NVIDIA is upgrading its Variable Rate Supersampling (VRSS) with support for headsets with eye-tracking, allowing the rendered application to improve performance by increasing quality where the user is looking, while decreasing it elsewhere.

Nvidia today announced the latest version of VRSS, a foveated rendering implementation that works with any of the company’s RTX series GPUs and any application which supports DirectX 11, forward rendering, and MSAA.

The first version of VRSS only offered static foveated rendering which increased the effective resolution at the center of the image (where the lens is the sharpest), while decreasing the quality outside of the central area, effectively concentrating the GPU power where it matters most. The foveated rendering region can be supersampled up to 8x.

Image courtesy NVIDIA

VRSS 2 adds support for dynamic foveated rendering which allows the system to move the supersampled area to wherever the eye is looking. Although lens sharpness drops off as the user looks away from the center of the lens, there can still be perceptual benefits to supersampling outside the center of the lens.

This of course only works for headsets equipped with eye-tracking tracking, which is not common in consumer-grade VR headsets today, but is expected to become more widespread in the future.

Out of the gate, Nvidia says that the dynamic foveated rendering in VRSS 2 will support HP’s new Reverb G2 Omnicept Edition headset. In the future we hope to see support added for HTC’s Vive Pro Eye and Varjo headsets, both of which include eye-tracking hardware.

VRSS 2 is supported as of GeForce driver version R465 which became available on March 30th. Users must enable VRSS via the Nvidia Control Panel (Manage 3D Settings > Global Settings > Virtual Reality – Variable Rate Supersampling > Adaptive).

SEE ALSO
'Moss' 1 & 2 Get a Visual Overhaul and 25% Discount in Time for Quest 3S Launch

Although eye-tracking headsets themselves appear to require per-headset integrations to support dynamic foveated rendering with VRSS 2, Nvidia says that applications don’t need to be modified in any way to get the benefits of VRSS 2, provided they support DirectX 11, forward rendering, and MSAA. That’s a good thing because it means developers don’t need to rely on any technology that’s specific to Nvidia GPUs in order to benefit from VRSS 2.

Developers with compatible titles need only to submit their application to Nvidia for consideration. If the application benefits from VRSS 2, Nvidia will whitelist the app to use VRSS 2 in a future driver update.

Nvidia today also published a new list with all games currently supporting VRSS:

Games Supporting NVIDIA VRSS – April 12th, 2021
Battlewake Raw Data
Boneworks Rec Room
Budget Cuts 2: Mission Insolvency Rick & Morty: Virtual Rick-ality
Doctor Who: The Edge of Time Robo Recall
Eternity Warriors VR Sairento VR
Hot Dogs, Horeshoes, & Hand Grenades Serious Sam VR: The Last Hope
In Death Skeet: VR Target Shooting
Job Simulator Sniper Elite VR
Killing Floor: Incursion Space Pirate Trainer
L.A. Noire VR Special Force VR: Infinity War
Lone Echo Spiderman Far From Home
Medal of Honor: Above and Beyond Spiderman Homecoming VR
Mercenary 2: Silicon Rising Talos Principle VR
Onward VR The Soulkeeper VR
Pavlov VR The Walking Dead: Saints & Sinners
PokerStars VR VRChat
Newsletter graphic

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. More information.


Ben Lang
Ben is the world's most senior professional analyst solely dedicated to the XR industry, having founded Road to VR in 2011—a year before the Oculus Kickstarter sparked a resurgence that led to the modern XR landscape. He has authored more than 3,000 articles chronicling the evolution of the XR industry over more than a decade. With that unique perspective, Ben has been consistently recognized as one of the most influential voices in XR, giving keynotes and joining panel and podcast discussions at key industry events. He is a self-described "journalist and analyst, not evangelist."
  • TechPassion

    Do you think the VR user will not notice something is changing or is in worse quality/blured on the edge of his eye focus area? I think this foveated rendering is a total bs. Not needed at all.

    • Hacker4748

      Do you think the VR user will not notice something is changing or is in worse quality/blured on the edge of his eye focus area?

      We actually know that with absolute certainty. https://uploads.disquscdn.com/images/e4520a151f176cabf461b4f2f70491af33c18018e3fba533e4dee2615849d5c6.jpg

      • JDawg

        That pic is magic

      • Tabp

        Witchcraft!

      • nejihiashi88

        best counter argument xD

      • JrSpaceMan

        Perfect example, well done!

    • benz145

      It depends on the specific implementation. If you go from 1x sampling across the whole frame and 8x sampling in the middle, users won’t see any difference in the periphery compared to if they weren’t using VRSS at all.

      Ostensibly it’s also possible to go to less than 1x sampling outside the middle which could introduce visible artifacts compared to not using VRSS.

      It depends on the content and the headset as well. Ideally you want the falloff to be equal to lens blur so that you benefit from reduced shading outside the center without any perceptual differences. This is difficult to tune however as each headset has different lens characteristics and everyone has different vision.

      Eye-tracking makes things a bit easier so it’s nice they are adding it.

      • Chris Omega

        You won’t notice less than 1x sampling outside of your fovea.

        The whole point of foveated rendering is that you are unable to notice the quality loss outside of your fovea.

        • benz145

          Yes, that’s the theory. In reality it’s actually very difficult to reduce quality without introducing perceptual artifacts. Here’s a pretty interesting article explaining some facets of this: https://www.roadtovr.com/nvidia-perceptually-based-foveated-rendering-research/

          I’ve seen many foveated rendering approaches with eye-tracking which are not fast or accurate enough to be invisible. I’ve seen a few which have nailed it though.

      • Adrian Meredith

        This is why a future version that uses dlss in addition to DFR will be groundbreaking. VR Games could potentially leap a whole generation (or two) graphically above flat screen games

        • Charles

          I’d say it would put VR right at the same level as the best of the latest-generation, assuming you want the retinal area to be about the density of a 1080p screen viewed from a distance.

          Which is much better than how it currently is, with PS2 graphics being common. The standard currently seems to be PS3-level graphics.

    • xyzs

      The figues are not precise but like 95 percent of your perception sits inside a 5 percent zone in the center of your view (the fovea).

      You think it’s bullshit to skip the heavy high quality rendering in the zones where you brain and eyes have such low perception that you are almost blind ?…

      No, it’s the best optimization that can happen to VR and this is thanks to humain vision limits.

      In the futur, I think this can be optimized further by applying several progressive levels of reduced quality rendering around the fovea instead of just 2 with a sharp transition like in the current examples.

      • Fovea is just 2 degrees of visual field, only occupies 1% of retinal area, but wired to 50% of visual cortex in brain.

    • Tabp

      If you don’t like it you can just wait 30 years for GPU technology to get good enough. Right now we have terrible resolution, terrible framerates, terrible field of view, terrible content detail, terrible hardware costs, etc. Not having foveated rendering means having big problems somewhere else.

      • mellott124

        I guess it depends on your point of view. It’s all amazing to me where we’re at today with VR. Is it all perfect… not at all. But terrible, I don’t think so.

    • Sofian

  • Have they fixed big stutters when vrss is enabled? I guess not. They hyped this tech to the oblivion and everyone forgot about it.

    • doug

      Seems so. I just ran Talos and Rick and Morty, no stutters. Although, I can’t be 100% sure it was active, since there’s no affirmative indicator.

  • Byaaface

    The technology is cool, but honestly having half the screen blue and the surrounding be tinted red is just going to be distracting. I wish they could just make the tint more subtle, but otherwise really think the blue/red will really help with the 3D effect!

    • JDawg

      Is this a joke or are you serious?

    • Bob

      I’m sorry to inform you but VRSS isn’t a technology. It is actually a game which you can launch through the Nvidia control panel! :)

    • xyzs

      Just please admit you are joking… for you own sake

      • Gabriele Pratticò

        to troll or to be banned, this is the problem :D

    • this is hilarious XD

    • david vincent

      Haha, this joke is as old as foveated rendering :)

    • Jim P

      It’s Virtual boy 2.

  • Ad

    I don’t think eye tracking is remotely worth it. I feel like we could get most of what we want from DLSS.

    • Sofian

      Actually we need both eye tracking and deep learning image reconstruction.
      Look for facebook s research on “deepfovea”.

      • Charles

        Put both together and you can do PS5 graphics with retinal resolution (at least before being downscaled to your headset’s limitations.

    • Martin355

      DLSS (and TAA in general) tends to look worse in VR than in regular flatscreen games. I guess people notice the blurring more when the display is 3D and covers your field of vision?

    • JrSpaceMan

      It allows better quality using less physical technology, creating less heat while using less power. This all equals smaller headsets running on less powerful computers and will allowing wireless headset batteries to last longer while providing a better quality image. This is another game changer that VR needs to bring more people in by ultimately lowering the price to entry. The tech will take time to mature but it’s absolutely worth exploring. Who knows what this will lead to next. Eye tracking has all sorts of other benefits within VR, including emulation and projection into the game that creates a more human interaction. The tech marches on…

      • Ad

        “Eye tracking equals less compute needed so smaller headsets” is exactly the kind of bandwagon and largely incorrect thinking that has led to so much wasted time and opportunities.

        • JrSpaceMan

          I think the companies spending money in R&D have come to a different conclusion.

  • This of course only works for headsets equipped with eye-tracking tracking,

    • Charles

      “Eye-tracker tracking”? So it tracks the eye-tracker?

      Personally, I’m holding out for eye-tracker-tracker tracking.

      • Martin355

        Yeah, it feels unsafe not to know where my eye tracker tracker is. Who knows what it’s up to?

  • nejihiashi88

    this with DLSS will provide at least %80 performance boost so a gpu with 20 tflops will preform like a gpu with 36 tflops, a huge advantage over flat screens even if we dont consider DLSS it will be at least 40% with foveated rendering.

  • Amazing piece of news!

  • Schadows

    VRSS (Variable Rate Super Sampling), be it v1 or v2, doesn’t decrease “the quality outside of the central area”. It just increase the quality inside the central area while leaving the rest alone.

    VRS (Variable Rate Shading) is the tech decreasing the quality in specific area (which could be set as the outside of the central area).