HTC Vive VR Kit Review: Experiences And Performance
VR GPU Performance
Those recommendations are on-par with what we’d consider to be the bare minimum for a good VR experience with the Vive. We spent some time testing four different graphics cards with the kit, including a GeForce GTX 970, a GTX 980 Ti, a Radeon R9 390, and a Radeon R9 Fury X. With the exception of a few more in-app stutters with the two lower-end cards in more visually demanding applications, the overall experience was quite similar across them all.
Ideally you’ll want to throw as much performance at the Vive as possible, like a 980 Ti or Fury X (or a new GeForce GTX 1080) with a Core i7 and 8GB of RAM or more for example, but even if you only meet the Vive’s minimum spec you should be fine – at least for a while. Once developers start pushing the envelope, you’ll inevitably need more GPU horsepower.
Valve's Steam VR Performance Test
To give you somewhat of an idea as to how the cards compared with an actual benchmark, we used the Steam VR Performance Test utility that’s publicly available for download on Steam. The Steam VR Performance Test measures a system's performance using a 2-minute sequence from Valve’s Aperture Robot Repair VR demo. After running the test, it determines whether your system is capable of properly running VR content at 90Hz and whether the visual fidelity can be increased to the recommended level for a given application. Both a system’s CPU and GPU are factored into the score.
There is a clear difference in each card’s performance according to the tool, but over and above the overall score, one of the more important points is how long each card spent below 90Hz. All four of the cards never dipped below that level in the benchmark -- the run details are posted below. We should mention that our test system consisted of an Intel Core i7-5960X with 16GB of DRR4-2133 memory, an OCZ SSD, and Windows 10 Professional x64.
 Radeon R9 390 |
 GeForce GTX 970 |
 Radeon R9 Fury X |
 GeForce GTX 980 Ti |
Notice how the performance curve gets smoother as the GPU gets faster -- that's what you want. A smooth, even delivery of frames is ideal, not only in terms of the experience, but to minimize the chance of VR-induced "motion sickness".
If you’re wondering why we’re not presenting FRAPS runs from actual apps, it’s because the numbers are mostly meaningless. FRAPS captures data before the GPU performs any asynchronous time warps, which change what you’ll see in the headset. And it also doesn’t account for the latency from when and application makes a draw call to when that frame is actually presented on the Vive’s display.
In short, until we have proven, reliable tools that’ll measure what’s actually important to the VR experience, quantifying “VR performance” is difficult. This stuff is all evolving quickly, though, and better tools should be available soon. Futuremark, Basemark / Crytek, and the GPU makers are all working on tools to help quantify VR performance.
