[HardOCP] AMD and NVIDIA GPU Vive VR Performance in Raw Data

[Arachnotronic]

As you can see from the HardOCP VR Leaderboard chart above, the pecking order of how to obtain the best VR experience is clear. The GeForce GTX 1080 is the king of the VR experience. The GTX 980 Ti and the GTX 1070 are pretty much trading punches for the number two spot, but the 980 Ti still edges it out. Surprisingly the R9 Fury X card is only slightly faster than the RX 480, but the Fury X still lags well behind the GTX 1060. For those of you looking for a "baseline" VR experience, it is hard for us to suggest the RX 480, but it can get the job done. If you are going to shell out the cash for the HTC Vive VR gaming system, and I suggest you do, you're best off putting an extra $60 towards the GTX 1060 as it is going to allow you a better visual fidelity than the RX 480 and a better experience.

http://www.hardocp.com/article/2016...ive_vr_performance_in_raw_data/6#.V596I4-cHZ9

Personally, I can't justify the huge cost of getting an HTC Vive, especially considering that there's pretty much no content yet. Would rather spend the $ on a high end monitor.

 

[Unreal123]

Wow just look at maxwell performance.

What i am seeing is that most of the VR games are being sponsored by Nvidia and recommending Nvidia cards only. However, i think there is only one VR game , that will be AMD sponsored is called Climb.

 

[sm625]

If I understand correctly, a ~99% reprojection rate basically means it is only running at 45 FPS with some sort of interpolation to get 90FPS? How can that be called VR capable?

 

[digitaldurandal]

Just to be clear this "HardOCP VR Leaderboard" is based on a single early access game?

@sm625 it is basically like Vsync. If the card doesn't hit 90+ it drops to 45 and shows each frame twice.

 

[Bacon1]

Just to be clear this "HardOCP VR Leaderboard" is based on a single early access game?

@sm625 it is basically like Vsync. If the card doesn't hit 90+ it drops to 45 and shows each frame twice.

Yes is a single Early Access game. Looks cool can't see how it turns out once released.

Reprojection is different than just pure doubling as it still re-renders part of the scene and "merges" them iirc. So it's not pure doubling.

 

[pj-]

UE4 is not the best engine for VR. Its deferred rendering is not good with VR for reasons I've forgotten. Oculus wrote a forward renderer for ue4 but I dunno if Raw Data uses it or if it works with vive.

I have a $700 144hz 1440p gsync monitor and a vive. I use the monitor a lot more but if I could only keep one I would keep the vive, no question.

It is low res and the games so far are very thin for what you pay, but the amount of pure fun I've had with it in 2.5 months is more than the last 5 years of regular games.

 

[antihelten]

This is clearly an issue with Raw Data being an early access game imho.

Generally speaking a game shouldn't use interleaved reprojection unless it has already dropped the adaptive quality level down by 3 to 4 steps.

A 480 or 1060 should never need to drop down 3 to 4 quality steps, given that games should target 970 level performance as the base quality level*. Anything slower than a 970 will then drop down in quality level (which basically involves reducing the target rendering resolution), and anything faster will increase in quality level (by increasing AA and resolution).

So if a 480, 1060 or Fury X goes straight to reprojection without first decreasing the quality level, then that looks like an early access issue and not a GPU issue.

For a description of adaptive quality in OpenVR and interleaved reprojection:
http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_Performance_GDC2016.pdf

*Saying that games should target 970 level performance might be a bit of an exaggeration, developers are of course free to target whatever level they desire, but 970 level is the recommended.

 

[pj-]

adaptive quality is a feature of Source 2 and the SteamVR unity plugin.

I don't believe valve has done anything for ue4

 

[bystander36]

Just to be clear this "HardOCP VR Leaderboard" is based on a single early access game?

@sm625 it is basically like Vsync. If the card doesn't hit 90+ it drops to 45 and shows each frame twice.

The reprojection system sounds most like a post process method to guess what the next frame will look like, so it is much better than nothing, but is prone to show artifacts.

I'd be interested in this type of tech for normal gaming, as I have the same motion sickness issues in any 1st person game until I hit 80+ FPS, although it's not bad at 60, it is really bad at 45 and below. 3D Vision was no worse than normal 2D gaming as far as I'm concerned.

 

[bystander36]

This is clearly an issue with Raw Data being an early access game imho.

Generally speaking a game shouldn't use interleaved reprojection unless it has already dropped the adaptive quality level down by 3 to 4 steps.

A 480 or 1060 should never need to drop down 3 to 4 quality steps, given that games should target 970 level performance as the base quality level*. Anything slower than a 970 will then drop down in quality level (which basically involves reducing the target rendering resolution), and anything faster will increase in quality level (by increasing AA and resolution).

So if a 480, 1060 or Fury X goes straight to reprojection without first decreasing the quality level, then that looks like an early access issue and not a GPU issue.

For a description of adaptive quality in OpenVR and interleaved reprojection:
http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_Performance_GDC2016.pdf

*Saying that games should target 970 level performance might be a bit of an exaggeration, developers are of course free to target whatever level they desire, but 970 level is the recommended.

I'd assume that players can turn settings down to attempt to avoid reprojection, but for the purposes for an apples to apples comparison, and to test the technology, they simply left all system on the same settings.

 

[Thala]

If I understand correctly, a ~99% reprojection rate basically means it is only running at 45 FPS with some sort of interpolation to get 90FPS? How can that be called VR capable?

It just means that the game is not optimized at all.
RAW Data was released into early access like 1 week ago...

 

[Yakk]

It just means that the game is not optimized at all.
RAW Data was released into early access like 1 week ago...

LOL... I guess the takeaway from that article is... the game is at least functional then.

 

[bystander36]

It just means that the game is not optimized at all.
RAW Data was released into early access like 1 week ago...

That or they just have to drop a few settings, so they don't have to rely on reprojection. It's a lot like Vsync with double buffering and a single GPU. If it can't meet your refresh rate, it drops to half FPS, but you can drop a few settings and counteract it. With the reprojection system, it does give you the option to use the higher settings, even if your system can't quite meet the demands.

It's pretty cool tech, and I'd like to see it in 2D as well. I suffer the same motion sickness in any 1st person game, 3D or not.

 

[Thala]

That or they just have to drop a few settings, so they don't have to rely on reprojection.

Indeed. RAW Data has quite a few gfx settings. I do not dare to chose the highest setting with my Nanos On the other hand, the highest settings do not improve quality anyway (not that i have noticed).

Would rather spend the $ on a high end monitor.

Sure, but no high end monitor will give you the "i am there" experience. Not even close.

 

[antihelten]

adaptive quality is a feature of Source 2 and the SteamVR unity plugin.

I don't believe valve has done anything for ue4

UE4 also uses the OpenVR/SteamVR API for VR, but they may not have implemented adaptive quality yet.

Valve does seem to work primarily with Unity and as such unity does seem to be the first to get new features.

Also to my knowledge the Source 2 engine doesn't support adaptive quality, remember The Lab game uses Unity not Source.

 

[pj-]

UE4 also uses the OpenVR/SteamVR API for VR, but they may not have implemented adaptive quality yet.

Valve does seem to work primarily with Unity and as such unity does seem to be the first to get new features.

Also to my knowledge the Source 2 engine doesn't support adaptive quality, remember The Lab game uses Unity not Source.

adaptive quality is an engine level feature, it is not the responsibility of steamvr itself. That is why valve had to release a plugin for unity for it to support the feature.

Source 2 does support adaptive quality and I believe it was done there before the unity plugin was created. The lab uses a mix of source 2 and unity. The robot repair demo is source 2, and a lot of the examples from the 2015 VR presentation were from that demo. I think the photogrammetry locations are also source 2 but I'm not sure.

 

[poofyhairguy]

Wish they would have tested a Hawaii card (390, 390x) too.

 

[Det0x]

Raw Data: Greatly-Anticipated VR Action Game Out Today; Uses NVIDIA VRWorks For 30% Faster Performance

http://www.geforce.com/whats-new/ar...sed-performance-accelerated-by-nvidia-vrworks

Nvidia Showcase: Early Access: VR first-person shooter Raw Data unleashed on Steam

http://www.pcgamesn.com/nvidia-show...st-person-shooter-raw-data-unleashed-on-steam

The only thing i can conclude from this thread is that gameworks™ seems to be working as intended..

Seems like you will only get a worthwhile player experience on a gtx 1080/1070 :thumbsup: (i try to avoid recommending legacy hardware)

Time to upgrade those 1060's again

 

[antihelten]

adaptive quality is an engine level feature, it is not the responsibility of steamvr itself. That is why valve had to release a plugin for unity for it to support the feature.

Source 2 does support adaptive quality and I believe it was done there before the unity plugin was created. The lab uses a mix of source 2 and unity. The robot repair demo is source 2, and a lot of the examples from the 2015 VR presentation were from that demo. I think the photogrammetry locations are also source 2 but I'm not sure.

Honestly, I always thought that the adaptive quality feature was a fairly light plugin for unity that mainly just called a number of features in steamVR (such as render target multiplier and the like), but looking at the asset store entry it does look like it pretty much replaces the unity renderer wholesale (since it's bundled with the forward renderer version of unity).

 

[Yakk]

http://www.geforce.com/whats-new/ar...sed-performance-accelerated-by-nvidia-vrworks



http://www.pcgamesn.com/nvidia-show...st-person-shooter-raw-data-unleashed-on-steam

The only thing i can conclude from this thread is that gamewreck™ seems to be working as intended..

Seems like you will only get a decent player experience on a gtx 1080/1070 :thumbsup:

Time to upgrade again folks

Ah!

There were go, that explains everything.

 

[digitaldurandal]

Ah!

There were go, that explains everything.

Yea, also explains the weird [H] recommendation of a VR card based on an early access title. Kyle really has a grudge against AMD and IMO the site has gone off the deep end because of it. Leading up to the launch of the 480 he made some crazy claims, some things he said a professional reviewer should just know better (things like 14nm is supposedly 30% faster than 16nm.) Which is very unfortunate because I actually prefer their review style over other sites hands down.

Anyway, not trying to take this thread off topic but it really made me scratch my head, I had expected based on the title of the chart that they had benched multiple VR games for the rankings which made me very surprised to see the Fury X under a 1060. When I saw it was a single game I was even more confused.

Nvidia supposedly does have some proprietary technologies up their sleeve for VR that the 10xx series can use to reduce workload for stereo rendering. Considering the placement of the 980ti im not sure this is actually taking advantage of it unless the 9xx series has the capability as well.

If it is using those technologies this is interesting in another way, maybe would have been better to focus the review on those technologies instead of selling the review as VR recommendation article. His site though so.. w/e.

 

[USER8000]

So Nvidia does well in a Nvidia VRWorks title is what I take from that.

 

[Stuka87]

So, they ran numbers for a single GameWorks VR game that is in early access. We know [H] is out for AMD, but this basically reads like an nVidia PR statement.

How can anybody give an honest suggestion for what card to get based on a sample set of 1? I have no doubt that a 1080 is going to be faster than an RX 480, but its obvious there is an issue with this game on AMD hardware when an RX 480 and a Fury get the exact same performance.

 

[sandorski]

According to this LinusTechTips review of the 480's VR capabilities, there are no established testing methodologies for VR yet. I would take this article with a huge grain of salt for that reason alone.

 

[master_shake_]

i tried my friends HTC Vive and he has a single 980.

smooth experience and do want.