SteamVR Performance Test: 16 GPUs Compared

[DuncDog]

If any Tom's employees are still floating through the comments on this post, might I suggest an addition of the (3/6GB) 970M and the (4/8GB) 980M to the data points.

For anyone who is interested, I was able to sit right at the top of the Capable block, just a hair off of the Ready area with my Asus ROG G751JY with the 4GB 980M. After a small OC of +135/+500 I was able to sit comfortably in the middle of the Ready block. Don't have the exact fidelity numbers (can grab them later and rerun the test since it has updated since I ran it last), but I do remember where I placed in the test.

Test System:
http://www.newegg.com/Product/Product.aspx?Item=N82E16834232849

 

[huilun02]

And if you compare results between the 680 and 770 and the whole thing goes south...

Also the test almost completely ignores CPU performance.
Which brings to light the whole motive behind it.

 

[InvalidError]

Also the test almost completely ignores CPU performance.
Which brings to light the whole motive behind it.

The main impact of VR is doubling the GPU load by having to render two images. From the CPU side of thing, the game logic only needs to get iterated once per scene regardless of how many viewports into the scene are open. Send the whole scene without hidden surface removal/culling to the GPU(s) - let the GPU's early discard take care of that on a per-viewport basis, setup your viewports and the CPU performance impact of rendering the game on a single display or stereoscopic VR should be slim to none.

 

[kcarbotte]

If any Tom's employees are still floating through the comments on this post, might I suggest an addition of the (3/6GB) 970M and the (4/8GB) 980M to the data points.

For anyone who is interested, I was able to sit right at the top of the Capable block, just a hair off of the Ready area with my Asus ROG G751JY with the 4GB 980M. After a small OC of +135/+500 I was able to sit comfortably in the middle of the Ready block. Don't have the exact fidelity numbers (can grab them later and rerun the test since it has updated since I ran it last), but I do remember where I placed in the test.

Test System:
http://www.newegg.com/Product/Product.aspx?Item=N82E16834232849

Laptop tests are definitely something we're looking into. This article was specifically about different GPUs though. Changing other variables of the PC makes it no longer about the GPU, but the computer itself.

 

[DavidGX_]

No Fury X?

 

[HaB1971]

I have an i7-2600 running at 3.4Ghz, EVGA Superclocked GTX980 ti, 16GB of ram and 1.5TB of SSD and it hits 11 with no problem on this test... proof of the pudding will be the real thing rather than this emulation. Would love a try before you buy but that won't happen... So what is it to be Oculus or Vive?

 

[kcarbotte]

No Fury X?

Don't have one.
We'll have one for proper VR tests though.

 

[kcarbotte]

I have an i7-2600 running at 3.4Ghz, EVGA Superclocked GTX980 ti, 16GB of ram and 1.5TB of SSD and it hits 11 with no problem on this test... proof of the pudding will be the real thing rather than this emulation. Would love a try before you buy but that won't happen... So what is it to be Oculus or Vive?

You won't have any trouble with that PC, though I would run the Oculus compatibility utility. I have the same CPU, which doesn't pass (but I doubt it has any major impact. The 2600k is still a very strong processor.) My motherboard isn't compatible with the rift though. something with the USB controller.

 

[IntelDiesel]

I have an original i7 920 @ 3.7 12 gig ram and an r9 390 and says I'm ready?!

 

[kcarbotte]

I have an original i7 920 @ 3.7 12 gig ram and an r9 390 and says I'm ready?!

It's far more about the GPU than it is about the CPU.
I wouldn't be surprised if it is.

 

[IntelDiesel]

8 year old rig READY LOL?

 

[kcarbotte]

8 year old rig READY LOL?

Don't see why not.
There isn't a tremendous diference in CPU performance between an i7 920 and an AMD FX 8350.

 

[beetlejuicegr]

I discussed the way steam approached this with some friends. Please read carefully

Steam is selling Steam machines. There is no better and more cheaty way that this benchmark.
They try to keep over 100fps while reducing detail on the fly as the demo goes. That doesn't mean the game won't play in VR gear if you are on low "fidelity" . For me it just means the experience won't be 100% smooth and possibly it will just be low fps (tearing or something similar).
Also i have for example i5.2500k @4.3Ghz. the details on the vr demo showed "no frame lost due to cpu bottleneck". But it also mentions "no cpu loss was calculated for positional calculations" well even cheap smartphone cpus can do it..

For me this is a very well served "buy new hardware along our gear" pushing us for possibly no reason.

I also agree with some of you "if you want vr gear then you can also buy new hardware" Well wrong !. If you are married and not rich you won't dump 600+ for vr gear AND a whole new pc way above average (1000 for cpu/ram/gpu) to play with a mask on the face.

I will definitely and eventually buy htc vive because it will let me move freely in my game room however! till i get it on a reasonable price, i bet more vr gear will provide that goody.

 

[kcarbotte]

I discussed the way steam approached this with some friends. Please read carefully

Steam is selling Steam machines. There is no better and more cheaty way that this benchmark.
They try to keep over 100fps while reducing detail on the fly as the demo goes. That doesn't mean the game won't play in VR gear if you are on low "fidelity" . For me it just means the experience won't be 100% smooth and possibly it will just be low fps (tearing or something similar).
Also i have for example i5.2500k @4.3Ghz. the details on the vr demo showed "no frame lost due to cpu bottleneck". But it also mentions "no cpu loss was calculated for positional calculations" well even cheap smartphone cpus can do it..

For me this is a very well served "buy new hardware along our gear" pushing us for possibly no reason.

I also agree with some of you "if you want vr gear then you can also buy new hardware" Well wrong !. If you are married and not rich you won't dump 600+ for vr gear AND a whole new pc way above average (1000 for cpu/ram/gpu) to play with a mask on the face.

I will definitely and eventually buy htc vive because it will let me move freely in my game room however! till i get it on a reasonable price, i bet more vr gear will provide that goody.

Steam is selling Steam Machines, and NOT A SINGLE Steam Machine will work with the Vive.
Steam Machines run Steam OS, which is not compatible with SteamVR.

Low fidelity is the lowest detail that the SteamVR Benchmark will run at, but the point I made is that developers are not beholden to that. Most demos that I have tested have absolutely no visual options to dial back.

Low frame rate isn't just less than ideal. Low frame rates results in diziness and motion sickness. Anything less than 90 is very noticeable on the Vive.

 

[Math Geek]

i just saw a write up of an amd vr headset that was totally self contained!! all in one vr. specs are a lot lower than what is required with the other units.

could be interesting once it is released anyway http://www.vg247.com/2016/03/15/amd-unveils-sulon-q-all-in-one-ar-vr-headset/

 

[InvalidError]

i just saw a write up of an amd vr headset that was totally self contained!! all in one vr. specs are a lot lower than what is required with the other units.

VR is little more than stereoscopic 3D with motion sensors added. Stereoscopic 3D has been done on the desktop over a decade ago with mixed degrees of success using far less powerful hardware than what is available today. The only reason the specs are "so high" on VR today is because the vendors want to offer reasonably smooth VR experience without sacrificing much detail.

 

[turkey3_scratch]

I discussed the way steam approached this with some friends. Please read carefully

Steam is selling Steam machines. There is no better and more cheaty way that this benchmark.
They try to keep over 100fps while reducing detail on the fly as the demo goes. That doesn't mean the game won't play in VR gear if you are on low "fidelity" . For me it just means the experience won't be 100% smooth and possibly it will just be low fps (tearing or something similar).
Also i have for example i5.2500k @4.3Ghz. the details on the vr demo showed "no frame lost due to cpu bottleneck". But it also mentions "no cpu loss was calculated for positional calculations" well even cheap smartphone cpus can do it..

For me this is a very well served "buy new hardware along our gear" pushing us for possibly no reason.

I also agree with some of you "if you want vr gear then you can also buy new hardware" Well wrong !. If you are married and not rich you won't dump 600+ for vr gear AND a whole new pc way above average (1000 for cpu/ram/gpu) to play with a mask on the face.

I will definitely and eventually buy htc vive because it will let me move freely in my game room however! till i get it on a reasonable price, i bet more vr gear will provide that goody.

Steam is selling Steam Machines, and NOT A SINGLE Steam Machine will work with the Vive.
Steam Machines run Steam OS, which is not compatible with SteamVR.

Low fidelity is the lowest detail that the SteamVR Benchmark will run at, but the point I made is that developers are not beholden to that. Most demos that I have tested have absolutely no visual options to dial back.

Low frame rate isn't just less than ideal. Low frame rates results in diziness and motion sickness. Anything less than 90 is very noticeable on the Vive.

That's why I believe a 980Ti and above is really necessity for a good enough experience.

 

[kcarbotte]

i just saw a write up of an amd vr headset that was totally self contained!! all in one vr. specs are a lot lower than what is required with the other units.

could be interesting once it is released anyway http://www.vg247.com/2016/03/15/amd-unveils-sulon-q-all-in-one-ar-vr-headset/

http://www.tomshardware.com/news/sulon-q-ar-vr-hmd,31406.html

Could definitely be interesting, but don't expect to see Vive and Rift level graphics on that headset.
Our article features a video of the demo the company plans to show this week.

 

[Math Geek]

cool, i'll have to give the article a read. i had not seen that one here yet. figured something was coming though as it always does

everything else i have seen i simply a repeat of the pr release and some quick conjecture. be nice to see a bit more info. i'm still not gonna buy but all in one that still does the job seems like a nice way to go.

 

[beetlejuicegr]

I didn't know that but that doesn't mean it won't become compatible. And i wouldn't be surprised if that was the case for this weird vr demo. I don't what you to see this as trying to protect my post, for me, chances are 50-50 i am right or wrong.

 

[kcarbotte]

I discussed the way steam approached this with some friends. Please read carefully

Steam is selling Steam machines. There is no better and more cheaty way that this benchmark.
They try to keep over 100fps while reducing detail on the fly as the demo goes. That doesn't mean the game won't play in VR gear if you are on low "fidelity" . For me it just means the experience won't be 100% smooth and possibly it will just be low fps (tearing or something similar).
Also i have for example i5.2500k @4.3Ghz. the details on the vr demo showed "no frame lost due to cpu bottleneck". But it also mentions "no cpu loss was calculated for positional calculations" well even cheap smartphone cpus can do it..

For me this is a very well served "buy new hardware along our gear" pushing us for possibly no reason.

I also agree with some of you "if you want vr gear then you can also buy new hardware" Well wrong !. If you are married and not rich you won't dump 600+ for vr gear AND a whole new pc way above average (1000 for cpu/ram/gpu) to play with a mask on the face.

I will definitely and eventually buy htc vive because it will let me move freely in my game room however! till i get it on a reasonable price, i bet more vr gear will provide that goody.

Steam is selling Steam Machines, and NOT A SINGLE Steam Machine will work with the Vive.
Steam Machines run Steam OS, which is not compatible with SteamVR.

Low fidelity is the lowest detail that the SteamVR Benchmark will run at, but the point I made is that developers are not beholden to that. Most demos that I have tested have absolutely no visual options to dial back.

Low frame rate isn't just less than ideal. Low frame rates results in diziness and motion sickness. Anything less than 90 is very noticeable on the Vive.

That's why I believe a 980Ti and above is really necessity for a good enough experience.

So far I'm finding that a GTX 970 is working just fine for practically all of the games. The only ones that haven't been running that well are not planned to be launch titles, so there's time for optimizations.

 

[MTC001]

Why isn't anyone testing with an overclocked GTX970? My GTX970 Windforce from Gigabyte gets to 7.3 without overclock and with OC to 1521MHZ and 3900MHz RAM it reaches 8.1(very high) http://abload.de/img/1521mhz1g7xte.jpg

 

[kcarbotte]

Why isn't anyone testing with an overclocked GTX970? My GTX970 Windforce from Gigabyte gets to 7.3 without overclock and with OC to 1521MHZ and 3900MHz RAM it reaches 8.1(very high) http://abload.de/img/1521mhz1g7xte.jpg

We tested that exact same card. it's in the graph.
6.9 on our system.

 

[Biscuit42]

They rate the lower-end hardware 'capable' to get users to invest in the VR headset. Then, when they're disappointed, they hope the user will shell out for an upgrade to a premium card.

Or am I just an old cynical guy who's seen this way too often?

 

[kcarbotte]

They rate the lower-end hardware 'capable' to get users to invest in the VR headset. Then, when they're disappointed, they hope the user will shell out for an upgrade to a premium card.

Or am I just an old cynical guy who's seen this way too often?

Valve gets no benefit from you buying a premium card, but I agree with the first part.