• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

PS4 Pro provides checkerboard rendering in hardware - no cost to devs

Afrikan

Member
In a rough, very small nutshell.

lnYHDdx.png


Render the red pixels.
Calculate everything else.

It's late and I'm probably not thinking this through....buuuut

Can Sony use this technique to stream 4K Video (Movies) on the PS4 Pro, with not needing to use a lot of bandwidth?

Making 4K Movies available to more people who don't have fast (25mb?) Internet?

Or maybe to add lossless sound for those who want it?
 

Afrikan

Member
Well, if you are talking of using a similar technique with sound, to reduce the size needed for the track... it wouldn't be lossless sound anymore, you are losing half of the information. :p

Lol No, I meant with the saved space of the 4k Video stream, be able to add full lossless sound.

Again just thinking out of my ass here.


One reason I still/only use BluRay Discs is because of Lossless sound.
 

cheezcake

Member
It's late and I'm probably not thinking this through....buuuut

Can Sony use this technique to stream 4K Video (Movies) on the PS4 Pro, with not needing to use a lot of bandwidth?

Making 4K Movies available to more people who don't have fast (25mb?) Internet?

Or maybe to add lossless sound for those who want it?

That would basically be a 2 decade regression in video compression technology, what we have now on major streaming platforms compresses much better and would have much higher quality.
 

double jump

you haven't lived until a random little kid ask you "how do you make love".
Prob been asked before but could this tech be used as a filter to upres older games from the sd era ?
 

Inuhanyou

Believes Dragon Quest is a franchise managed by Sony
This :/

I really did not like the look of quantum break at all.

Its not like Rainbow Siege or quantum break, the implementation and concept is different even though they both use similar general concepts. For one thing, Quantum break and Rainbow six were both using sub HD samples to render the image they had in place, it ended up not even looking close to 1080p.

Sony are using 2 1080p frames to output the final image, so the fidelity and lack of blurring would be significant even if they were using Rainbow Six's technique
 

cheezcake

Member
Its not like Rainbow Siege or quantum break, the implementation and concept is different even though they both use similar general concepts. For one thing, Quantum break and Rainbow six were both using sub HD samples to render the image they had in place, it ended up not even looking close to 1080p.

Sony are using 2 1080p frames to output the final image, so the fidelity and lack of blurring would be significant even if they were using Rainbow Six's technique

I think you need to reword this lol, "the implementation and concept is different even though they both use similar concepts", haha

Btw do we actually have any technical details of the checkerboard render Sony is using? Because obviously the concept is identical to what R6: Siege and QB did but people seem to be talking as if the technique is worlds apart.
 

pottuvoi

Banned
I think you need to reword this lol, "the implementation and concept is different even though they both use similar concepts", haha

Btw do we actually have any technical details of the checkerboard render Sony is using? Because obviously the concept is identical to what R6: Siege and QB did but people seem to be talking as if the technique is worlds apart.
As far as I know there has not been a presentation or implementation details on it.
 

Afrikan

Member
That would basically be a 2 decade regression in video compression technology, what we have now on major streaming platforms compresses much better and would have much higher quality.

Oh OK cool.

I believe the requirement to stream 4k used to be 25mbs... But I thought I read somewhere that it has dropped to 15mbs is it?
 
It's late and I'm probably not thinking this through....buuuut

Can Sony use this technique to stream 4K Video (Movies) on the PS4 Pro, with not needing to use a lot of bandwidth?

Making 4K Movies available to more people who don't have fast (25mb?) Internet?

Or maybe to add lossless sound for those who want it?

I think you wouldn't see any benefit in using such a thing in videos. Video compression algorithms are already much more advanced than this. For the simple reason that compressing a frame in a movie is much easier than a frame from a video game because in a movie you have all the frames at once and you can use this knowledge to be more efficient. In a game, you see the current frame but you have no clue at all what would be next.
 

consoul

Member
Modern video compression, while marvelous, still results in less pixels of original detail than PS4Pro's 'uprendering' technique or whatever we're going to call it.

You won't hear the haters complaining about Pro's cheating method of 4K bemoaning the fact that most of what they're actually watching on 4K Blu-ray is calculated, not captured from original source details.
 

onQ123

Member
It's late and I'm probably not thinking this through....buuuut

Can Sony use this technique to stream 4K Video (Movies) on the PS4 Pro, with not needing to use a lot of bandwidth?

Making 4K Movies available to more people who don't have fast (25mb?) Internet?

Or maybe to add lossless sound for those who want it?

Prob been asked before but could this tech be used as a filter to upres older games from the sd era ?

http://www.freepatentsonline.com/y2016/0005344.html

In some embodiment, the uprendering achieves a higher resolution uprendered image. This uprendering can be applied to substantially any image to generate a higher resolution image, frame or the like. For example, the uprendering can be applied to legacy interactive electronic games, video content, digital photographs, and/or other such images. In some instances, the uprendering is applied to legacy PlayStation® electronic games that often generate frames having a resolution of [640×448]. Accordingly, the uprendering can be applied to increase the resolution of frames as the game is being played back to generate higher resolution images that can take advantage of displays and televisions with greater resolution and/or increased size that were available at the time the gaming titles were originally generated. As such, the uprendering allows images of PlayStation legacy gaming titles to be uprendered and displayed with higher definition graphics while still maintaining a correct emulation model of the original legacy hardware on which the titles were based. Additionally, in some embodiments, the uprendering through the cooperation of the reference image and the shifted images can retain the deterministic method of point-sampling that some image systems and/or legacy titles utilize to render correctly. Further, the uprendering provided through the coalescing of the reference image and the one or more shifted images often minimizes the introduction of artifacts, which are common to other methods of uprendering.
 

Leonidas

Member
Do you guys think that this upscaling would look nice on a 2560x1440 monitor? That'd provide some supersampling and potentially make the image look closer to native, no?

Of course this is assuming WQHD monitors are even supported.
 

pottuvoi

Banned
Been wondering how often GPU can change the MSAA subsample shading rate.
If you can change it per object or tens of times per frame there might be some interesting saving opportunities. (Decouple shading from resolution.)

With 8x MSAA used in chekerboard rendering you could fill 4x4 pixel quads.
Meaning rendering resolution of 980x540 8xMSAA to get 4k after checkerboard and temporal gives double samples.

Resolution and maximum shading would be identical as would sample positions.
Minimum shading could be 4x less even if you would do same trick with the original 2x version. (1/16th of native 4k.)

Do you guys think that this upscaling would look nice on a 2560x1440 monitor? That'd provide some supersampling and potentially make the image look closer to native, no?

Of course this is assuming WQHD monitors are even supported.
Usually you want to have perfectly dividable pixel dimensions for downscaling for it to look good.

Great sampling can lessen the artifacts, but the result must be a bit soft for it to look good.
 

EatChildren

Currently polling second in Australia's federal election (first in the Gold Coast), this feral may one day be your Bogan King.
Ummm. Quantum break looked blurry as Fuck. Rainbow six looks soft as well compared to uncharted and infamous.. No thanks.

This :/

I really did not like the look of quantum break at all.

Well yes, not native rendering is always going to look blurrier than native. QB/R6 are not running at native 1080. U4 and Infamous are.

And in turn checkerboard 4K games also look blurrier than native 4K.
 
With all this checkboard rendering stuff I've been wondering why no one has implemented the stuff that the force unleashed 2 guys developed: http://www.eurogamer.net/articles/digitalfoundry-force-unleashed-60fps-tech-article

I'm not saying they should interpolate 30fps to 60-- they should treat it like a video codec and render the areas that have a lot of change on the screen, just use calculations to render the areas that have very little movement.

A lot techniques make use a temporal data, I believe GG cloud rendering does it and insomniac had some paper on it for shadow map generation.

Unless you mean something different?
 

Fafalada

Fafracer forever
pottuvoi said:
If you can change it per object or tens of times per frame there might be some interesting saving opportunities.
Variable shading rate is - opposite of what you want here though - you have 2.3x the shading power, you want to know you scale in that region at all times, else you're just dealing with randomness, and that's basically a PC - ie. hw optimization anti-thesis.
Decoupled shading has nice benefits on its own of course - but it's only really interesting if you can apply it across all your SKUs.

infinityBCRT said:
I'm not saying they should interpolate 30fps to 60-- they should treat it like a video codec and render the areas that have a lot of change on the screen, just use calculations to render the areas that have very little movement.
While that could work - you end up with a wildly varying performance profile (anytime camera moves for instance, there are no low-movement areas on screen). Video has the benefit of such peaks being predictable (on account of - non-interactivity and all), in an interactive app I'm not sure you could amortize peaks effectively enough.
 

dr_rus

Member
You said it was not encoded into the hardware. We have Sony and devs directly saying that hardware acceleration for implementation of this specific technique is encoded into the hardware.
I said I very much doubt that it's a fixed function specialized h/w which they are referring to. I doubt that it's even there but at least it's possible, ff h/w would just be stupid.

What's terrible about it?
It will create more issues in load balancing then it will solve in cache conjestion. Running such reconstruction across all CUs would be very much preferred option. It's also not "free" at all to take several CUs out of general graphics rendering.
 

SpotAnime

Member
With all this checkboard rendering stuff I've been wondering why no one has implemented the stuff that the force unleashed 2 guys developed: http://www.eurogamer.net/articles/digitalfoundry-force-unleashed-60fps-tech-article

I'm not saying they should interpolate 30fps to 60-- they should treat it like a video codec and render the areas that have a lot of change on the screen, just use calculations to render the areas that have very little movement.

This is an excellent article, thanks for sharing. Really interesting to see what dev teams were able to achieve last gen, especially in light of the resolution and framerate issues this gen.

I don't know if anyone already stated this - I know the thread started out with comparisons to QB and KZ - but I'm thinking the implementation of this in the PS4Pro will be more effective because it's baked into the hardware, rather than achieved in-engine. Eurogamer did state that RotTR did look soft but was still the most impressive third-party title shown at the event, which gives confidence this rendering will look just fine.
 

onQ123

Member
I said I very much doubt that it's a fixed function specialized h/w which they are referring to. I doubt that it's even there but at least it's possible, ff h/w would just be stupid.


It will create more issues in load balancing then it will solve in cache conjestion. Running such reconstruction across all CUs would be very much preferred option. It's also not "free" at all to take several CUs out of general graphics rendering.


They come out & say that it's extra hardware that help achieve 4K level fidelity https://youtu.be/AJWD6RpAt6E?t=1m20s
 

cheezcake

Member
Oh OK cool.

I believe the requirement to stream 4k used to be 25mbs... But I thought I read somewhere that it has dropped to 15mbs is it?

I don't really know tbh, this might be the step from h264 to h265 (gap still sounds a little too big) but it hasn't seen big adoption yet coz not many devices support it, but it'll grow over time.
 
People more hardware savvy than I, here's a question:

If the custom hardware found on PS4Pro GPU proves to be very efficient, leaving the 4.2 TFLOPS free to do other stuff and the Scorpio had to render things at 4K, or even upscalling without those custom enhancements, could the performance gap between the two be narrowed?
 

RoboPlato

I'd be in the dick
People more hardware savvy than I, here's a question:

If the custom hardware found on PS4Pro GPU proves to be very efficient, leaving the 4.2 TFLOPS free to do other stuff and the Scorpio had to render things at 4K, or even upscalling without those custom enhancements, could the performance gap between the two be narrowed?

It's a mitigating factor for subnative resolutions. The percent difference between the Scorpio and PS4 Pro GPUs is almost identical to the PS4 and Xbox One. I expect we'll just see higher native resolutions on Scorpio. #4k does look quite solid, but there are still artifacts and a softer look than native.

That said, Scorpio could do something similar.
 

onQ123

Member
People more hardware savvy than I, here's a question:

If the custom hardware found on PS4Pro GPU proves to be very efficient, leaving the 4.2 TFLOPS free to do other stuff and the Scorpio had to render things at 4K, or even upscalling without those custom enhancements, could the performance gap between the two be narrowed?

Yes but who is to say that Scorpio isn't using even more hardware to help it along.
 

cheezcake

Member
People more hardware savvy than I, here's a question:

If the custom hardware found on PS4Pro GPU proves to be very efficient, leaving the 4.2 TFLOPS free to do other stuff and the Scorpio had to render things at 4K, or even upscalling without those custom enhancements, could the performance gap between the two be narrowed?

Not to any meaningful extent.
 

dr_rus

Member
People more hardware savvy than I, here's a question:

If the custom hardware found on PS4Pro GPU proves to be very efficient, leaving the 4.2 TFLOPS free to do other stuff and the Scorpio had to render things at 4K, or even upscalling without those custom enhancements, could the performance gap between the two be narrowed?

Upscaling and even reconstruction are generally cheap and whatever PS4Pro will have Scorpio will be able to get as well, even without any "specialized h/w" it can just use its shader core to perform the same operation and it won't cost them a lot of performance. So no, I don't think that this will affect the gap much.
 
Top Bottom