Eyetracking When?

[wilcovertegaal]

@mirage335

Cost of foveated rendering is on the CPU single-thread side .

Unless AMD/nVidia implement GPU hardware support …

We’re going in circles.

You seem to insist that there can be only one purpose for eye tracking. I think there are more and I would welcome the possibility to explore them all.

[mirage335]

@PimaxUSA I would be extremely curious to see how eye tracking foveated rendering works on DCS World with my (relatively CPU intense) use cases on the best available hardware (which I have). Would not take long to test.

[noro]

Any clue if eye-tracked FV needs RTX features? like FFV where RTX cards are mandatory? (got an 1080ti)

[Schadows]

@Cdaked

Didn’t Pimax say that the module achieved a minimum of 25% more performance? Which I think, for Pimax’s FOV, is very little.
For example, Oculus is trying to increase graphics card performance by 67% using eye tracking and AI.

There is not so much you can do in the rendering pipeline at Pimax level. For impressive gains, and possibly less annoying issues (flickers and such), it will need to be implemented at the game level (and will need some standardized features for that).
Nvidia has worked tremendously in that direction for years, but without much compatible hardware out there, there been no incentive spend time (and money) to implement fovated rendering directly in games for now.

@mirage335

By the time eye tracking foveated rendering can be implemented, even better graphics hardware will be available.

You probably didn’t mean it that way, but it sounds like you believe future hardware will solve anything performance-wise. It won’t. Just look at what is happening with normal games, which requires better hardware every 2-3 year to take advantage of their beauty.
Same with VR, with more powerful hardware, will come more ambitious games, and the need to work around these new limitations.

@noro

Any clue if eye-tracked FV needs RTX features? like FFV where RTX cards are mandatory? (got an 1080ti)

I would be surprised if it were to be compatible with non RTX cards, at least for games that didn’t implemented it themselves.
FFR was supposed to arrive on GTX 10x0 cards, probably by thinking using the Multi Res Shading feature, but seemed to have been too hard (or not giving satisfying gains) in the end.
VRS, as implemented in RTX cards, is fully dynamic, and the “mask” can be updated frame by frame if necessary, but I see no features which could be dynamic in non RTX cards so far.

[Cdaked]

@Schadows

but without much compatible hardware out there, there been no incentive spend time (and money) to implement fovated rendering directly in games for now.

Fortunately, that’s changing.

[Heliosurge]

Also not entirely true. FoveVR already had some of this support long ago but couldn’t gain traction with their own headset.

Oculus already supports agnostic gpu FFR techniques in their sdk for game development. Headset driver implementation of course will not be as good as game direct support as even Nvidia had demonstrated it’s gpu implementation on games is hit and miss.

[mirage335]

The existing combination of CPU/GPU hardware out there now has hard limits which almost no amount of money can overcome. VR apps are often limited to one core, one GPU, in the most intolerable situations, leaving the thinnest possible margins to continue extracting adequate performance out of with any available hacks.

Unlike the ongoing desire for shiny graphics, this is a case of ‘completely broken’, not ‘room for improvement’.

[Heliosurge]

Yes those old programs that are stuck with unoptimized code will eventually be replaced with programs that utilized modern computers.

The limitations are based around shortsightedness.

[Schadows]

@Heliosurge

Also not entirely true. FoveVR already had some of this support long ago but couldn’t gain traction with their own headset.

Oculus already supports agnostic gpu FFR techniques in their sdk for game development. Headset driver implementation of course will not be as good as game direct support as even Nvidia had demonstrated it’s gpu implementation on games is hit and miss.

That’s why (FoveVR, and also Vive Pro Eye) I said “without much compatible hardware out there”. The number of those headsets in the wild are hardly something you can count on to tilt the balance.

Still, it will happen … slowly, but still.

[androo87]

Any updates on when the Pimax eyetracking module might be available?

I am a researcher and developer, and I would love to sign up for a beta tester unit if available.