SteamVR Performance Test

  • Thread starter Deleted member 13524
  • Start date
D

Deleted member 13524

Guest
There's a performance test for SteamVR.
It's listed as a "tool" for Steam so it doesn't show in the front page or anything, so you can just follow these links:

https://steamdb.info/app/323910/
steam://install/323910


It runs a windowed VR demo with a robot from Portal being repaired:

SxuEpTi.jpg



The result I initially got, with a Core i7 4820K and two R9 290X:

Wonm1A2.png


But I noticed in MSI Afterburner that only one GPU was being used, and I found that if we put "-multigpu" in the launch options, the test will use both GPUs. The result for that was:

I2uXze9.png


Which is the highest possible score :)
It seems the multigpu option isn't turned on by default because they claim most games don't support it yet. Let's hope that by H2 2016 this will be different.
 
Test seems to favor Nvidia , my r290 gets me a 5.8 with a i7 3820@ 3.6ghz
Could it be that your v-sync is forced on in the drivers? I had mine in both tests, and only remembered it after I took both screenshots and uninstalled the tool.
 
If a 2500k and a 970 picked up a 6.7 score, I'm gonna come in here later tonight and make AMD owners angry when I unleash my 4.5GHz 3930k and the overclocked 980Ti G1 :D Seriously though, if I remember to, I'll post in here later to see how it ranks.
 
Could it be that your v-sync is forced on in the drivers? I had mine in both tests, and only remembered it after I took both screenshots and uninstalled the tool.

thanks for the suggestion but according to the radeon settings its on off unless application specifies . It seems to be common that the r290 is in the mid to high 5 and the x model sometimes makes it into the low 6s.

I'm sure I will be fine. I can play things with similar frame rates to my friends with a GeForce 970 and 980
 
LoL, 980Ti is a flatline of 11... Somehow it was CPU bound for a single frame though. I even retried the test, and a single frame still showed up. Not sure what that's about...
 

Attachments

  • Screencap.png
    Screencap.png
    30.2 KB · Views: 40
I ran the test in interactive mode and was impressed by the fine geometry and texture detail, though shading is nothing to write about. I can't notice any ambient occlusion being used.

Here's an example of dynamic asset detail adjustment from the test, noticeable on slower systems, triggered by a frametime threshold:

 
Another 980Ti here. I don't get the single CPU bound frame (5820K @ 4.4Ghz) but I do only get a 10.5 score. My GPU is at reference clocks though.

@albuquerque what are your gpu/mem speeds at? I'd like to OC temporarily to see if I can match your flat 11 result so I can see if it's that or perhaps some background process causing it. Also, you seem to have 13k frames compared to everyone elses 11k or less.

4S787hJ.png
 
Last edited by a moderator:
I'm guessing the higher performance allows the system to render more frames in the allotted time? I'm guessing. Ran it this time and there were no CPU-bound frames, and also notice I squeezed a few more frames into the render time (13334 compared to 13245 last time...)

Ran it with MSI Afterburner around so you can see the clockspeeds. Big picture attached! Overclock is pretty conservative, 1465MHz core and ~3.6GHz (7.2GHz effective) on the memory. I can give the memory quite a bit more, but it doesn't seem to help much and in very rare cases it ends up making the display driver reset.
 

Attachments

  • BigScreencap.png
    BigScreencap.png
    1.8 MB · Views: 20
Thanks. I matched your settings apart from power limit which I couldn't set to more than 110. Monitoring power it reached 99% once. Only got a 10.7 with a bit over 11k frames. Is the higher power limit something you tweaked in afterburner or a feature of your card?

...and cheers for the edit @BRiT :p
 
If I'm not mistaken, the higher power limit is something unique to the firmware in the Gigabyte G1 card; I'm reasonably confident there are other high-end 980Ti OC cards with similar capabilities. Afterburner simply reads the cap from your firmware and provides the slider; if you cannot go above 110, then it's a firmware limitation.
 
No CPU bound frames here on a stock 2500K. My 670 is woefully inadequate though apparently. Scored at the bottom end of not ready.
 
Strange.... not the overall result but the fidelity graph. I've run it twice just to make sure, and the graph was the same on both runs. Nice stable graph though :)

2016-02-25.png
 
Back
Top