AMD vs Nvidia - VR edition

Associate
Joined
6 Feb 2004
Posts
1,328
Location
Toon
Like most I'm sure, I've been keeping a close eye on the new GFX releases from both sides. If they'd been available, I'd have a 3080 by now, but lack of availability has afforded pause for thought.

I've got a x570 - 3700x - 5700XT system at present, and for VR have a Quest 1 which I use mainly for sim racing via the link cable. I have a G2 pre-ordered though, day 1 order so should be here in the next week or two.

Based on price/performance AMD look like the right choice based on what little we currently know. However I'm got nagging doubts about their VR performance in general. I didn't really see the massive gain I was hoping for in VR when going to the 5700XT from a 980TI, which makes me think perhaps Nvidia are in general a better choice for VR?

Additionally to further complicate matters I may well add a new generation Zen to my x570 depending on availability, which will unlock those new AMD only features.

Of course I'm not really testing native VR here, it's just streaming via link to the quest - should I expect a big difference in performance when running 'natively' in VR?

Obviously we only have real-world data for the green team at the min, but I was hoping this thread could serve as a place where VR performance in specific games can be reported as and when we find out more about any of the new cards. I'm going to hold off buying anything on the GFX side until more is known about VR performance. However as VR is still clearly pretty niche, there is not a huge amount of info out there on the subject.

Anyone already using the 3080 for VR here? Or planning to go AMD?
 
Everyone seems to be in much the same boat then. I'd still probably go Nvidia if I had the choice of both right now, but not sure what I'm basing that on. Any more thoughts on general AMD Vs Nvidia VR performance? A few seem to think Nvidia are inherently better for VR, is this mindshare or real? Any benchmarks etc from previous generations which illustrate this?

P.S. 1000 posts! Only took me 20 years :)
 
The key thing to remember is that these are marketing slides. Benchmarks done by AMD. Wait for reviews before making your mind up. Saying that I do think the RDNA 2 cards are looking very good.

Of course. And even if they are spot on, still no real idea what VR performance we'll get. Roll on some proper reviews, ideally VR centric.
 
Hmm interesting. So the GPU itself cannot directly output a compressed video stream via its internal USB port, it has to back and forth communicate with the CPU to do so?

This is one of the things I'm hoping improves performance when I get the G2, coming from the Quest 1 + Link cable. Instead of streaming compressed video with the overheads that brings, the image is going directly from graphics card to the HMD. Not sure exactly what sort of difference I'd expect to see, certainly I think there should be a very noticeable visual improvement, aside from the obvious huge jump in resolution.

As it seems unlikely that I'll have a new Graphics card by then, I should be able to test on a similar resolution to the quest by downsampling, and get some idea of how performance might be affected by the link's compression.
 
According to this, "the PC compositor encodes the eye textures -- surfaces that an application submits to the compositor -- using H.264 video compression, and sends the images as slices to the remote client over USB."

I think that compression would probably be done on the GPU wouldn't it?

*edit* same question asked (and answered) here
 
I fundamentally disagree. It's not a trope... fact of the matter is that it's an absolute bear to run in VR compared to eg AMS2, Project Cars 2 or RF2 for example. It's great that you have it running well enough for you but you have a frikkin 1500 quid graphics card buddy.

It's still my favourite game of the lot, but despite a perfectly respectable spec I struggle much more to get good performance than in any of the others listed (all of which I also personally run in VR) and have to make sacrifices like not having blooming wing mirrors so don't go telling me I'm peddling tropes! I'm talking from my own personal and extensive experience with it and other games in VR, not parroting some ******** I read some guy say on a forum somewhere. I'm a long way from being anti VR and a long way from being anti ACC - it's a simple fact that it's just very poorly optimised to run in VR relative to pretty much any of the competition.

+1. I've got AC, AMS2, PC2 and RF2 which I can run happily on my 5700XT, I bought ACC hoping I could do the same but spent about an hour messing around with absolute minimum settings before consigning it to the 'play when system is beefy enough to handle it' pile. Seems like it will be great, but not yet for me.
 
So I managed to snag a 3080FE in today's drop. I know last drop there were problems with over-ordering so could be a further wait, but if not, there's a decent chance I will be getting the G2 and the 3080 on the same day, or close enough. All I need now is for the 5900X to suddenly become available for the full house.

Looking forward to comparing to the 5700XT, I'm expecting great things.
 
Thought this may be of interest:

old RX480 score on Steam VR performance test:

DiB1TGd.png



And now the new 3070:

9eXkPqd.png



However i have to say, in Air Car running over VD, I still can't push pixel density to 1.0 otherwise I cant get 72 fps. (VD on medium, 80 bitrate).

Nice idea, just what we need in this thread. This is what I get on the steam test and the open VR test for my 5700XT. 3080 arrives tomorrow, so will be interesting to see a comparison

npb08gW.png
CwHJvx0.png
 
The Steam VR test isn't really very good. It's only for lowest baseline of performance.

Noted. I'll ignore that with the new 3080 then. The OpenVR benchmark is way over double the 5700XT's score though:

BD2vwIj.png

Had a brief play on some games yesterday, RDR2 allowed me to push the settings much higher but honestly it looks so good anyway it's hard to tell the difference.

in VR I only really tried Asssetto Corsa with the SOL Mod and I was able to hold the Quest's forgiving 72hz no matter what settings I threw at it, which is definitely a big improvement. Today I'm going to try ACC again, having given up trying in VR with the 5700XT. Also downloaded Control to see what Raytracing/DLSS looks like :) And still the G2 to come, hopefully on Tuesday all being well :)
 
You might be interested in this video as you're on a 3700X. He only does I racing, but it's got me considering rebuilding my system over to AMD once the top end cpu become more available.


Wow thanks. I'm also in the market for a CPU upgrade, got the funds ready for a 5900X when it's possile to get hold of one. Mostly I was thinking it would future-proof me a bit as after this raft of upgrades I'm not planning on making any big purchases for some time. I wasn't thinking I'd be seeing such an improvement, but that is a big difference. Can't wait to get hold of a new CPU now :D
 
Last edited:
It popped up with a message that my Supersampling needed to be reset when I ran it, and I said yes. I assumed that was the default resolution for a quest, but it seems it should be 1440 X 1600. Steam VR seems to be doing some weird stuff with res - I'm on a G2 Facebook group and it seems that due to a bug, you need to set Steam resolution to around 50% to get it to render in the correct res for the G2.

Just had another go there, 100% is giving me 1808 x 2000, which is clearly wrong. Closest I can get it is 64%, which is 1444 x 1600. I have to ignore the message from the benchmark that I should set Supersampling to 100%, or it resets. Not sure what's going on tbh :rolleyes:

5Z49o1T.png
 
It is the Oculus link software that sets the rendering resolution and framerate in the benchmark. This means that if we're comparing settings we have to be mindful of the Link settings.

Here is a series of tests on my hardware. The performance is clearly dependent on resolution and not refresh rate at all, which is interesting. I assume this is because it doesn't matter what the refresh rate is set to in my case, as the GPU is not reaching it at these tested resolutions.

TZkNpXt.png

Native resolution on the Quest 2 is about 7 million pixels, so Im getting around 50 fps average at that resolution on the benchmark (reading off my chart).

Note that the 'auto' setting on link, is lower resolution if you select a higher refresh rate. That's why the first three points on my chart are spaced out. The next two tested resolutions were the same at each refresh rate (but higher multipliers shown on the slider).

'4K' is about 8.3 million pixels, so some of the higher link resolutions are well beyond 4K. I stopped at 4464x2256 but it goes higher still. At its highest its 50% more pixels than 4K.

That could explain it I guess, in the oculus debug tool I could have something funky going on. Will have another look tomorrow
 
It is the Oculus link software that sets the rendering resolution and framerate in the benchmark. This means that if we're comparing settings we have to be mindful of the Link settings.

Here is a series of tests on my hardware. The performance is clearly dependent on resolution and not refresh rate at all, which is interesting. I assume this is because it doesn't matter what the refresh rate is set to in my case, as the GPU is not reaching it at these tested resolutions.

TZkNpXt.png

Native resolution on the Quest 2 is about 7 million pixels, so Im getting around 50 fps average at that resolution on the benchmark (reading off my chart).

Note that the 'auto' setting on link, is lower resolution if you select a higher refresh rate. That's why the first three points on my chart are spaced out. The next two tested resolutions were the same at each refresh rate (but higher multipliers shown on the slider).

'4K' is about 8.3 million pixels, so some of the higher link resolutions are well beyond 4K. I stopped at 4464x2256 but it goes higher still. At its highest its 50% more pixels than 4K.

Sorry not sure what the link Slider is? I've used the debug tool to increase pixel density before, but not used a slider? Also I presume this is the Quest 2 you are using? I'm still on the quest 1, so a lower resolution
 
Oh im not sure if you get the slider on a Quest 1? Its in the Link software itself now, on the latest version. When you select your headset there are options for changing the resolution.

I don't get any of that, unless there's something I'm missing. Not too bothered anyway as the Quest will be unused as of tomorrow when the G2 arrives. And will be replaced with a Quest 2 before Christmas :)
 
I’ve basically decided whoever can offer me an msrp FE/MBA card first will get my cash this time.

6800XT reference or 3080FE... either will be a dramatic step up from my 2070S

Very similar to what I've done going from 5700XT to 3080FE. Would have happily grabbed a 6800XT had one been forthcoming. Was such a relief to turn off the discord notifications after snagging the FE - but have just turned them on for the 5900X now haha
 
Anyone know what sort of improvement you can get going from 3900x to 5900x? - I'm mainly a sim racer with some flight simulation as well.

Trying to decide how best to spend up to about £1000 - either CPU or GPU, or just wait!

According to this video @lokiss found, it's a pretty significant difference - certainly from 3700x anyway.

 
5 months since the 3000 series release and I still have no real idea if AMD would be worth a buy this gen, mostly as the AMD cards seem even more rare than the Nvidia ones.

In hindsight I'm extremely pleased I managed to snag the 3080FE but I still feel like there's quite a lot more to come with drivers, just hope there's some real focus on improving VR performance, still dropping frames all over the place even with the newest driver which is meant to address that.

Anyone on here got their hands on an AMD 6000 series card and tried in VR?
 
Back
Top Bottom