The PSVR launch thread

On a side note, I noticed that the OLED screen can of course produce perfect black (easily visible in many apps and games) however when it comes to near black, greys or otherwise darker tones, it gets quite grainy.
 
iWaggle3D made some fantastic Move controller use analysis in the past, and this new video focussing on VR and Move is great in how it overlays the real move controller's position and the virtual hands they control.

 
Shifty Geezer said:
Not with pupil tracking. That's what's needed to allow players to look around naturally. Latency may not be too much a problem as there's a focussing delay in the eye after a saccade that may well be enough for the GPU to adapt.
Click to expand...

No not with pupil tracking, but I bet the gains from that in games might be smaller than you think, as for a lot of games you are looking mostly in the center anyway.
 
Arwin said:
No not with pupil tracking, but I bet the gains from that in games might be smaller than you think, as for a lot of games you are looking mostly in the center anyway.
Click to expand...

Indeed. You really have to contort your eyeballs and actively look for it. Naturally you only move your eyeballs ever so slightly, thus never really looking away from the area of the screen where the level of detail is at its highest. It's your head that's taking care of larger movements.

The only game where I've noticed it was Arkham VR.
 
Is that because of the way the games are structured? In RL my eyes use a lot of their available range instead of fixing forwards and relying on the head to direct them. eg. Looking at this monitor! I can only visually resolve a small part of the screen based on where I'm looking (fovea). To read a page, my eyes scan across and down, without my head moving. Most of the FOV could be rendered low res as long as that few percent where I'm actually looking is sharp.

I seriously doubt anyone is using anything close to 'foveated' rendering. Likely just a shrinking of the sharp area to the more central part of the display, maybe 30 degrees FOV, and rendering the other 60 degrees FOV low quality. But that's still very far from the ideal and a couple orders magnitude more processing power wasted on unresolved visual data.
 
The detailed area in these games is way larger than the small area your eyes are usually focused on when looking straight ahead. So yeah, it's not really the super advanced 'foveated' rendering you're thinking of, but using your numbers, 30 degrees out of 90 is still 60 degrees worth of savings.
 
Reducing the resolution of the render target on the periphery is an optimization that should be getting done anyways as the barrel distortion ends up undesirably biasing the sampling to the peripheries of the image. You can't be too aggressive with this however as the eyes do have a surprisingly large range where they naturally rest (especially vertically, in order to prevent the neck and back from having to tilt the head downward). This is further emphasized when you start dealing with content that has you walking around and using your hands. Almost every form of natural interaction you have with your environment on a daily basis occurs below the horizon line of your head. Even the eyelids accommodate this by closing mostly from the top, allowing your eyelids to close partially and your eyes to sit more comfortably in a lowered position for longer periods without drying out. I suspect this also factored heavily in the choice to use fresnel lenses for Rift/Vive, as the flatter profile accommodates a wider area of full focus. Using the DK2 with hand tracking is a pretty rotten experience as you're having to make uncomfortable deliberate motions of your head in order to focus at what your hands are doing.

Foveated rendering is a different beast as you're talking about an area that's the size of a coin held at arm's length - that's a fraction of a percent of your total view. While pretty well all VR optimizations that currently exist are there to reduce the overhead that occurs with VR rendering (over and above traditional monitor rendering), foveated rendering could actually make VR orders of magnitude more efficient than traditional displays.

https://www.shadertoy.com/view/4dsXzM
 
milk said:
Holy shit, it's happening! Exciting times!
Click to expand...

Happening... eventually ;P

www.youtube.com/watch?v=RS-9g9kx8jw&t=1h40m54s

It's one thing to have a foveation technique that looks relatively inoffensive to the eyes, it's another to have the tracking be robust enough to never fail, and deliver a dramatic performance boost to make all the legwork worth while. Also have to factor in that the predictive reprojection techniques that are employed now to deliver the perceived low latency will need to be rethought.

Who knows.. maybe with this method of motion-vector synthesized frames we maybe could see a very different pipeline to draw and composite frames in VR. Looking at just what ASW can do has my gears turning for how such a thing could be used to accelerate path tracing or other such rendering schemes that accumulate samples and converge over time. Instead of starting with an empty canvas with each frame you could synthesize a pretty good guess and then devote your fresh rays/samples to the required area.

I wonder if we've had a performance surplus in hardware accelerated 3D for so long (basically since Voodoo1), that we've probably left an awful lot of old-school hacky optimizations on the table for extracting performance. A lot of this reprojection and synthesized frame stuff feels like the sort of thing you'd read about from the 80s or early 90s where every alternative to brute forcing fully updated frames was considered.
 
Try this if you want to see your fovea for yourself, and how it compares to your full field of vision.

This is pretty far afield from PSVR, though.
 
idsn6 said:
Try this if you want to see your fovea for yourself, and how it compares to your full field of vision.

This is pretty far afield from PSVR, though.
Click to expand...

That's quite interesting. At normal monitor distance the fovea is tiny, however with your face pressed right up to the screen as it is with VR, it's quite big. I seem to remember reading that pupil tracking and foveated rendering may reduce VR rendering requirements to as little as 25% of current which sounds realistic at a stretch.
 
pjbliverpool said:
That's quite interesting. At normal monitor distance the fovea is tiny, however with your face pressed right up to the screen as it is with VR, it's quite big. I seem to remember reading that pupil tracking and foveated rendering may reduce VR rendering requirements to as little as 25% of current which sounds realistic at a stretch.
Click to expand...

By moving closer to the screen with that application what you're doing is making the object motion in your periphery larger and more apparent, such that your fovea is no longer required to discern motion. In other words, the fovea is needed to see fine detail, but if there is no fine detail then you're not really going to see a benefit. The potential performance savings from foveated rendering dramatically increase once the panel resolutions are of a sufficiently high resolution, at which point you're talking about a lot more than a 2x or 4x. Even Nvidia's static MRS/LMS offers something on the order of 25-50% improvement right now. Foveated rendering, as Abrash said, should be something on the order of a 10x improvement eventually.
 
pjbliverpool said:
That's quite interesting. At normal monitor distance the fovea is tiny, however with your face pressed right up to the screen as it is with VR, it's quite big.
Click to expand...
Not really - that's a different effect. Your fovea is tiny and resolves only a small portion of what you see in full detail. Beyond that you have lower acuity. This demo is showing the brain's processing of the info. When close enough, there's enough variance in the low-fidelity areas for the brain to resolve what's happening correctly. The fovea is biologically fixed at something like 5 degrees FOV.
 
Has anyone already mentioned a little problem with surround sound?

I installed the PSVR the other day, didn't touch my settings. Last night I wanted to play a normal game and for the life of me I couldn't get 5.1 surround out of it. Checked my settings and somehow the Audio was set to HDMI OUT (instead of Optical which is how I've always set it up to get my surround juices running). I assume it set itself to HDMI Out because of the 3D audio on the PSVR, but it's a bit of a pain now to have to switch it back and forth when I'm playing normal games, or movies.

And no, Sony, the solution is not to buy an HDMI receiver. My good old 15 yo Sony surround system works mighty fine with Optical!
 
I think that may be because the binaural audio downmix to earphones takes an uncompressed 5.1/7.1 LPCM source, and can't do that with a compressed AC-3 one, which is the only thing that can go through optical SPDIF.

I guess the mistake here is Sony not having the system revert to your previous settings when PSVR is turned off. Maybe this will be fixed through a software update eventually.


You could just spend like $20 in a splitter/converter, though:
http://www.ebay.com/itm/HDMI-to-HDM...io-Extractor-Converter-Splitter-/311437978922
 
For me it reset the audio u output to hdmi out only once .

Once i set it to use optical again , everything's fine .

The DVD player also says Dolby digital .

Edit
Holy shit the bed chrome android is messing up my SwiftKey keyboard punctuation .

More edit
Oh my God the typing also feels like t-shirt in the new chrome , complete with weird auto connect .

Typing on another apps works fine . Weird .
 
You must log in or register to reply here.

Similar threads

Arwin
Official E3 Sony Conference Thread 2010
15 16 17
Replies
326
Views
38K
green.pixel
green.pixel
M
  • Locked
PS3: Phil Harrison addresses the real-time vs CGI issue
7 8 9
Replies
177
Views
19K
Mythos
M
B
  • Locked
Full Q/a With Tommy Tallarico Ps3 And 360
4 5 6
Replies
112
Views
22K
blakjedi
B
R
Reverend at The Pulpit #9
Replies
5
Views
2K
3dcgi
3
Back
Top