Nvidia releases FCAT VR performance metrics tool

Performance testing for VR apps is a tricky problem, and one we've not quite found a satisfactory answer for. As it turns out, Nvidia may have the solution. The GeForce guys just made good on that promise from GDC and released FCAT VR as a free download. As most readers of the site have probably already guessed, FCAT VR is a frame-time-based performance capture and analysis tool intended specifically for virtual reality apps.

Nvidia says that FCAT VR taps into its own drivers, Event Tracing for Windows, and the SteamVR APIs to capture performance data for the Oculus Rift and HTC Vive. The company notes that FCAT VR requires no external hardware (save for a VR headset) and should work on "any GPU." The app can distinguish among new frames, dropped frames, and synthesized frames created by technologies like the Rift's Asynchronous Spacewarp. Nvidia says this complete data analysis should illuminate the real performance picture for VR apps on any given system.

Along with the release of the FCAT VR software itself, Nvidia is offering up a fairly extensive guide to walk people through setting up and using the application. The company says that capturing data is fairly straightforward, but that analyzing it is more involved. That's no surprise, as the same could be said for our own methods. If you're eager to start testing your machine's VR performance, you can hit the FCAT download page. Don't miss the guide, either.

Comments closed
    • tsk
    • 3 years ago

    PCper has already tested the 480 vs the 1060 using this tool. The 1060 wins [url<]https://www.pcper.com/reviews/Graphics-Cards/FCAT-VR-Measuring-VR-Performance-GTX-1060-and-RX-480[/url<]

      • toastie
      • 3 years ago

      I don’t think we really needed a VR measuring tool to know how that would turn out. The 1060 has beat the 480 in every other benchmark, and it isn’t close. Why would VR suddenly change that?

        • Goty
        • 3 years ago

        “Every other benchmark” is missing the mark by a wide margin…

        Nvidia’s cards do better in VR environments and it isn’t even close, but the 480 traded blows with the 1060 when the 1060 first came out (the 1060 6GB held a ~4% advantage in 99th-percentile frametimes over the 480 8GB in October, per TR’s testing) and they’re basically on par with one another now in non-VR gaming.

          • renz496
          • 3 years ago

          now funny thing is is still remember some people said and insist nvidia GPU cannot do VR at all because it does not have “LiquidVR”.

      • Bumper
      • 3 years ago

      d’oh. 1060 is like 980 level performance. 480 is like 970 level (or 390ish).

      outside of that I’ll be happy when they release the source code soon, something about a vendor released benchmark feels icky…but I’m sure this information will end up doing good things in the end.

        • Voldenuit
        • 3 years ago

        Ryan at pcper used hardware capture equipment to confirm that the readings provided by FCAT VR were reliable, and they were.

          • Goty
          • 3 years ago

          Yeah, the original FCAT was fine, so I don’t suspect this would be any different.

            • Voldenuit
            • 3 years ago

            It’s definitely good journalistic practice for him to verify the tools that nvidia provided, though.

          • Bumper
          • 3 years ago

          ok. I should have read that article first. well this is super cool then and props to NVidia.

            • Tirk
            • 3 years ago

            Don’t give in that quickly, even the author in the conclusion cautions against overly broad conclusions do to the limited testing,

            “Our first results, though not covering nearly as many games as I would like at this point”

            This capture method, provided by Nvidia, showing Nvidia winning in a metric they wish to focus on should give everyone pause for continued testing to get more of a grasp on the whole picture. Note the 480 performed very closely to the 1060 in Edge of Nowhere which may indicate that a broader test suite could show more of a parity.

            But I guess we could just consign ourselves to focusing on only the metric the company wishes us to. Let’s only ever buy CPUs that top the cinebench multi-threaded tests shall we? As it must be the only true way to see its real performance it all scenarios.

    • TwoEars
    • 3 years ago

    I wonder what that Scott Wasson guy over at AMD thinks of this.

      • tsk
      • 3 years ago

      He’s busy focusing on his vega(n) diet at the moment.

      • the
      • 3 years ago

      He’s busy tiling his basement.

      • tritonus
      • 3 years ago

      He’s busy inside his second big thing.

Pin It on Pinterest

Share This