Sony Playstation Meeting September 7 2016 [PS4 Slim, PS4 Pro, Rumors, Speculations, and News]

Some (misinformed) people considered Killzone Shadow Fall interlaced rendering equal to 960x1080. Checkerboarding is an improved version of interlacing. Interlace pattern only differs. Both techniques render 50% of pixels per frame, reproject last frame to fill the gaps and interpolate occluded pixels.

But nobody questions whether 1080p Blu-ray movies are native or not. But modern video compression codecs reproject up to 90% of previous frame pixels. It's a waste to store or calculate a pixel again if it hasn't notably changed since the last frame.

We will see similar techniques in games as well. Calculate variable amount of new data per frame and not use evenly spaced sampling grids. Place new samples where it counts the most.

 

If you bought a BRD and it was soft and you found the data was stored at 900p, you'd probably be rightly pissed!

The difference with video is you know how much change there is on a pixel and can choose to calculate/compress it or not. Reconstruction isn't skipping unchanged pixels but skipping pixels that could need to be calculated, certainly when the camera is moving. Reconstruction is instead trying to 'cheap out' on the rendering in the hopes that the interpolated value is sufficiently close to the value it should be to not be noticeably different. And in most cases it probably is (not noticeably different), and in most cases it's probably a very good idea.

 

I was talking about more advanced methods (no fixed sampling grid).

First, you could simply render (refresh) every pixel that failed reprojection (never interpolate). This would be simple, but not efficient to implement if you are rasterizing your image with hardware ROPs (stencil buffer masking small areas is not fast and doesn't reduce vertex processing cost). Pure compute shader based renderers could do with high efficiency (point clouds, SDF, etc). This would solve the occlusion case.

Perfect error metric is impossible to calculate at runtime as you don't know the exact goal (frame is not yet rendered). You'd have to render a full frame to calculate perfect error. Obviously not worth it. But you could estimate the error by various means and/or boost error metric artificially to improve quality on areas where errors are most likely seen (high contrast, geometry edges, shadow edges, motion vector discontinuities, etc). I believe we will eventually get pretty close to the reuse factor of video codecs.

 

Break into warping thread?

 

Just realized how super sleek and nice looking the Glacier White PS4 Slim is. Best looking console of "this gen" imo.

 

It's even more amazing when you put it next to an Xbox One S, it's so small!

 

I'm not sure that's true:

"The dimensions of the PS4 are 275 x 53 x 305mm, compared to the 265 x 39 x 288mm of the PS4 Slim."

http://www.valuewalk.com/2016/11/ps4-pro-vs-ps4-slim-vs-ps4/

 

Only 1.1 inch deeper?

Volumes seem comparable.

 

Good thing I'm OK with junk in the trunk....

 

PS4 Slim is more than 50% smaller by volume compared to XBO S.

 

Err... PS4 Pro is much closer in volume to the Xbox One S, than the PS4 Slim is in volume to the Xbox One S. With the Slim being much smaller, obviously

 

Beyond3D Console forum. Focussing on the important technical issues.

 

12.36% deeper, and that's in the z-axis so you don't tend to see it. y-axis delta is 40% less than XBOS in one of the two higher priority dimensions. Total volume is 63.93% of XBOS. PS4 Slim Glacial White is objectively far smaller than One S. It's also 15.44% whiter than XBOS, while the blacks are only 3.73% less black, meaning contrast in the styling is overall more than 10% greater in the PS4 Glacial White Slim.

PS4 White Glacial Slim has better numbers. Objective facts (other facts are available).

I've tried!

 

Could you write down your method of measurement? I think you used DCI-P3 and BT.2020 for each values. Black should be 3.11% less black (at max) if you measure using the same color space.