Oculus Rift VR Benching – AMD vs. NVIDIA – Part 2

Intro

Since BabelTechReviews purchased an Oculus Rift last December, we have been playing 27 VR games using six top NVIDIA and AMD video cards, and we have completed performance benchmarking for Showdown, Alice VR, Batman Arkham VR, Chronos, DiRT Rally, Robo Recall, and Serious Sam: The Last Hope. We posted our original evaluation a few months ago and since then, we have been exploring ways to accurately benchmark VR games.  PC benchmarking methods such as using Fraps do not accurately convey the performance that the user actually experiences in the HMD (Head Mounted Display), and until recently, there were no logging tools available for the Oculus Rift as there are for the HTC Vive.

Fortunately, NVIDIA has recently released FCAT VR as an accurate VR benching tool that we will compare with our own video benchmarking which uses a camera to capture images directly from a Rift HMD lens.  For each of our six test video cards, we will use Showdown to compare NVIDIA’s FCAT-generated frametime graphs and performance results directly with the Oculus Performance DeBug tool overlay running in real time in our videos.  We will also benchmark 5 other games using the GTX 1080 Ti FE, the GTX 1080 FE, the GTX 1070 FE, a EVGA GTX 1060 SC, the Fury X, and a Gigabyte G1-RX 480 at RX 580 clocks with our Core i7-6700K at 4.0GHz where all 4 cores turbo to 4.6GHz, an ASRock Z7170 motherboard and 16GB of Kingston HyperX DDR4 at 3333MHz.

In part 1 of our evaluation, we gave an overview of Virtual Reality (VR) and looked at the complexities of Oculus Rift VR benching.  We used the few synthetic benchmarking tools then available to roughly compare video card performance of nine AMD and NVIDIA video cards.

In this part 2, we will go into more depth as we attempt to perform accurate video VR benchmarking, and by using NVIDIA’s FCAT VR, we believe that we will be able to convey to BTR’s readers an accurate sense of what is actually happening in the HMD while VR benchmarks are being run.  And the FCAT generated charts will be clearly able to show the performance difference between competing video cards.

Until NVIDIA released FCAT VR in March, there was no universally acknowledged way to accurately benchmark the Oculus Rift as there are no SDK logging tools available.  To compound the difficulties of benchmarking the Rift, there are additional complexities because of the way it uses a type of frame reprojection called asynchronous space warp (ASW) to keep framerates steady at either 90 FPS or at 45 FPS.

Both NVIDIA and AMD agree that Fraps benchmarking only provides a slight correlation between what the user actually experiences and what is displayed by the Windows desktop.  Fraps only measures the framerate of the desktop window, not what appears in the HMD so the performance that a user sees in the desktop window is not a valid proxy for the experience in the HMD.  OCAT, Performance Mon, Afterburner, and other PC benching software do not accurately report nor measure what the user experiences in the HMD.   And it is important to be aware of VR performance since poorly delivered frames will actually make a VR experience quite unpleasant and the user can even become VR sick.

VR sickness is similar to motion sickness with a sick sensation in ones stomach or with a headache or both. Common symptoms are general discomfort, headache, stomach upset, nausea, fatigue, drowsiness, and disorientation. If a player gets VR sick, they may have to quit playing in VR for the rest of the day, or it may take less time to recover depending on the person. Taking a nap often helps but extreme sickness can even lead to vomiting.

Let’s take a close look at NVIDIA’s VRWorks as well as at AMD’s LiquidVR, and then at our own video benching before we look at our 6-pack of video cards performance with six VR games plus Showdown using FCAT VR and video benchmarking.