5 ms·

For the people worrying about GPU usage, it also says "Foveated driving logic for VR and AR applications was implemented".

by Jyaif 9y ago

egeozcan 9y ago
I had to google foveated driving logic: It means tracking the user’s eye movements and only rendering the most central area to the eye in high-definition. Source: https://www.roadtovr.com/google-shares-new-research-foveated-rendering-techniques/ https://www.roadtovr.com/google-shares-new-research-foveated...
- kyrra 9y ago
  Real source: https://research.googleblog.com/2017/12/introducing-new-foveation-pipeline-for.html https://research.googleblog.com/2017/12/introducing-new-fove... (the one you linked came up first for me too when searching around, took a little more digging to find the original article).
- boobsbr 9y ago
  That sounds really nice.
- mistercow 9y ago
  Huh, I've been wondering for like a decade if we'd start doing this, but the eye-tracking latency must have to be insane.
- mnx 9y ago
  Is this foveated as in Field-Of-View-ated?
  - gordo4 9y ago
    https://en.wikipedia.org/wiki/Fovea_centralis https://en.wikipedia.org/wiki/Fovea_centralis
    - mnx 9y ago
      thanks!
  - vokep 9y ago
    In laymen terms, sure! Its a useful enough way to think about it. Technically though the fov is refering to fovea, not field of view.
petermcneeley 9y ago
Ya this is effectively a non issue. https://www.roadtovr.com/nvidia-perceptually-based-foveated-rendering-research/ https://www.roadtovr.com/nvidia-perceptually-based-foveated-... The primary issues in this space is latency not raw power.
- mncharity 9y ago
  > Ya this is effectively a non issue. Yeah. If the new panel is presented as 3x or 4x 1080p (two foveal insets and one or two backgrounds), and tolerates 30 fps, I could run it off the old integrated graphics I'm typing at now. > The primary issues in this space is latency For immersive VR, yes. But as one moves away from gaming, and immersive, and VR, design constraints relax a lot. There's so much ambient confusion about the shape of the broader design space, and misattribution of constraints. I usually use my WMR (and Vive before that) on an old laptop with Intel integrated graphics. A duct-taped-on 30 fps camera-passthrough AR serves for balance, permitting <30 fps rendering for a software development 3D "desktop". Latency, and its variance, is just not a challenge here. I can even do subpixel rendering on the WMR. Lens focus is so bad, subpixels are only worth bothering with for like (0.5 kP)^2/eye (of the ~1 kP square/eye usable area, of the 1.4 kP square/eye panel). There's no point in even rendering full native resolution over much of the visual field. And not caring about immersion, there's no need for barrel and chromatic correction, nor for blending away seams between resolutions. Such a vast amount of effort is spent on dealing with "horrible immersion-breaking visual artifacts", which for some parts of design space, are like "meh, sooo don't care". Remember the skeuomorphic user interface fad of a decade ago? A calender app should look like a leather-bound calendar book? And... remember all the claims that calendar apps wouldn't be viable for years, because of the GPU demands of faithfully rendering light reflection off the leather. Yeah, me neither. That would have been silly. Just like expecting a 3D/HMD professional "desktop" environment to prioritize immersion.
mncharity 9y ago
Here's an optical illusion to show how small your fovea's FoV is: https://www.shadertoy.com/view/4dsXzM https://www.shadertoy.com/view/4dsXzM

For the people worrying about GPU usage, it also says "Foveated driving logic for VR and AR applications was implemented".

by Jyaif 9y ago

For the people worrying about GPU usage, it also says "Foveated driving logic for VR and AR applications was implemented".

egeozcan 9y ago

I had to google foveated driving logic: It means tracking the user’s eye movements and only rendering the most central area to the eye in high-definition. Source: https://www.roadtovr.com/google-shares-new-research-foveated-rendering-techniques/ https://www.roadtovr.com/google-shares-new-research-foveated...

kyrra 9y ago
Real source: https://research.googleblog.com/2017/12/introducing-new-foveation-pipeline-for.html https://research.googleblog.com/2017/12/introducing-new-fove... (the one you linked came up first for me too when searching around, took a little more digging to find the original article).
boobsbr 9y ago
That sounds really nice.
mistercow 9y ago
Huh, I've been wondering for like a decade if we'd start doing this, but the eye-tracking latency must have to be insane.
mnx 9y ago
Is this foveated as in Field-Of-View-ated?
- gordo4 9y ago
  https://en.wikipedia.org/wiki/Fovea_centralis https://en.wikipedia.org/wiki/Fovea_centralis
  - mnx 9y ago
    thanks!
- vokep 9y ago
  In laymen terms, sure! Its a useful enough way to think about it. Technically though the fov is refering to fovea, not field of view.

petermcneeley 9y ago

Ya this is effectively a non issue. https://www.roadtovr.com/nvidia-perceptually-based-foveated-rendering-research/ https://www.roadtovr.com/nvidia-perceptually-based-foveated-... The primary issues in this space is latency not raw power.

mncharity 9y ago
> Ya this is effectively a non issue. Yeah. If the new panel is presented as 3x or 4x 1080p (two foveal insets and one or two backgrounds), and tolerates 30 fps, I could run it off the old integrated graphics I'm typing at now. > The primary issues in this space is latency For immersive VR, yes. But as one moves away from gaming, and immersive, and VR, design constraints relax a lot. There's so much ambient confusion about the shape of the broader design space, and misattribution of constraints. I usually use my WMR (and Vive before that) on an old laptop with Intel integrated graphics. A duct-taped-on 30 fps camera-passthrough AR serves for balance, permitting <30 fps rendering for a software development 3D "desktop". Latency, and its variance, is just not a challenge here. I can even do subpixel rendering on the WMR. Lens focus is so bad, subpixels are only worth bothering with for like (0.5 kP)^2/eye (of the ~1 kP square/eye usable area, of the 1.4 kP square/eye panel). There's no point in even rendering full native resolution over much of the visual field. And not caring about immersion, there's no need for barrel and chromatic correction, nor for blending away seams between resolutions. Such a vast amount of effort is spent on dealing with "horrible immersion-breaking visual artifacts", which for some parts of design space, are like "meh, sooo don't care". Remember the skeuomorphic user interface fad of a decade ago? A calender app should look like a leather-bound calendar book? And... remember all the claims that calendar apps wouldn't be viable for years, because of the GPU demands of faithfully rendering light reflection off the leather. Yeah, me neither. That would have been silly. Just like expecting a 3D/HMD professional "desktop" environment to prioritize immersion.

mncharity 9y ago

Here's an optical illusion to show how small your fovea's FoV is: https://www.shadertoy.com/view/4dsXzM https://www.shadertoy.com/view/4dsXzM