Baseless Next Generation Rumors with no Technical Merits [pre E3 2019] *spawn*

incredible feat.
Excited to see more details on this one

 

What exactly is the feat? What's the hardware? What's being raytraced?

 

On CES they had an 8k demo without raytracing. This is probably 4 PS5 devkit. I think the game will be 4k 120 fps with hardware photon mapping 2d tile. Photon mapping was not far from being ready for been use real-time in-game in 2013 when Nvidia tested all the different photon mapping method on Geforce 670. They tested some complex scene but all static. KD-tree generation is too slow for a dynamic scene in compute shading slower than BVH. I think here this is KD-tree photon maps generated via something like the scene hierarchy generator of Power VR. Photon mapping is fast. In SIGGRAPH 2008 they did a comparison between path tracing and photon mapping.


http://casual-effects.com/research/Mara2013Photon/Mara13Photon.pdf

Nvidia search about the different photon mapping method on a Geforce 670, the best one is 2d tiles on a GPU

https://patents.google.com/patent/US9633471B2/en

Sony patent with hardware 2d tiles photon mapping, reading the patent it seems rasterizing is only used for direct illumination and visibility for shadow. Full indirect photon mapping lightning

http://fileadmin.cs.lth.se/cs/Education/EDAN30/papers/photonMappingTutorial-SIG08.pdf

It goes much faster than path tracing, sometimes two orders of magnitude faster

Advantage:
_ Much faster than path tracing/raytracing
_ fully independent of geometry scene complexity
_ In 2D tile method resolution is not a real limitation. This is the number of photons you generate during the photon tracing pass.
_ Easy to integrate inside an existing rasterizing or raytracing pipeline
_ In 2D tile possibility to use tile rendering inside on-chip memory, I believe this what Sony is doing reading the patent
_ No need to do some denoising when the number of photons is too low the light blur, like into the REVISION 2019 raytracing 4k winning demo.

Problem
_ KD-tree construction is slow(probably generation with something like scene generation in Power VR GPU)
_ Can't do glossy effects like eye reflection or certain refraction. Can't do direct illumination. The two effects demand too much photon.
_ Difficult to parameter the number of photon for each surface, it kills photon mapping as a global method for offline rendering. This is only used for some effect like volume rendering where it is great and much faster than path tracing. Importance sampling can be a solution to this problem.

 

I think 4 PS5 devkit, in CES 2019 they had an 8k 120 fps HDR demo too on Crystal LED like this one but without raytracing...

EDIT: This is not the first time Sony uses Gran Turismo for demo linking multiple consoles...

 

8K 120fps is a pretty big feat. Lots of pixels per second.

not sure what the load is on ray tracing.

Even doing something successfully like SFR over multiple GPUS is generally excitiing, i'm not sure if those pixels are stitched or each PS5 has it's own quadrant in that showcase

But if the load is real, then I guess the approximation is 1080p/120hz Ray Traced per unit, or an equivalence of 4K/30fps ray traced. Fairly significant power for a console.

 

That looks like the regular game. I'm calling BS on the RT claim.

EDIT: for a confirmed GT ray tracing demo look here:

https://www.gtplanet.net/gran-turismo-tech-demo-real-time-ray-tracing-replay-movie-editor/

 

So that would mean 4K 120fps for PS5. With checkerboard rendering it's totally feasable. That's about only 4x more pixels processed than PS4 1080p 60fps (1.84 * 4 = 7.36). They could even have enough juice for some Ray tracing, depending how it's done obviously.

 

Have we any information that it's even realtime? There's absolutely nothing about what we're seeing here! Could be a frickin' CGI render based on what's presented - a video demoing a TV with no qualifications in the description. It's very likely to be some random who saw this video and made the assumption it's PS5 - it's a nothing channel where you'd expect a proper tech site to cover something significant like a first public showing officially labelled PS5.

 

This is native... it means 4k 120 fps like I said if this is photon mapping this is much faster than path tracing or raytracing. The global illumination is totally independent of geometry photon map an Avatar movie scene and a PS1 scene at the same resolution if you have more photon in the PS1 scene it will be slower to do the photon mapping pass than on the Avatar movie scene, only dependent of number of photon and resolution but resolution is not a big problem with 2d tile photon mapping...

 

How do you know it's native from a badly compressed youtube video ? Native 8K would be utterly stupid. What a waste of processing. If it's native then they used 8 PS5 dev-kits, not 4. Assuming it's even possible.

 

Like in CES 2019 this is not labeled as PS5. Read a little about photon mapping you will understand why it can be very interesting for real-time. The biggest problem is kd-tree construction which made it unusable for dynamic scene without a fast KD-tree generator. In SIGGRAPH 2008 something needing 100 rays per pixel to render in 5 minutes in path tracing to be noise free, took 5 seconds and 10 rays per pixel in photon mapping without blur.

 

Because on the right side of the screen there is the resolution and the number of fps display on the Crystal led screen...

Read the paper of photon mapping it can be 100 times faster than path tracing...

 

Someone needs to provide proof this is raytraced, and realtime. I see absolutely no evidence other than someone labelled the video as such. Until there's any confirmation, this discussion is ludicrous. I may as well shoot some 4K phone footage and post it on YouTube as "Xbox Anaconda realtime raytraced demo"...

How do you know that text is the footage render speed and not the display details? How do you know this is realtime and not a video? How do you know it's ray-traced? How do you know the demo isn't a demo of the display's upscaling abilities showing what a PS4 game would look like upscaled to 8K and frame-interpolated to 120 Hz? Where are the details of what's being shown so we know how to interpret it?

 

In 2013 Nvidia said Photon mapping could be use for all Global illumination soon on high end hardware without RTX in videogame. It was 1080p scene on a Geforce 670 but static scene because the problem was KD-tree generation too slow in compute shading. 6 years ago on a mid range GPU without RTX...

With full BSDF it was a little slow but with Lambertian only it was pretty fast.

Because at CES it was presented as native 8k Gran Turismo demo at 120 fps but without raytracing with probably multiple devkit for rendering...

And this probably not raytraced but use photon map, raytracing is too slow...

EDIT: Photon tracing phase is like raymarching.

It was not SIGGRAPH 2018 but SIGGRAPH 2008 for the demo taking 5 minutes to render with path tracing and 100 rays per pixel enough to eliminate the noise and 5 seconds and 10 rays per pixel to eliminate the blur on Photon tracing... Sorry for the typo...

 

The year after the 2015 hardware 2D tile Photon mapping patent, Sony Interactive Entertainment hired an employee of Imagination/Power VR which did his master thesis on Progressive Photon mapping using CUDA in a GPU. He is a raytracing/Photon mapping specialist...

https://www.linkedin.com/in/carlovloet

https://carlovloet.wordpress.com/

On raytracing gem book there is a chapter about Photon Mapping on RTX with a 2080 TI but it does not use the 2D tile method.

https://github.com/ananthaks/DXR-PhotonMapper

A photon mapper on Titan X, slowest part kd tree construction. I read the Kun Zou 2008 paper about KD-tree construction on GPU I don't think this is fast enough, the GPU is only as good as a multicore CPU.

If someone resolve the KD-tree construction this a much better solution than raytracing and the only way to solve it is to use something fast enough and able to build BVH every frame like the scene hierarchy generator of Power VR but for KD-tree.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.118.6575&rep=rep1&type=pdf

When I say the Photon mapping pass will be slower with more photon on a PS1 scene at same resolution than a scene of the Avatar movie, this is not a joke... This is fully decoupled from scene complexity, you can use tesselated, displacement shading or any form of complex geometry or primitive this will not be the problem and the limitation. It will be the number of photons linked to the number of ray.

2D tile Photon mapping has less dependancy on resolution than other photon mapping method and probably other global illumination method. The resolution will be the one you can reach with the direct illumination rendering pass.

Edit: In offline rendering the biggest problem of Photon mapping it was too difficult to parameter and it took too much artist time . But the inventor of the method told it is possible to solve problem with importance sampling but I think offline rendering choose path tracing only more power was needed after all. Artist time is too precious.

 

I am posting just to prevent the 8th post in a row of @chris1515

 

Sorry, I tried to explain myself. All people coming like pitbull not even knowing what is Photon mapping. And why it looks like the best candidate for global illumination in real-time rendering.

I suppose Nvidia did not use it maybe because of patent problem, the scene hierarchy generator is a Power VR exclusivity. But in 2013 it was described by Nvidia itself as the best candidate for fully indirect lightning global illumination.

And there is a reference to the Nvidia study inside the Sony patent and reading the patent you understand they use 2D tile Photon mapping method.

 

Clearly PS5, first pastebin rumour states XBOX has larger lv1 cache for AI and such, 3rd paste bin says one has larger CPU portion for increased cache.

It says larger cache in two of the rumours, nothing about either CPU being more powerful, if both are clocked the same I doubt there will be any difference at all.

 

Larger L1 seems like too fundamental a change compared to what we've seen with their L3. (Zen 2 seemingly has gone with higher associativity and smaller L1).