NVIDIA Releases Advanced VR Benchmarking Tool ‘FCAT VR’

At a press and analyst briefing held a couple of weeks ago in San Francisco, NVIDIA talked at length about some of its latest technologies, and in particular, what exactly it is that makes the company’s brand-new GeForce GTX 1080 Ti so damn powerful. Beyond that, we were also shown some development tools, as well as a benchmarking tool that anyone using VR can take advantage of.

That tool is called FCAT VR, and it’s the brainchild of NVIDIA’s Tom Peterson (the man behind G-SYNC and a number of other green team technologies). Like the company’s regular FCAT tool, for monitoring the performance of 2D content, FCAT VR aims to prove to the world which GPU actually earns the “Ultimate” moniker for overall performance – that is, performance with zero dropped frames.

Losing frames in 2D content is bad, but it’s even worse in VR, where your entire field of vision is of gameplay. If there’s any slowdown, it won’t bode well. If the issue is bad enough, you could begin to feel nauseous, which is pretty much the opposite feeling you should have from playing a fun game.

FCAT VR is made up of a couple of components. For starters, you need a VR headset to make use of the software, as the feed is directly monitored. That means that all benchmarking will be manually conducted – no timedemos here. That in turns means that there could be slight variance from run to run, but if the same content is tested with all of the time, it shouldn’t be that hard to produce comparable results.

If you have deep pockets, you can use a hardware version of FCAT VR that will allow you to capture your video in a lossless format for later inspection. NVIDIA is not selling the hardware; instead, the software just makes use of other vendor hardware to get the job done. However, we were told that the solution NVIDIA itself used for testing cost about $2,000, so it’s not for everyone. It’s not expensive for the sake of being expensive, either – it’s expensive because the capture card needs to be able to capture two video feeds at once, at 90 FPS each.

For the vast majority of FCAT VR users, the software component will be opted for instead. In talking to NVIDIA, I asked whether there’d be any difference in the quality of the results between the hardware and software solutions, and I was told no. The hardware solution would just capture lossless video that could be useful to actual content producers, or the makers of the VR kits.

It’d help to look at a result example to help explain this better:

In this particular test, NVIDIA is showing how use of its MRS Multi-Res Shading technology could improve VR performance without impacting the IQ (it renders only the important bits that your eye will see). With no MRS, the game drops about 50% of its frames, on a GTX 1080. As different levels of MRS are introduced, the number of dropped frames plummets.

Ideally, your benchmark result would only ever show a solid green line, with no dropped frames whatsoever. However in this particular test, the game’s Ultra detail settings were used, which is why the GTX 1080 still can’t guarantee no dropped frames. However, that result is still quite good, so some might still want to game with it. Hitting 0 dropped frames with top quality in certain games is going to be tough. So if you don’t think VR content “looks that good” right now, bear these kinds of results in mind. To have 4K per eye, we’re going to require some serious GPU horsepower.

We’re going to take a deeper look at FCAT VR in the near-future, so stay tuned. If you want to give the solution a try for yourself, hit up NVIDIA’s blog, but be warned: this is not for the feint of heart. It’s not hard to setup, per se, but does require a lot of time and reading. It should prove worth it if you want to get to the bottom of your VR performance.