. I was curious and I like it, also managed to get on a good price, but its way more powerful than the necessary performance level in most of my cases.

In vam for example with amd cards you can set tessellation in driver wich can translate to an extra 20-40% performance. On nvidia I did not looked yet, dont know if I can or would mean anything. I assume may be it handles better and probably this is the benched performance edge for them.
 
I could get an action price of ~900€ right now - pretty attractive relative to the current "moon-prices" of Nvidia's 50xx cards.

The 9070XT is coming out in like a week, for 600$US and its faster in raster and raytracing... but it has 16gb. FSR4 is going to be as good as DLSS4
 
I could get an action price of ~900€ right now - pretty attractive relative to the current "moon-prices" of Nvidia's 50xx cards.

The 9070XT is coming out in like a week, for 600$US and its faster in raster and raytracing... but it has 16gb. FSR4 is going to be as good as DLSS4

Yeah, eagerly waiting for the tests myself. As regards performance, it looks like it might equal the Nv 5070Ti in rasterization and raytracing performance - and if that's the case, I'll try to get one - but I doubt it'll beat the 7900XTX in rasterization performance. The XTX rasterization perf is between the 4080 TiS and the 5080 - it maybe be old, but it's still a fsking beast.
 
The XTX rasterization perf is between the 4080 TiS and the 5080 - it maybe be old, but it's still a fsking beast.
oO i heard it got better after drivers\stuff updates, tho dunno how about VaM...
Just after XTX release there was someone in this thread comparing it with basic (no ti\super whatever) 4080. And in VaM 4080 was better.
 
Vrperfkt has forks that work with radeon! At least I used with rdna2. I dont know regarding rdna1.
Yeah, no joy for the rdna1 - I looked for quite a while.
On the other hand since you asked here I assumed you ask only vam relevant comparison.
Correct.
Regarding other games it depends on titles. You can check in reviews how that plays on given cards, but the 7900 xtx will still be the king of amd lineup after rdna4 comes out.
On pure rasterization perf, it certainly will. As regards my fears of FSR 3.1 locking me out of upscaling tech - probably unknowable.

DLSS is the best upscaler. period. But you cant use that in vam and fsr is also usable.

Oh? Srsly? DLSS upscaling don't work in VaM. That's ... usefull info. Thank you.

Multiframe generating is also an option on amd, more so you can use Lossless Scaling from steam in any title. Its not a native app so you need to fiddle it, but if you interested a great option.
I am interested, and if I get the XTX I'd be interested in any advice you can/want to give.

To be honest I am still not convinced that I need the 4090. I was curious and I like it, also managed to get on a good price, but its way more powerful than the necessary performance level in most of my cases.
Hmmmh - you're CPU-bottlenecked, right?

In vam for example with amd cards you can set tessellation in driver wich can translate to an extra 20-40% performance. On nvidia I did not looked yet, dont know if I can or would mean anything. I assume may be it handles better and probably this is the benched performance edge for them.
oh! Again: I'll look into that if/when I get either the XTX or the 9070XT.

May I get back to you in that case?
 
oO i heard it got better after drivers\stuff updates, tho dunno how about VaM...
Just after XTX release there was someone in this thread comparing it with basic (no ti\super whatever) 4080. And in VaM 4080 was better.
Not sure about VaM performance - I looked at aggregated benchmarks by a German hardware mag (PCGH) of rasterization perf in HD and WQHD - and in those Benches, it comfortably beat the 4080. But I heard others mention that VaM seems to favor Nv cards, so I wouldn't be surprised if the 4080 beat the XTX in VaM.

Then again, that same "Vam Nvidia bonus/Radeon penalty" would likely apply to the 9070XT, right? :cry:

And in that case, the sheer rasterization performance of the XTX might be more important for VaM than any potential FSR4 benefits ...

Anyhow: thanks @mostvanvege, @lolmao500 and @trety for your advice!

Edit: Just saw performance estimates in an aggregated rasterization benchmark (20 games in 4 resolutions - HD/WQHD/UWQGD/UHD) and according to that "benchmark", th 9070XT beats 5071 Ti, 4080 Super and is almost on par with the 7900XTX - so @lolmao500 might be right! If that holds true ... :love:
 
Last edited:
Not sure about VaM performance - I looked at aggregated benchmarks by a German hardware mag (PCGH) of rasterization perf in HD and WQHD - and in those Benches, it comfortably beat the 4080. But I heard others mention that VaM seems to favor Nv cards, so I wouldn't be surprised if the 4080 beat the XTX in VaM.

Then again, that same "Vam Nvidia bonus/Radeon penalty" would likely apply to the 9070XT, right? :cry:

And in that case, the sheer rasterization performance of the XTX might be more important for VaM than any potential FSR4 benefits ...

Anyhow: thanks @mostvanvege, @lolmao500 and @trety for your advice!

Edit: Just saw performance estimates in an aggregated rasterization benchmark (20 games in 4 resolutions - HD/WQHD/UWQGD/UHD) and according to that "benchmark", th 9070XT beats 5071 Ti, 4080 Super and is almost on par with the 7900XTX - so @lolmao500 might be right! If that holds true ... :love:
Seeing the announced rdna4 prices I strongly think - if it is an option - its better to wait for their release. Rumors were that amd will price them agressively and it seems they have made it. So it will probably make a price pressure on the competing cards too.
Regarding performance there are still only guesses no real reviews. I think the xt may land near 4080/5070Ti level and thats good. If you think 16GB is enough you wont give up much performance related to the xtx.
 
Seeing the announced rdna4 prices I strongly think - if it is an option - its better to wait for their release. Rumors were that amd will price them agressively and it seems they have made it. So it will probably make a price pressure on the competing cards too.
Regarding performance there are still only guesses no real reviews. I think the xt may land near 4080/5070Ti level and thats good. If you think 16GB is enough you wont give up much performance related to the xtx.
Well ... is 16Gb enough, though? (genuine question).

Also, I'm a bit confused as to which upscaling tech benefits VaM at all - neither DLSS nor FSR is supported in VaM? (Or maybe I'm missing the point) Afaics, what really seems to help is very good video encoding, a bloody high bandwidth and upscaling the image via the Snapdragon in a Quest 3 (iirc, Virtual Desktop offers that).
 
Seeing the announced rdna4 prices I strongly think - if it is an option - its better to wait for their release. Rumors were that amd will price them agressively and it seems they have made it. So it will probably make a price pressure on the competing cards too.
Regarding performance there are still only guesses no real reviews. I think the xt may land near 4080/5070Ti level and thats good. If you think 16GB is enough you wont give up much performance related to the xtx.

Hmmmpf ... 9070XT is within 4% of the 5070Ti and within 9% of the 7900XTX in rasterizing - the Benchmark I saw even had a massively overclocked "9070XT(X)" that beats the crap out of both the 5070Ti and the 7900XTX. Good OC/UV potential, very power-efficient, "grown-up" AI upsampling, finally "adequate" raytracing performance - and, very important for us VR-fiends: Apparently, it doesn't have rdna 3.1's troubles with hardware encoding.

Superb little card ... but ... scuttlebutt has it that the "aggressive pricing" was likely wishful thinking. Allegedly, the cheapest 9070XT will be offered at 815€+19%VAT = 969€ in Germany. That's a cool 42€ over "the last affordable 7900XTX in Central Europe" I snagged yesterday (the price for the same model jumped by 167€ 222€ overnight - stocks of XTX seem to be running low).

Got one :love: Pricing was ... OK - a ways above MSRP, but also not "Nvidia pricing".
Edit: Fuckme - the card I bought for 869€ an hour ago is now 1.357€ on Amazon! :sick:
 
Last edited:
No CPU patch:

View attachment 467584


CPU patch applied:

View attachment 467552

[threads]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=2
applyMorphs=4
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,
affinity=1,2,3,4,5,6,7,8
engineAffinity=1,2,3

[threadsVR]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=2
applyMorphs=4
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,
affinity=1,2,3,4,5,6,7,8
engineAffinity=1,2,3
[profiler]
enabled=0

Copied this config: https://hub.virtamate.com/threads/c...r-physics-up-to-60-more-fps.49738/post-151411

Any suggestions?
Yeah, Dont copy config for an 7th gen intel i7 lol
 
No CPU patch:

View attachment 467584


CPU patch applied:

View attachment 467552

[threads]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=2
applyMorphs=4
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,
affinity=1,2,3,4,5,6,7,8
engineAffinity=1,2,3

[threadsVR]
computeColliders=4
skinmeshPart=4
skinmeshPartMaxPerChar=2
applyMorphs=4
applyMorphMaxPerChar=6
#affinity=1,2,3,4,5,6,7,8,
affinity=1,2,3,4,5,6,7,8
engineAffinity=1,2,3
[profiler]
enabled=0

Copied this config: https://hub.virtamate.com/threads/c...r-physics-up-to-60-more-fps.49738/post-151411

Any suggestions?
[threads]
computeColliders=8
skinmeshPart=8
skinmeshPartMaxPerChar=8
applyMorphs=8
applyMorphMaxPerChar=8
#affinity=1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16
affinity=1,3,5,7,9,11,13,15

[threadsVR]
computeColliders=8
skinmeshPart=8
skinmeshPartMaxPerChar=8
applyMorphs=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15
[profiler]
enabled=0
 
[threads]
computeColliders=8
skinmeshPart=8
skinmeshPartMaxPerChar=8
applyMorphs=8
applyMorphMaxPerChar=8
#affinity=1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16
affinity=1,3,5,7,9,11,13,15

[threadsVR]
computeColliders=8
skinmeshPart=8
skinmeshPartMaxPerChar=8
applyMorphs=8
applyMorphMaxPerChar=8
affinity=1,3,5,7,9,11,13,15
[profiler]
enabled=0
Benchmark-20250310-232438.png


That's better! Thanks!

I was considering leaving some threads for Virtual desktop and Steam for VR though. I need to use Steam if I want hand/finger tracking to work.
 
I am benchmarking my shiny new PC (Ryzen 7 9800X3D & RTX 5090) and trying to get all caught up on everything to ensure I'm making the most of it.

I am still blown away by how much VAM slows down just due to having a lot of vars in AddonPackages. The gains I see in the benchmark results from my cpu-patched normal/full installation vs. a cpu-patched run with just the basic addons is nearly as big a jump as it is between the non-cpu-patched and cpu-patched that are both on the full addons folder.

Without CPU-patch
With 793GB in AddonPackages
1741813071251.png


With CPU-patch
With 793GB in AddonPackages
1741813131185.png


With CPU-patch
With only 863MB in AddonPackages (the defaults + benchmark and its dependencies)
1741813379785.png

The addons bloat really shows up in the PhysicsTime performance, such as in Baseline 3. Just having more vars is nearly doubling the PhysicsTime in that scenario. I may just end up taking that hit, but I'm considering trying the VARManager features of BrowserAssist again, which can dynamically offload vars, putting them outside AddonPackages and VAM's view while they're not actually needed.
 
I just finish building my new PC, I guess it supposed to be a beast but... it seems like I'm missing something because 4K and 1080p are not too far away from each other.
No overclocking, no change of any settings on 3rd party software or hardware beside XMP enabled on Bios, everything is so new that I barely have the machine on for a day now.

I guess this could affect some "extra points" but I'm not 100% sure:
The uptime of the PC is already 4+ hours (I installed windows, drivers, softwares etc..) before I tried VAM's benchmark so it wasn't "chill" maybe without using the PC before I could earn some extra points, but oh well...


4K res:
Benchmark-20250313-070717.png


1080p res:
Benchmark-20250313-082035.png
 
Hmmmpf ... 9070XT is within 4% of the 5070Ti and within 9% of the 7900XTX in rasterizing - the Benchmark I saw even had a massively overclocked "9070XT(X)" that beats the crap out of both the 5070Ti and the 7900XTX. Good OC/UV potential, very power-efficient, "grown-up" AI upsampling, finally "adequate" raytracing performance - and, very important for us VR-fiends: Apparently, it doesn't have rdna 3.1's troubles with hardware encoding.

Superb little card ... but ... scuttlebutt has it that the "aggressive pricing" was likely wishful thinking. Allegedly, the cheapest 9070XT will be offered at 815€+19%VAT = 969€ in Germany. That's a cool 42€ over "the last affordable 7900XTX in Central Europe" I snagged yesterday (the price for the same model jumped by 167€ 222€ overnight - stocks of XTX seem to be running low).

Got one :love: Pricing was ... OK - a ways above MSRP, but also not "Nvidia pricing".
Edit: Fuckme - the card I bought for 869€ an hour ago is now 1.357€ on Amazon! :sick:
Could you maybe be so kind as to do us all a favor, and post a vr benchmark result..? I think many of us are on the fence and waiting for benchmarks to pop up, I know I am at least

Edit: nvm, I ordered one anyway, so I will be doing a vr benchmark myself when I get it up and running
 
Last edited:
Back
Top Bottom