120hz or 1080p ++?

why 120fps ?
If we dont move the head fast... 60 is largly enough

Its gotta be 1080p or consumers wont bite. But keep in mind even at 1080p it still wont look as good as a monitor. There are ways around this by designing 3d models and textures for the rift and not for a monitor. That I think is the biggest mistake so far is you have to design not only the game play around the rift but the graphics as well play on the weakness so to speak.

"KBK" wrote:
I mean, 1080p smartphones will be 'so last year' in about 6 months.

Are you sure about that? I am not entirely sure what would drive increased resolution for smartphone displays (which have a natural upper limit in display size of about 5"). I currently own an S4, and it is absolutely impossible for me to spot any pixels no matter how close i get. While i can certainly see the need for a higher resolution display for HMD use, i am not sure why consumers would want a 2650x1600 smartphone display in the 4" - 5" range? Most likely it wont look better, and it will drain the battery somewhat faster.

Short question. What is the actual developer kit resolution at the moment ? 720p or 1920x1200?

Greetings
Dominik

"cbell77" wrote:
Its gotta be 1080p or consumers wont bite. But keep in mind even at 1080p it still wont look as good as a monitor. There are ways around this by designing 3d models and textures for the rift and not for a monitor. That I think is the biggest mistake so far is you have to design not only the game play around the rift but the graphics as well play on the weakness so to speak.

Lets not forget on a monitor its 1080p per eye.
Where on rift its 1080p for BOTH eyes. (so 1080p/2 per eye)

And that only 2/3 of the screen is used.

So its would be only one third of the 1080p resolution we can see on a monitor.

but its always better than the actual "one third" of a normal 720p resolution.

"Tgaud" wrote:

"cbell77" wrote:
Its gotta be 1080p or consumers wont bite. But keep in mind even at 1080p it still wont look as good as a monitor. There are ways around this by designing 3d models and textures for the rift and not for a monitor. That I think is the biggest mistake so far is you have to design not only the game play around the rift but the graphics as well play on the weakness so to speak.

Lets not forget on a monitor its 1080p per eye.
Where on rift its 1080p for BOTH eyes. (so 1080p/2 per eye)

And that only 2/3 of the screen is used.

So its would be only one third of the 1080p resolution we can see on a monitor.

but its always better than the actual "one third" of a normal 720p resolution.

No doubt that 1080 screen will be better than what we have now. But it still will not be great I think in the long run until they upgrade with an even higher res screen that better use of textures and overall good modeling techniques will help alleviate the lack of graphical fidelity.

I think resolution is the absolute priority for now. Things farther than 10 meters away start to look like a light bright. The sailboats in the tuscany demo are literally 6 pixels tall if viewed from the handrail. At that distance you only know they are boats through context, visually they are just blocky white triangles.

LG Display claims a world's first with 2,560 x 1,440 LCD for smartphones. Dear Oculus Developer Team plz consider this opportunity for the next generation beyond 1080p.

my feature requests:
higher resolution;
positional motion tracking;
external camera (probably feeds back video over USB or something);
built-in headphones + microphone;
...



otherwise, I am left to wonder about an idea, but have not yet tested it.
this is basically to try to make high-framerate rendering more plausible (without imposing as much on the main renderer).

instead of rendering the views directly, we render into a dome-map (flat texture with a fairly large FOV, eg, 120 or 180 degrees), then, the view for the Rift is rendered from this (in its own thread, or otherwise asynchronously), potentially allowing lower-latency (the dome-maps will not be updated as quickly as the main view from the Rift is redrawn).

possible problems:
problematic for stereo-rendering (less of an issue for parallax-warping, but granted, this produces artifacts, *1);
driver reliability problems when using OpenGL from multiple threads (*2).

*2: I was having problems with AMD drivers with a multithreaded renderer, but it seemed to work ok with NV drivers.
could need more testing, as it could be something that can be worked around...

basically, you would need 2 rendering threads:
a scene renderer, which updates the cube or dome map(s) in the background;
an HMD renderer, which basically just redraws the HMD view, which should be able to go faster as it is mostly just doing a single big draw using a shader.


*1: this is simpler with a single map (color+depth), and using a tweaked form of parallax-occlusion mapping, but this introduces ugly-looking artifacts (noisy/distorted edge details).

if using dome-maps and wanting full stereo, this could be possible:
render dome-maps from a slightly larger eye-distance, and with "predicted" eye positions;
render specific view using an image interpolated from both domes, based on the current view-angle and position.

compensating for eye-offset, the depth values should be "similar" between domes, but the differences should be usable in terms of blending the two images in order to generate a new pair of images for the current eye positions. similarly, the estimated normal should indicate which image is given more priority WRT the interpolated pixel color.

the idea being that maybe even if the main renderer only pulls off maybe 20 or 30 fps, it should still be hopefully possible to get 60+ fps from the Rift via interpolation.

or such...

Hey, guys.

Speaking of parallel computing for VR, you've seen this kickstarter project, right?

http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone?ref=category

Even if those games are 10 years away, the hardware seems to have arrived fresh out the oven! :P