SteamVR Performance Test

CAT-THE-FIFTH · 22 Feb 2016 at 23:10

Seems a bit of a stupid test if it does not work with SLI or XFire which does not use AFR in VR,since it dedicates one card to each display.

So realistically its not really that great a test,unless OFC SteamVR won't support dual cards which is dumb.

stephenb · 22 Feb 2016 at 23:15

Interesting to see how much further ahead 970's are then my 780, both here and elsewhere. 1254 is a pretty decent clock for a 780, with lower clocks than that I beat the GPU score in 3D Mark that was part of the Oculus recommended spec they put up. Must be taking advantage of some architectural advances in Kepler or Nvidia drivers are better optimised for VR+900 series.

SaintJimmy · 22 Feb 2016 at 23:15

Looking good...

r7slayer · 22 Feb 2016 at 23:23

Looks like im set then

shankly1985 · 22 Feb 2016 at 23:24

Lokken86 · 22 Feb 2016 at 23:27

CAT-THE-FIFTH said:
So basically XFire does not work for this test??

WTF??

If you read this Anandtech article,it says VR does not AFR and would mean one card per display,so there should be no scaling issues:

http://www.anandtech.com/show/9874/amd-dual-fiji-gemini-video-card-delayed-to-2016

I only have 1 card.

On a side note this was 15.7.1 drivers. Someone suggested using crimson for best results so Il give that a go later.

CAT-THE-FIFTH · 22 Feb 2016 at 23:34

Lokken86 said:
I only have 1 card.

On a side note this was 15.7.1 drivers. Someone suggested using crimson for best results so Il give that a go later.

Its more the case Fury Nano XFire scores around 10 and the Fury Nano is around R9 Fury speed so it means XFire does not work.

SLI for Kepler seems borked too with the GTX780 SLI setup doing worse than a single GTX780.

So it either means its a pointless benchmark as BOTH Nvidia and AMD have been selling dual cards for VR as it means one display can be handled by each card or SteamVR cannot handle dual cards.

If that is the case,and the OR can support dual cards,then its a major problem for the Vive if its relying on SteamVR.

james.miller · 22 Feb 2016 at 23:34

Struggling to see how that would correlate to real world performance but there it is.

millermilla · 22 Feb 2016 at 23:46

i5 6400@ 4.5
Single GTX 960 1438 / mem 1975
8mb RAM

I will be interested to see what the real world performance is on a budget setup?

VR%20High%20setting%204.5%20OC%20Stable.jpg

triss · 23 Feb 2016 at 01:22

stephenb said:
GTX780 will pass, you just need to make sure it is suitably clocked. Interesting to note that while I can pass with both SLI enabled and disabled I get a lot more frames with it disabled. Framerate throughout the test being higher and memory usage lower.

Thanks had forgotten to clock mine, interestingly memory clock seems to help this a lot
Seems my 780 is ready just

http://imgur.com/O4H8Mdg

Lokken86 · 23 Feb 2016 at 01:25

triss said:
[/SPOILER]

Thanks had forgotten to clock mine, interestingly memory clock seems to help this a lot
Seems my 780 is ready just
http://imgur.com/O4H8Mdg

Result.

CAT-THE-FIFTH · 23 Feb 2016 at 01:51

millermilla said:
i5 6400@ 4.5
Single GTX 960 1438 / mem 1975
8mb RAM

I will be interested to see what the real world performance is on a budget setup?

I got the same card as you but run a Xeon E3 1230 V2/Core i7 3770.

Stock Results.

Having said that the score was 2.9 when I benched the card from cold.

Results with GTX960 overclocked to around 1.45GHZ and aroubd 1.973GHZ for the VRAM

There seems to be nearly a 50% increase in score overclocking the core around 125MHZ more and the VRAM around 450MHZ more!!

Interestingly you seem to be scoring nearly 19% more than me,which interestingly is around the same as the clockspeed difference between our CPUs.

I wonder if the benchmark is single threaded limited or not??

Klunt Bumskrint · 23 Feb 2016 at 06:51

Nvidia keep banging on about multi GPU setups for best results in VR but lately they've really been neglecting SLI users.

I really hope they pull their thumb out of their bum and focus on it!

pgi947 · 23 Feb 2016 at 07:43

10.8, stock 980Ti, frames would have been capped at 120. Shame I'll likely not be going near VR at its current price point.

eddyr · 23 Feb 2016 at 08:24

CAT-THE-FIFTH said:
I got the same card as you but run a Xeon E3 1230 V2/Core i7 3770.

Stock Results.

Having said that the score was 2.9 when I benched the card from cold.

Results with GTX960 overclocked to around 1.45GHZ and aroubd 1.973GHZ for the VRAM

There seems to be nearly a 50% increase in score overclocking the core around 125MHZ more and the VRAM around 450MHZ more!!

Interestingly you seem to be scoring nearly 19% more than me,which interestingly is around the same as the clockspeed difference between our CPUs.

I wonder if the benchmark is single threaded limited or not??

Watching the benchmark it appears to be turning up/down settings on the fly and seems (with my hd7950) to want to sustain highest graphical quality at ~105 fps. Score seems significantly weighted on above medium capability at over 90fps.

Incidentally can just squeeze a 4 score @ 1300/1500 on my 7950 with [email protected].

stephenb · 23 Feb 2016 at 09:33

CAT-THE-FIFTH said:
Its more the case Fury Nano XFire scores around 10 and the Fury Nano is around R9 Fury speed so it means XFire does not work.

SLI for Kepler seems borked too with the GTX780 SLI setup doing worse than a single GTX780.

So it either means its a pointless benchmark as BOTH Nvidia and AMD have been selling dual cards for VR as it means one display can be handled by each card or SteamVR cannot handle dual cards.

If that is the case,and the OR can support dual cards,then its a major problem for the Vive if its relying on SteamVR.

Developers need to implement dual GPU support directly into the game engine, most likely using LiquidVR/GameworksVR if it's based on DX11. For DX12 they should be able to do this natively so to speak. This test is based on the Source2 engine? Valve obviously haven't implemented support for multi GPU setups, maybe someone should ask them if they will.

stephenb · 23 Feb 2016 at 09:45

Some useful info here: https://www.reddit.com/r/Vive/comments/46yzeb/steamvr_performance_test_on_steam/d095k41

Looks like I may need to retract my previous comments. Try -multigpu as a launch option if you have 2 or more GPUs. Anyone able to test this out and see if there is an improvement?

CAT-THE-FIFTH · 23 Feb 2016 at 09:49

eddyr said:
Watching the benchmark it appears to be turning up/down settings on the fly and seems (with my hd7950) to want to sustain highest graphical quality at ~105 fps. Score seems significantly weighted on above medium capability at over 90fps.

Incidentally can just squeeze a 4 score @ 1300/1500 on my 7950 with [email protected].

It seems weird that I get a much lower score as their are people Phenom ii and Q6600 CPUs getting with a few percent of those with faster Core i7 CPUs using a GTX970.

Thunnylaf · 23 Feb 2016 at 09:55

i7 [email protected]
gtx980sc (factory defaults)

Cooper · 23 Feb 2016 at 09:56

stephenb said:
Looks like I may need to retract my previous comments. Try -multigpu as a launch option if you have 2 or more GPUs. Anyone able to test this out and see if there is an improvement?

It does work (-multigpu) but weirdly my score doesn't increase. :confused:

in single gpu mode my 970 is at 99%, with -multigpu, its roughly 50% on each card.

I get this kind of score whether in single or multi gpu mode: