AMD & NVIDIA GPU VR Performance: Valve's Robot Repair @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
54,871
AMD & NVIDIA GPU VR Performance: Valve's Robot Repair - Robot Repair is a mini-game that you can download for free from Steam to use with your HTC Vive VR system. It is highly likely that anyone using a new HTC Vive will play the mini-games inside Valve's The Lab. This is a Source 2 engine demo and perfect to validate prior GPU performance we have tested.
 
I love The Lab as an intro to VR demonstration. That and Job Simulator are both great interactive demos for new people who have never tried it before.

I would say for non-interactive, theBlu is amazing.
 
Kyle's ending at his review, was exactly what i've been thinking:
Raja Coduri must have been drinking (i had something else in my mind but i preferred to use a "light" term instead) the day he made the statement about a superior VR performance!! :wacky:
 
Kyle, I know you mentioned in the Raw Data performance review that you have not unboxed your Oculus Rift due to the lack of room scale/touch controllers but I'd be very curious to see if there are any performance differences between titles that work on both due to the hardware/runtimes associated with each.

Any plans on doing some 'boring' apples to apples comparisons between the Rift and the Vive any time soon? Something that can show us the difference between the Vive's reprojection technique vs. the Rift's Time Warp perhaps.
 
Yeah it would seem that way now. AMD keeps on disappointing me although all they have this gen is the mainstream 400 series so far, still you can see they have loads of work to do still. Hopefully DX12/Vulcan VR games will look better for them, but I aint holding my breath. Tired of hoping they can turn things around only to be disappointed in the end.
 
I am really interested to see how well the 480 will fair on a VR game that favors more on AMD architecture. Things just aren't looking good for AMD in the VR front.

On the plus side, Titan X in VR ranking! That means I will looking forward to the GPU review!
 
After reading the last several VR articles, I keep asking myself is there is any point to the Titan X for VR?

I've been getting back into PC gaming after a 5 year hiatus entirely because of my interest in VR. Long story short, I picked up at GTX 1080 and the performance is excellent. I then grabbed a Titan X as I thought it would provide a better experience but given the content, it doesn't appear to be the case. Am I missing something?

Another great read, the [H] continues its best in class analysis!
 
Sonicks said:
Kyle, I know you mentioned in the Raw Data performance review that you have not unboxed your Oculus Rift due to the lack of room scale/touch controllers but I'd be very curious to see if there are any performance differences between titles that work on both due to the hardware/runtimes associated with each.

Any plans on doing some 'boring' apples to apples comparisons between the Rift and the Vive any time soon? Something that can show us the difference between the Vive's reprojection technique vs. the Rift's Time Warp perhaps.
Click to expand...
No, I am not going to be digging into the HMD hardware. There are simply a ton of others covering that, and a large amount of user reviews in this forum alone.

Edit: FYI, I sold the Oculus while I could still get what I paid for it. Unless I see some big reason to buy one again after the motion controllers come out, I am sticking with the Vive. If you are buying into VR any time soon, make sure you get motion controllers. It greatly opens up the world of VR gaming.
 
Last edited:
VitaminTee said:
After reading the last several VR articles, I keep asking myself is there is any point to the Titan X for VR?

I've been getting back into PC gaming after a 5 year hiatus entirely because of my interest in VR. Long story short, I picked up at GTX 1080 and the performance is excellent. I then grabbed a Titan X as I thought it would provide a better experience but given the content, it doesn't appear to be the case. Am I missing something?

Another great read, the [H] continues its best in class analysis!
Click to expand...
From the games I have tested so far, no, the Titan X does not give a better VR experience using the Vive. That said, once we see that are truly pushing VR visual fidelity, it will start to show off its chops. One game where it might show its prowess is Raw Data, when pushing the visuals "all the way up."
 
The Rift HMD is better but until their motion controllers come out I can't argue with going with the Vive.

Happy the 980 Ti is holding up so well in these reviews, looks like 4K flat panel gaming will be what drives the eventual upgrade.
 
Sonicks said:
Fair enough. I'm in the market for a VR headset and I'm leaning towards the Rift only because it's entry price is much lower than the Vive's and I'm sure with it's motion controllers the price will even up, but at least the initial cost is much less.
Click to expand...
Keep in mind that you will likely have to purchase additional sensors for room-scale VR as well with the Oculus.
 
Thank you Kyle - about as straightforward as it gets. The only, ONLY thing I can even possibly consider for the poor 480 performance against AMD's claims (and keep in mind, I am intentionally stretching things here), is that they really want mGPU working, and are banking on being able to have "1 GPU per Eye" to compete with the higher end NVidia cards. But even then, simplest answer is always the best - AMD is just trying to move product and put lipstick on a pig. There's a reason Raja was moving already to Vega before Polaris launched.
 
Alternate title for this could have been: "Nvidia is King and AMD Sucks: Here's Why"

But I have to ask: Given there has been no change on the financial/personnel levels of each company relative to each other, why would you really expect the situation to change drastically? So long as Nvidia has more money and people, it will likely be able to hold the performance crown and provide better driver support. That said, I still think AMD offer great products for the money and were right to target the bulk of the market using a smaller die on a new node with the 480. Doom on my OC'd FX 8370/R9 290 with Vulkan has been excellent, and I'm confident Quake Champions will perform similarly on it.

I hate to say it, but I suspect when Zen finally arrives I expect a review here that sounds a lot like this one. And yet for the vast majority of consumers it will probably offer an excellent value.
 
Zion Halcyon said:
Thank you Kyle - about as straightforward as it gets. The only, ONLY thing I can even possibly consider for the poor 480 performance against AMD's claims (and keep in mind, I am intentionally stretching things here), is that they really want mGPU working, and are banking on being able to have "1 GPU per Eye" to compete with the higher end NVidia cards. But even then, simplest answer is always the best - AMD is just trying to move product and put lipstick on a pig. There's a reason Raja was moving already to Vega before Polaris launched.

Vega does have me curious though, mostly because I am not sure if Vega is still another hold over from Pre-Raja given the development cycle. I can't say I blame AMD - they are trying to stay afloat, but we really won't know the whole story until Raja finally have a chip he designed from conception to sale, and I think in terms of the usual R&D, Vega may be too far along in the R&D process when Raja came on board for him to put his stamp on it. I think Navi may be the first real Raja chip since AMD started investing in R&D again.

Kyle, have you heard any Vega rumors akin to yours for Polaris, or anything confirming (or refuting) my suspicions?
Click to expand...

Ok so you are going to wait to see if AMD get mGPU shit together, but if that fails then Vega might be the answer? Or maybe Navi? While AMD fails to deliver, Nvidia keeps dick punching them with no sign of letting up.

AMD keeps dangling out the same bait and people keep biting it. "Just wait for [next thing], it will be fine." Then they release the 480 with Premium VR (horrible) and premium components (horrible cooler). I'm surprise AMD can even keep a straight face at this point.
 
Last edited:
Well, with VR the proof is in the pudding and we have numbers to verify that AMD is full of shit on these claims.
 
Last edited:
AzixTGO said:
its proof of nothing. I wouldn't expect any different from what we see in some dx11 games. Its down to the software. No matter what the hardware can actually do, its always down to the software that runs on it. People act like that basic fact is complicated. We know what happens with dx11 already, which is why I don't think VR is going to ever mature on the api. That's how you will get things like this running consistently solid on lower end hardware (and cheaper headsets of course), opening up VR to the masses. As it is you are talking almost $1000 dollars for so-so graphics with the 1070 up.

I still don't get why the hell they even go with dx11 at the end of its life at the forefront (well besides being easier for early development I guess).

Its not about mGPU either. You cannot draw universal conclusions from this type of thing. It doesn't mean the hardware is inadequate till you at the very least use a lower level api to say you actually tapped it out or nearly did. Also wonder about the effects of using gameworks vs or liquid vr, if that would improve things.
Click to expand...

That is certainly true, but for AMD to tout VR Ready, it better be VR Ready and at this moment it is not. You are correct, current VR games that Nvidia is doing well is by no means AMD won't win in the future but for a lot of us, we buy for the present, not the future and at the present, Nvidia is winning.
 
AzixTGO said:
its proof of nothing.
Click to expand...
You are exactly correct, real world gaming performance has and will never be an indicator of how well a GPU actually works! We are in no position to speak to how well these GPUs work today until we see how well those perform weeks, months, possibly even years from now. The truth is out there. :rolleyes:
 
For all intents and purposes, the "$200" RX 480 4GB is a figment of the imagination if you actually go to purchase one. Why in the world would you want to "save" $10 and purchase the RX 480 8GB card over the GTX 1060 6GB for VR usage? Let me answer that for you, you wouldn't.
Click to expand...
For all intents and purposes, the "250$" 1060 is figment of the imagination if you actually go to purchase one.
Same goes for "$380" 1070 and "$650" 1080.

The "$240" 480 8GB also seeem to be a figment - right now the only 480 8GB in stock at newegg costs 289$ - which probably explains why it's in stock: a 1060 is both faster and cheaper (and overclocks better and uses significantly less power). I just noticed that product photos of 480s on NewEgg have a small "PREMIUM VR" logo in the bottom-right corner. Pretty ironic.

radeon-vr-ready-premium-red-logo-2016-210.png
 
Simplex said:
The "$240" 480 8GB also seeem to be a figment - right now the only 480 8GB in stock at newegg costs 289$
Click to expand...
In my defense, when I published this morning, there were at least two RX 480 8GB cards in stock at $249.
 
SighTurtle said:
No 390s or 970s for testing?
Click to expand...

Heheh, with the 480 in such bad shape do you really think you'd have any worthwhile experience on the previous gen 390?

The 970 is the absolute minimum for both headsets on the nVidia side so it's easy to guesstimate where in the [H] ranking it'll land.....right above the 480 :p
 
AzixTGO said:
its possible. The 390/390x and 480 have different performance characteristics and obviously hardware. Often the hawaii chips do well in some situations and 480 in others. In dx12 I think the older chips might actually be faster currently.
Click to expand...
Odd, I did not see that referenced in AMD's Premium VR press release.
 
Simplex said:
For all intents and purposes, the "250$" 1060 is figment of the imagination if you actually go to purchase one.
Same goes for "$380" 1070 and "$650" 1080.

The "$240" 480 8GB also seeem to be a figment - right now the only 480 8GB in stock at newegg costs 289$ - which probably explains why it's in stock: a 1060 is both faster and cheaper (and overclocks better and uses significantly less power). I just noticed that product photos of 480s on NewEgg have a small "PREMIUM VR" logo in the bottom-right corner. Pretty ironic.
Click to expand...

Naturally, the only cards that always sold out are the ones close to MRSP. Popular models too like MSIs. Definitely wouldn't buy them at $330+, seeing as the cheapest 1070 is $399. I bet everyone's camping on those models.

Germanicus said:
Alternate title for this could have been: "Nvidia is King and AMD Sucks: Here's Why"

But I have to ask: Given there has been no change on the financial/personnel levels of each company relative to each other, why would you really expect the situation to change drastically? So long as Nvidia has more money and people, it will likely be able to hold the performance crown and provide better driver support. That said, I still think AMD offer great products for the money and were right to target the bulk of the market using a smaller die on a new node with the 480. Doom on my OC'd FX 8370/R9 290 with Vulkan has been excellent, and I'm confident Quake Champions will perform similarly on it.

I hate to say it, but I suspect when Zen finally arrives I expect a review here that sounds a lot like this one. And yet for the vast majority of consumers it will probably offer an excellent value.
Click to expand...

Because of context. If they market it as 'lowers the barrier for DX12/Vulkan', sure, I'm certain Kyle will give them a free pass for it. In fact, if you haven't noticed, the RX480 and 470 are [H] Gold award winners.

Sadly, they marketed it wrong - 'lowers the barrier' to VR (remember Radeon Rebellion?), and yet giving poor performance for value on actual VR tests? Why would [H], given the context of this review (*a VR review*) give them a free pass?
 
Last edited:
It seems like something is wrong with either the test or the game. The slide on the first page says that the default setting is native res + 4x msaa, and a 970 stays at or above that level. The implication there is that it maintains 90fps or better and sometimes can use the higher settings. Why on earth would a Fury X be reprojecting half the time? If the adaptive IQ was broken, you'd think it would be running at the default "0" level, which should be no problem for Fury X in such a simple game.

Have you tried running the SteamVR performance test to see if the results you get are in line with what other people have gotten?
 
I'm kind of curious how widespread VR headset use is on here at this point.

Have you guys been rushing out to buy them, or are they still pretty rare in the community?
 
Couldn't help but notice the presence of a Titan X in this lineup.

Will a full [H] review of the Titan X be forthcoming?
 
pj- said:
It seems like something is wrong with either the test or the game. ................................
Click to expand...


Yeah, in 4 VR tests that are made so far, in which AMD finishes in all of those tests at the bottom, we should blame the tests / games and not AMD !!! AMD always does everything fine!!o_O
 
pj- said:
It seems like something is wrong with either the test or the game. The slide on the first page says that the default setting is native res + 4x msaa, and a 970 stays at or above that level. The implication there is that it maintains 90fps or better and sometimes can use the higher settings. Why on earth would a Fury X be reprojecting half the time? If the adaptive IQ was broken, you'd think it would be running at the default "0" level, which should be no problem for Fury X in such a simple game.

Have you tried running the SteamVR performance test to see if the results you get are in line with what other people have gotten?
Click to expand...

Shot in the dark here but you're assuming an apples to apples comparison as if these were standard desktop games.

Maybe the Fury X architecture isn't very efficient at VR, meaning it can't render two simultaneous full resolution scenes as well as a 970 can.
 
Sith'ari said:
Yeah, in 4 VR tests that are made so far, in which AMD finishes in all of those tests at the bottom, we should blame the tests / games and not AMD !!! AMD always does everything fine!!o_O
Click to expand...


I would still like to know why this is the case though. Is it un-optimized drivers ? Is it the game? Maybe its the hardware? Could be that its all of the above ?
 
Sith'ari said:
Yeah, in 4 VR tests that are made so far, in which AMD finishes in all of those tests at the bottom, we should blame the tests / games and not AMD !!! AMD always does everything fine!!o_O
Click to expand...

I'm not trying to defend AMD, I'm trying to rationalize something that doesn't make sense to me. The last AMD product I owned was a radeon 9600 non-pro. I just got a pascal Titan X so AMD's performance doesn't affect me at all. It is just sad if it's the case that a Fury X can't handle something as simple as robot repair without reprojecting. I hope for the sake of AMD and AMD owners that something is wrong and can be addressed with software fixes or testing methodology changes.

Sonicks said:
Shot in the dark here but you're assuming an apples to apples comparison as if these were standard desktop games.

Maybe the Fury X architecture isn't very efficient at VR, meaning it can't render two simultaneous full resolution scenes as well as a 970 can.
Click to expand...

I guess, but Valve has clearly set up robot repair specifically as a tool to test VR and it has existed for years now. I expect it to be one of the best optimized experiences for both GPU vendors. Reprojection is supposed to be a last resort when there's no way to maintain 90FPS. Why would they list a 290 as the AMD minimum for Vive if even a Fury X can't handle a barely interactive tech demo? Just doesn't smell right to me.
 
Spyhawk said:
I would still like to know why this is the case though. Is it un-optimized drivers ? Is it the game? Maybe its the hardware? Could be that its all of the above ?
Click to expand...
VR requires two scenes to be rendered, one for each eye. My understanding is the geometry is processed twice, once for each scene (a problem that is alleviated with nVidia's SMP). Aren't AMD products way behind nVidia with respect to geometry performance? Perhaps that's a major issue.
 
pj- said:
I'm not trying to defend AMD, I'm trying to rationalize something that doesn't make sense to me. The last AMD product I owned was a radeon 9600 non-pro. I just got a pascal Titan X so AMD's performance doesn't affect me at all. It is just sad if it's the case that a Fury X can't handle something as simple as robot repair without reprojecting. I hope for the sake of AMD and AMD owners that something is wrong and can be addressed with software fixes or testing methodology changes.
Click to expand...

Since you mentioned FuryX, you have to consider that even though it's AMD's top-gpu (*and the most expensive one), it can not compete even with the half-priced RX480 (in this specific VR test i mean, not in general).
So since you like rationalizing (*and i like it myself as well ;)), since we have 2 AMD's GPUs, and the medium-ranged one surpasses the top-range one, then in my opinion, the problem is with AMD's cards and not with the tests/games. EDIT: hmm, i just noticed that FuryX, finished last in this evaluation, probably because of its high price and not only for pure performance metrics! )
 
Last edited:
Sith'ari said:
EDIT: hmm, i just noticed that FuryX, finished last in this evaluation, probably because of its high price and not only for pure performance metrics!
Click to expand...
Just to remind you....

Our VR Leaderboard something new for us and currently it is a narrow look at VR gaming. We are going to compile a subjective rankings chart for GPU performance in VR that is based on objective data at its heart. This VR Leaderboard will surely grow and change as we update our data with new games as well as new GPUs and video cards...... Our chart here will be based on real world gameplay performance and experience, data collected from that gameplay, and our subjective thoughts and experiences.
 
pj- said:
It seems like something is wrong with either the test or the game. The slide on the first page says that the default setting is native res + 4x msaa, and a 970 stays at or above that level. The implication there is that it maintains 90fps or better and sometimes can use the higher settings. Why on earth would a Fury X be reprojecting half the time? If the adaptive IQ was broken, you'd think it would be running at the default "0" level, which should be no problem for Fury X in such a simple game.

Have you tried running the SteamVR performance test to see if the results you get are in line with what other people have gotten?
Click to expand...

I'm pretty sure it's a bottleneck on AMD hardware + software. The testing methodology is extremely unlikely to be faulty. Certainly not the first time AMD has been hampered by issues with software/hardware that cause underutilization.

The results are in line with pretty much all the other VR games tested so far.
 
Something just dawned on me. If the RX 480 is a "Premium VR Experience," what does that make all the NVIDIA Cards on our chart?
 
You must log in or register to reply here.
Back
Top