Ps5 and Local ray Ray tracing for Ps5 ? Spiderman next gen ray tracing ?
- Thread starter geordiemp
- Start date
Rea
Member
I really wish they use Temporal Injection for 4k/60 just like SM ps4 and retain RT.
3liteDragon
Member
Or the 60FPS mode does also have ray-tracing but it's checkerboarded 4K with the GPU natively rendering at 1440p resolution.
Griffon
Member
Software patents are cancer. It's just logical math solutions to problems. Imagine the world if someone owned the concept of 2+2=4. That's exactly how software engineers feel when they discover a clever piece of code they made happened to be already patented by some asshole company.
Last edited:
Elog
Member
The point is that the current PS5 showcases have a surprising amount of RT if you assume that the GPU of the PS5 has the standard AMD RDNA2 RT set-up. Either the PS5 has done something on the hardware side, the software side or both.
The alternative is that the Minecraft RT showcase was criminally unoptimised when it ran on the XSX.
The alternative is that the Minecraft RT showcase was criminally unoptimised when it ran on the XSX.
Rea
Member
To be correct, Insomniac doesn't use CB rendering, they use temporal injection method which is abit better than CB rendering.
Bo_Hazem
Banned
No one works for free, you won't as well. You have like a cut/tax when used by other companies if you want to or after certain time. Without that glory and money we wouldn't have reached this level of tech.
VFXVeteran
Banned
Inputs fromVFXVeteran would be nice, its his scene.
OK.
So here's the deal:
You can get incredible performance during reflection ray-tracing by assuming every shader has a mirror reflected ray. That means you don't need to cast multiple rays around the reflected vector.
But that is not good enough. You avoid all the extra rays around that reflected vector which is more to real life. So what happens is you got a game that shows mirror-like surfaces all over the place. This is the typical first step to ray-tracing reflections but if you have it all over the place then it just looks bad.
A much better approach is to spend more of the ray budget to cast multiple rays around that reflection vector (i.e. Glossy in the image) to get blurred reflections, but these consoles don't have the bandwidth.
This would make the games look way more like real life because most surfaces that reflect aren't mirrors.
Some company may do some tricks to blur the reflection I suppose.
One other thing to note. With every local ray cast for a color of an object, a developer will try to avoid evaluating various levels of complexity in the shader. You see that with this girl walking in the mirror. It looks wrong to see her fully reflecting her environment but you don't see that reflection level in the mirror. They compleley just used her diffuse contribution. You will see a LOT of this in upcoming games. This should be your clue that we are still many generations away from having the bandwidth to render even reflections within reflections.
So keep an eye out for the things I mentioned. They will become readily noticeable now that you know.
3liteDragon
Member
alabtrosMyster
Member
That spiderman costume is so shiny!
Bo_Hazem
Banned
OK.
So here's the deal:
You can get incredible performance during reflection ray-tracing by assuming every shader has a mirror reflected ray. That means you don't need to cast multiple rays around the reflected vector.
But that is not good enough. You avoid all the extra rays around that reflected vector which is more to real life. So what happens is you got a game that shows mirror-like surfaces all over the place. This is the typical first step to ray-tracing reflections but if you have it all over the place then it just looks bad.
A much better approach is to spend more of the ray budget to cast multiple rays around that reflection vector (i.e. Glossy in the image) to get blurred reflections, but these consoles don't have the bandwidth.
This would make the games look way more like real life because most surfaces that reflect aren't mirrors.
Some company may do some tricks to blur the reflection I suppose.
One other thing to note. With every local ray cast for a color of an object, a developer will try to avoid evaluating various levels of complexity in the shader. You see that with this girl walking in the mirror. It looks wrong to see her fully reflecting her environment but you don't see that reflection level in the mirror. They compleley just used her diffuse contribution. You will see a LOT of this in upcoming games. This should be your clue that we are still many generations away from having the bandwidth to render even reflections within reflections.
So keep an eye out for the things I mentioned. They will become readily noticeable now that you know.
Do you think the rumored AMD Big Navi HBM2e, 2TB/s bandwidth, 24GB VRAM could provide a jump on that?
VFXVeteran
Banned
2TB/s bandwidth? Of course. I'd be really shocked if Big Navi had that kind of hardware. That would be 2X the bandwidth of the 3090. It would be quite a sight to see!
EDIT: Let me be clear before someone comes in and points out that I'm wrong about this since I answered adhoc.
When I mention bandwidth, I mean every aspect of moving data and not just memory accesses. I mean rays/s, triangle verts/s, pixels/s, etc.. It's not just moving memory alone. So since I don't know the ray budget, I can't say but I somehow doubt it. As the 3090 I don't think will have enough to ray-trace a game the proper way and still move 4k/30FPS.
Last edited:
Bo_Hazem
Banned
My Radeon VII already has 1TB/s bandwidth HBM2 16GB VRAM (no RT, of course). I think that's happening to the upper model.
Last edited:
Griffon
Member
You misunderstood what I talked about. But then again, I was in the wrong thread to bring the subject of software patent (which is not the same thing as software IP and copyright).
Thugnificient
Banned
How is the Minecraft demo remotely comparable? The demos mostly had limited RT such as reflections whereas Minecraft had the full suite. You can't seriously compare ray traced puddles and half-res reflections to what was being done in Minecraft, come on.
Bo_Hazem
Banned
My Radeon VII already has 1TB/s bandwidth HBM2 16GB VRAM (no RT, of course). I think that's happening to the upper model.
Further about Radeon VII, they just released a new revision as a professional card, more proves that it's a glorified workstation graphics card repurposed into gaming. That's why I bought it in the first place, honestly.
— AMD Radeon™ Pro VII delivers up to 26 percent higher performance in Blackmagic Design DaVinci Resolve than the competition1, and is optimized to power demanding broadcast workloads, complex CAE simulations and HPC application development —
— New AMD Radeon™ Pro Software for Enterprise delivers enhanced performance for current-generation AMD Radeon™ Pro graphics cards; New plug-ins enable Radeon™ ProRender software integration into SideFX® Houdini™ and Unreal® Engine —
— New AMD Radeon™ Pro Software for Enterprise delivers enhanced performance for current-generation AMD Radeon™ Pro graphics cards; New plug-ins enable Radeon™ ProRender software integration into SideFX® Houdini™ and Unreal® Engine —
Last edited:
Rea
Member
Sry i can't answer, but according to digital foundry video on Spider-Man ps4, Temporal Injection looks closer to native 4k than using CB rendering and looks abit better.
Time stamped: 7:41
onesvenus
Member
Doing calculations in reverse, from light to eye, makes no sense. We are doing raytracing eye to light because doing it that way we are optimizing the number of rays needed because the only thing we know at the start is that we are only seeing rays that end at the eye position. Throwing rays from the light in random directions would be a waste because most of them would not meet the eye.
VFXVeteran
Banned
My Radeon VII already has 1TB/s bandwidth HBM2 16GB VRAM (no RT, of course). I think that's happening to the upper model.
See my edit above
SilentUser
Member
If the CB rendering is good like lots of games this gen, I will totally use it!
ShadowWolf712
Member
Fam, this company's been using it for a good while now. Funnily enough, this was actually one of the "sources" of the claims that the PS5 was using Software Raytracing instead of Hardware back in the day. Folks pointed to this company and their Spider-Man demo as evidence.
From back in August 2019:
I wouldn't put too much stock in it though, especially as it looks like it's far more geared toward the mobile market. I'm sure Sony would def go more proprietary for any software-based solutions anyways. Might be for Marvel's future mobile titles though.
Last edited:
Bo_Hazem
Banned
geordiemp
Member
Who is fam ? Is that an american thing ?
THe new info is localray have done a NDA deal with 1 mobile manufacterer and one console manufacterer and there is new info and screenshots. And no it is not just geared towards the mobile market
PC & Laptops | Adshir | Local Ray
www.adshir.com
And the 2020 demo is not mobile, it says clearly its console / pc market
Last edited:
geordiemp
Member
OK.
So here's the deal:
You can get incredible performance during reflection ray-tracing by assuming every shader has a mirror reflected ray. That means you don't need to cast multiple rays around the reflected vector.
But that is not good enough. You avoid all the extra rays around that reflected vector which is more to real life. So what happens is you got a game that shows mirror-like surfaces all over the place. This is the typical first step to ray-tracing reflections but if you have it all over the place then it just looks bad.
A much better approach is to spend more of the ray budget to cast multiple rays around that reflection vector (i.e. Glossy in the image) to get blurred reflections, but these consoles don't have the bandwidth.
This would make the games look way more like real life because most surfaces that reflect aren't mirrors.
Some company may do some tricks to blur the reflection I suppose.
One other thing to note. With every local ray cast for a color of an object, a developer will try to avoid evaluating various levels of complexity in the shader. You see that with this girl walking in the mirror. It looks wrong to see her fully reflecting her environment but you don't see that reflection level in the mirror. They compleley just used her diffuse contribution. You will see a LOT of this in upcoming games. This should be your clue that we are still many generations away from having the bandwidth to render even reflections within reflections.
So keep an eye out for the things I mentioned. They will become readily noticeable now that you know.
Of course I agree we have to be realistic about consoles, I mean only a few months ago we were at Ps5 would struggle to do 4K native and there would be no ray tracing just screen space if that.
They have to use economical means to do this stuff, its just not feasible unless you drop to 1080 and blow the whole frame budget on rays and even then. I am sure the 20 + TF beasts coming on pC will have something to say, but for consoles this would be a huge performance benefit.
Last edited:
yurinka
Member
Yes, they confirmed that Spiderman: Miles Morales was going to use raytracing.
Raytraced puddles, like in Spiderman: Miles Morales!
But in the PS4 age Marvel and Sony signed to handle Spiderman as a Sony published brand getting their 1st party treatment, even if we don't know during how many years or how many games. We saw that the deal allows to use the character in other Marvel games for other platforms, like in Marvel Ultimate Alliance 3.
Raytraced puddles, like in Spiderman: Miles Morales!
Yes, it was in the past.
But in the PS4 age Marvel and Sony signed to handle Spiderman as a Sony published brand getting their 1st party treatment, even if we don't know during how many years or how many games. We saw that the deal allows to use the character in other Marvel games for other platforms, like in Marvel Ultimate Alliance 3.
Last edited:
geordiemp
Member
And what else has been shown for consoles, and what do you think we will get ?
Expectations are strange, we have only seen snippets of next gen like Miles, Ratchet with clank and his shiny head, a few shiny floors reflecting in Ratchet gameplay and Halo with shiny craig lol of course. What do you expect ?
I am quite happy with Miles morales 4k60 dynamic temporal and a few puddles here and there / shiny stuff just to upset Alex at DF. But lets not get carried away, reality, and its much more than anyone predicted or believed not so long ago.....
Last edited:
yurinka
Member
They had that announcement trailer and released a few promotional screenshots since them. They seem to be from in-engine cutscenes using raytracing, but who knows it they are CG or in-engine running in a NASA pc instead of in the console devkit.
But according to Sony, everything shown in their PS5 games conference was captured from the console (devkit), so in theory what we saw from Ratchet, Gran Turismo VII, Horizon 2, Spiderman: Miles Morales, Project Athia, Pragmata and so on was captured from a PS5 devkit. And there was a lot of raytracing there. I suggest you to watch the PS5 announcement stream, there was a good amount of pretty games there.
Regarding the RT puddles, they mentioned it in an Insomniac account tweet and as I remember they were visible in the assets I mentioned.
We know that Spiderman: Miles Morales will have an optional performance 4K 60fps mode:
But who knows if it's going to be native 4K or not or if the 4K 60fps will have raytracing enabled. In any case, I prefer games like this to run at solid 30fps and not native 4K but instead look extra beautiful. If to go native 4K 60fps means to look ugly as Halo Infinite, I prefer them to avoid native 4K 60fps.
I assume that they will have some kind of next gen improved solution of what checkerboard upscaling was for PS4 Pro, something that may help them to render at a smaller resolution but to look way better than checkerboard.
And even if they don't use tricks like these ones, remember that people said that was going to be impossible to see native 4K 60fps on PS4 Pro and we saw many of them, and even more with non-native 4K (which is ok for 99% of the people because without watching DF or similar videos people don't spot the difference or care about it).
Last edited:
MastaKiiLA
Member
How do game engines incorporate RT? Are they shooting rays to all visible surfaces, or are they only shooting them at visible reflective surfaces? Based on VFX's post about simulating accurate light scatter, wouldn't the preferred shortcut be to assign each visible reflective surface a roughness value that then simulates diffusion via opacity or alpha value, rather than actually generating multiple rays off said reflective surface?
Given a cityscape, for example, it seems pointless to waste your ray budget shooting rays at street or stone surfaces. They're not going to reflect anything. And you should be able to quantify the number of reflective surfaces that are visible in the scene, and focus your ray budget on those, to improve both granularity of the effect, as well efficiency. I'd assume as well that in the case where Spiderman is on the side of a glass building, with the entire view comprising the character and effectively a large mirror, that you'd distribute your available rays across that surface, in such a way that the densest allocation is where you expect the focal point to be.
As you can tell, I'm not a game coder, so I'm just trying to figure out the conceptual mechanics of a RT process in a 3D environment. I appreciate any clarity that can be brought to my current state of confusion.
Given a cityscape, for example, it seems pointless to waste your ray budget shooting rays at street or stone surfaces. They're not going to reflect anything. And you should be able to quantify the number of reflective surfaces that are visible in the scene, and focus your ray budget on those, to improve both granularity of the effect, as well efficiency. I'd assume as well that in the case where Spiderman is on the side of a glass building, with the entire view comprising the character and effectively a large mirror, that you'd distribute your available rays across that surface, in such a way that the densest allocation is where you expect the focal point to be.
As you can tell, I'm not a game coder, so I'm just trying to figure out the conceptual mechanics of a RT process in a 3D environment. I appreciate any clarity that can be brought to my current state of confusion.
Elektro Demon
Banned
Fake
Member
Actually, Sony can choise not use both of them. The PRO use checkerboard rendering because he have a inbuild hardware to doing so. IDK if PS5 have such an exclusive hardware to do, but is probably by the sucess of Nvidia both Sony and Microsoft would use an AI reconstruct tech. Will save loads of perfomance at the cost of almost zero graphic loss.
I suggest you watch the Death Stranding PC version from Digital Foundry. Thats is what Sony and Microsoft need to use next gen because is amazing.
sircaw
Banned
onesvenus
Member
The thing is that you don't know what a given pixel is seeing before shooting a ray. When you shot a ray and find those kind of surfaces you won't need as much rays as if it were a reflective surface but you'd need more than one to compute a good lighting approximation
VFXVeteran
Banned
Yes. Every surface will have a shader attached to it. The shaders that are purely diffuse with no specular+reflection will have that checkbox set to OFF. It does no good to evaluate the reflection term when there is no reflection.
Polygonal_Sprite
Gold Member
I can't even imagine wanting native 4k/30fps over dynamic / temporal injection 4k/60fps. How close do you sit from your display lol. I wear glasses and struggle to see the difference between 'fake 4k' and native. Spider-Man will play sooo much better at 60fps too due to the type of game it is. That's what options are for I guess
geordiemp
Member
I can't even imagine wanting native 4k/30fps over dynamic / temporal injection 4k/60fps. How close do you sit from your display lol. I wear glasses and struggle to see the difference between 'fake 4k' and native. Spider-Man will play sooo much better at 60fps too due to the type of game it is. That's what options are for I guess
Yes options are good, I would also pick 4k60 dynamic. The main point of the thread is some Ray tracing could be possible on consoles at low frame time cost.
RaySoft
Member
The difference in Navi 2's RT and Nvidia's RT seems to be that Nvidia has a dedicated "RT core" meaning they still has the normal hardware to render a scene and a dedicated block to do RT.
On the AMD side, it seems it's more up to the devs how to distribute the power. Every CU's has RT capabilities, so the more CU's you reserve for RT, the better it gets, but on the cost of everything else.
Which is better is a good discussion, but In my mind I think AMD's solution is better. At least it's easier to reach peak perf. (no hardware left unused)
Last edited:
MastaKiiLA
Member
What do you mean by you'd need more than 1? Do you mean more than 1 ray? Are you referring to the direct bounce, or the diffused array that VFX noted in his post? Sorry, just want to make sure I fully understand your response.
Thanks. For realtime gameplay, I'm just wondering how valuable the diffuse or glossy ray bounces in your earlier post would be for the end result, as opposed to the single ray approach shown for the specular reflection. As opposed to cinematics, where fidelity is generally more important. Would the end result be poor enough, in your opinion, to make SSR a better option? I have only rough idea of how SSR works anyway.
I appreciate the feedback, guys. I'm just looking to learn more, but if the explanation is too lengthy to answer my questions, you don't have to bother.
BootsLoader
Banned
With such music I thought the robot will have some porn with the statue. Wtf they were thinking of when choosin the music.
Elog
Member
I am. Pathtraced Minecraft at 1080p (which is what the XSX data is from) gives roughly 45 FPS on a 2060 super card.
That is where Pathtraced Minecraft on the XSX is according to MS. The snippets we have seen though are slightly worse than that so who knows what the truth is. Point is - that is really early RT capabilities from last generation.
This contrasts quite starkly to what we have been seen pieces of from the PS5.
The Minecraft performance might be down to lack of software optimisation but why would MS show something they do not believe represent the capabilities of the machine?
Last edited:
I Love Rock 'n' Roll
Member
It makes total sense. I just can't wait to know more. Thank you OP.So what we know so far, people can join the dots however they please
People can believe whatever they want, if its not logical and thoughtful then they embarrass themselves.
- LocalRay company have announced a deal with a console manufacturer which is not announced yet and under NDA
- Sony have 3 RECENT games, Spiderman, Miles Morales and Avengers exclusive DLC using this IP
- Sony make the spiderman films
- New photos for New York Spiderman demo 2020
- Video and the full New York spiderman demo has not been released yet,,,mmm
- Spiderman getting upgrade on ps5, Miles Morales 4k60 and shows RT in the short teasers.
- Insomniac said on Twitter they have ray tracing in the game
This technology works with little performance cost, as seen on the phone demo,. and as such if employed widely would be a massive boost for console games that use it over traditional FP 32 BVH.
Most games shown at the sony first party conference had ray tracing and were native 4K, which would kill a 2080ti FPS using traditional BVH, so something is up s Ps5 is punching well above its weight.
Imagine next spideman game with these RT, it would be glorious. I wonder when the NDA and demo for this will get released and at what show ?
funking giblet
Member
i would use a stochastic approximation to generate my output based on simple hyper-parameters in monte carlo computation of output vectors., It's not accurate, but it's certainly able to output something comparable to any method, although it's helpful to use the additional speed to generate more temporarily stable images, rather than push framerates.OK.
So here's the deal:
You can get incredible performance during reflection ray-tracing by assuming every shader has a mirror reflected ray. That means you don't need to cast multiple rays around the reflected vector.
But that is not good enough. You avoid all the extra rays around that reflected vector which is more to real life. So what happens is you got a game that shows mirror-like surfaces all over the place. This is the typical first step to ray-tracing reflections but if you have it all over the place then it just looks bad.
A much better approach is to spend more of the ray budget to cast multiple rays around that reflection vector (i.e. Glossy in the image) to get blurred reflections, but these consoles don't have the bandwidth.
This would make the games look way more like real life because most surfaces that reflect aren't mirrors.
Some company may do some tricks to blur the reflection I suppose.
One other thing to note. With every local ray cast for a color of an object, a developer will try to avoid evaluating various levels of complexity in the shader. You see that with this girl walking in the mirror. It looks wrong to see her fully reflecting her environment but you don't see that reflection level in the mirror. They compleley just used her diffuse contribution. You will see a LOT of this in upcoming games. This should be your clue that we are still many generations away from having the bandwidth to render even reflections within reflections.
So keep an eye out for the things I mentioned. They will become readily noticeable now that you know.
I'm not sure what the above relates to in an actual method of solving the above, as they are all valid outputs, the idea being to output the right vectors for the right angle and material. Raytracing Mirrors shouldn't use glossy reflections.
Ray Tracing in modern Consoles, and GPUs for that matter is a jittery mess.
Elog
Member
Short question since you seem knowledgeable: Is the reason for the mess the fact that modern GPUs try to do two things at once? The old-school rendering pipeline as well as RT at the same time? I.e. a card that actually just ditched the old pipeline and went all in on RT would probably be the right thing to do (with the caveat of being bad at running 'old' games...).
onesvenus
Member
Sorry for not being clear enough.
Let's put ourselves in the situation of having shot a ray and intersected a non-reflective surface. To know which color that ray contributes with we need to at least shot a ray from there to each light and see for example if they are in shadow for a given light or not.
Even if we had a single light and found that surface to be in shadow, we would like to shot multiple rays to random directions to account for the diffuse lighting that's getting there.
There are a lot of ways to optimize all these processes and limit the number of needed rays but shooting only one ray per pixel is not enough to have convincing shading.
If you want to learn more about this, take a look at Peter Shirley's raytracing course here: https://raytracing.github.io/books/RayTracingInOneWeekend.html#diffusematerials
Even if coding is not your thing there are a lot of helpful explanations there.
ShadowWolf712
Member
Who is fam ? Is that an american thing ?
THe new info is localray have done a NDA deal with 1 mobile manufacterer and one console manufacterer and there is new info and screenshots. And no it is not just geared towards the mobile market
PC & Laptops | Adshir | Local Ray
And the 2020 demo is not mobile, it says clearly its console / pc market
If that's not mobile then.... yikes.
Temporal correction may be (and usualy is) non-trivial.
SlimySnake
Flashless at the Golden Globes
not that i disagree with the rest of this post, but just wanted to clarify this. native 4k 30 fps with RT can be easily done on even a 2080 let alone a 2080 ti. its when you push to 60 fps where you thing start to get rough. Everything shown at the ps5 reveal was native 4k 30 fps with the exception of gt7.
geordiemp
Member
I was NOT knocking a 2080ti as it would be doing full FP32 RT, I was suggesting consoles need to cheat and take whatever short cuts they can. Lets face it, even PCs use DLSS which is upscaling.
Consoles are not going to have dedicated RT cores.
Last edited:
Iced Arcade
Member
That's exciting, spiderman 2 on the PS5 is going to be amazing
Sony doesn't own spiderman, they 100% own the movie rights to Spiderman thoughSony has owned the Spider-Man character since 1985, and just because they made a deal with Marvel doesn't mean they sold him back.
'Spider-Man' returns to the Marvel universe as Sony, Disney strike new deal
The pact comes a month after the two studios parted ways amid a reported disagreement over profit-sharing.www.nbcnews.com
RaySoft
Member
Sources?
Jtibh
Banned
Ass lickerThank you for enriching the thread with your inputs.