pawel86ck
Banned
There is even one game that use tensor cores for denoising?
-
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.
- Status
Dim
Banned
I am not against what Cerny says,i am against the free translation from some people that say CERNY SAY WE HAVE HARDWARE RT IN PS5
Is different to have dedicated cores for RT only and different to have a hybrid solution
Vega support hybrid solution but is not RTRT as RDNA 2
Why so many years nobody talk about Vega RT Hardware then?
Dim
Banned
BattleScar
Member
20 CUs are 43mm2. So each CU is roughly 2.15mm2. If RT adds 8-10% to each CU, then each would grow to between 2.33-2.36mm2. That would make 20CUs roughly 46-48mm2.
Still haven't factored in another two 32-bit GDDR6 Phys - which are quite large might I remind you.
I know the Render back end is the same. Which is why I never mentioned it.
Something about the math does t add up to me. The difference cannot be just 50mm2. Microsoft would have to be running an APU around 400mm2. Around 350-360 just doesn't seem feasible.
Panajev2001a
GAF's Pleasant Genius
AMD is likely going for adding additional HW to accelerate ray intersection tests and dynamically updated acceleration structures (or some extra GPU memory that can be dedicated to it according to developer needs) that store a description of the current scene. The idea is to reuse the same HW to do all sorts of computation that involve shooting a “ray” (or whatever the dev wants it to represent) reusing the same HW with a small bit of additional logic. GCN and RDNA == SW based raytracing, RDNA2 RT (portion of the RDNA2 specs) == HW accelerated RT.
nVIDIA’s solution is A ray tracing implementation hard coded in the GPU HW (in the form of RT cores) while AMD plans more reusable universal shader cores (hence why the RT support die area increase estimates are quite conservative).
Last edited:
Dim
Banned
GTX 1080 8.8TF have better perfomance than Vega 64 13 TF in games
Thank you again, my mistake
Gamernyc78
Banned
Gudji
Member
You don't know what you're talking about though. The document is not confirmation of anything.
TheStruggler
Report me for trolling ND/TLoU2 threads
Buying talent is great but MS has a leadership problem. Its the sane company that wouldbt cancel crackdown yet Phil kept saying its his baby and fav franchise yet it was brought out to die. As I am hopeful, MS needs to change internally more than anything else and establish an identity that isn't childish
Blindarchangel82
Member
To be fair, they don't have games to talk about. So they cling to the hardware like it's mommy's teat.
Rightisright
Banned
Go and rewatch it. it's animation blows anything this gen away.
Panajev2001a
GAF's Pleasant Genius
Vega has no hybrid solution, Vega supports using the compute shaders to perform raytracing in SW. Nobody would refer to GPU accelerating RT or RT acceleration in the GPU HW when talking about Radeon Rays.
RDNA2, from what AMD has said so far and from their parents, will have the mentioned GPU acceleration for RT. nVIDIA did “An” implementation for RT support using dedicated additional cores while AMD is adding logic to the current CU’s (shader cores and TMU’s).
If you consider RT cores the only way to have real HW RT then neither XSX nor PS5 will have real HW RT by your definition, happier?
Else you are implying a non technologist VP is meant to have more technically correct terminology than a software engineer and platform architect who has never been found to be misleading consumers or developers in his statements.
Last edited:
Darius87
Member
so confirming that's is not software also doesn't mean it's hardware RT?
what's comes down to what you call RTRT is it have ray - intersection engines which means dedicated transistors on silicon, we all know cerny denied software based rendering
that means PS5 indeed hardware RTRT, if confirming
not good enough for you
now question is same solution as RDNA 2.0 is or some custom variant by sony.
it_wasn't_me
Banned
Additional 64bit for bus is ~16mm².
So 46-48mm² + 16mm² = 64mm².
That doesnt exactly fit, but if XSX does not have 3D audio hardware and obviously no DDR4 then its still some mm² back (although no doubt less then 14mm²)
Again, its estimate from him. He said Navi 10 is 250-255mm² back in January and big Navi is 505mm² and now everyone is quoting him. He has long history of TSMC leaks...
Rightisright
Banned
Also last gen Sony made changes to the Cu's on their GPUs which meant they were larger in size than normal. Not sure if they will do the same thing this time.
bitbydeath
Gold Member
So you’re admitting the AMD data leak is bogus?
Reindeer
Member
Sorry, would have to to disagree. There's nothing there that Naughty Dog or other top developers have not been able to accomplish.
BattleScar
Member
Dim
Banned
I am not repeating friend
Random people blame me the same things about Cerny
And i just say Cerny use the same words AMD use when reveal to people the Radeon Rays
RT acceleration built in in GPU
I cant find now the Radeon Rays reveal post from AMD but take a look at AMD description what is RT acceleration built in support in GPU
Radeon™ Rays
The lightweight accelerated ray intersection library for DirectX®12 and Vulkan®.
gpuopen.com
I just every time i read inside here Cerny says i make search to find any new statement diferent that obvious
Maybe some people need to learn what really Cerny says and that his statement was not clear in HW RT or Hybrid Solution in PS5
Is different thing to talk about RTRT and RT aceleration in many things as final GPU performance and RT perfomance
AM i clear to you now?
Nope. I am saying your reasoning is bogus.
Dim
Banned
Radeon™ Rays
The lightweight accelerated ray intersection library for DirectX®12 and Vulkan®.
gpuopen.com
Amd say it not me
bitbydeath
Gold Member
I’ve been constantly referring you to the data and asking you for proof of your claims which you are yet to provide.
If you’re just running off a gut feeling just say so, otherwise provide the receipts for your claims and I don’t mean PS4 APU information.
Dim
Banned
Rightisright
Banned
If you havent already seen those changes from the launch of the OG Xbox One and now, you haven't been looking.
Reindeer
Member
Apparently Crackdown 4 might be happening lol. I myself enjoy Crackdown (guilty pleasure) so the more the merrier!
Dim
Banned
dedicated hardware raytracing is that who AMD call in RDNA2 RTRT and that RDNA 1 dont have it and use the hybrid solution
Means that HW RTRT dont affect the actual performance of GPU and RT have better perfomance
Dim
Banned
TheStruggler
Report me for trolling ND/TLoU2 threads
The most important thing to a console is great games and yet they have lacked all generation, they may have changed in one way but completely missed the mark
No, you have been constantly told to stop being obtusive and check the data for yourself, which you reject for a week.
There have been multiple sites with access to it and giving you details (among them DF), so for you to still be so obtusive without even looking into file is actually incredible.
Believe what you want to believe. Data, while deleted from twitter and github, is still there. Anyone can find it and check for themselves, so can you.
Reindeer
Member
I think they'll have game going forward, but they're way too much about live service and multiplayer these days.
from what I remember there’s a single unit per TMU called the “TD” in the AMD patent. It seems to handle both BVH and triangle intersections, but does rely on the shader units for some parts of the pipeline it seems.
BattleScar
Member
That's fine I guess. But isn't there some information out there implying that the PS5 is somewhere around 316mm2. If I recall correctly, its the one with the detailed memory configuration leaks (overkill VRM, the chip layout etc.).
Regardless, ~300mm2 for PS5 and ~350mm2 for XSX is the window. That's not a lot of room to play with. The APU by my estimations should be closer to 400mm2. I don't see how 350mm2 is enough for 56CU with RT hardware and the 8 Zen2 cores.
No, he's been saying that Oberon is a newer version of Ariel. He's been saying they're the same fundamental design, but at different stages of completeness. Ariel was the first prototype and then Oberon is a revised and slightly improved prototype.
They're is no evidence to suggest that the PS5 will use two GPUs.
Just because there are two codenames, doesn't mean that both chips are going to be used in the same machine. That is faulty logic.
bitbydeath
Gold Member
I have found it and I’m calling you out because you are trying to twist the data and are providing nothing to back yourself up.
You won’t provide data because you can’t, it simply doesn’t exist.
bitbydeath
Gold Member
The data states there are two GPU’s.
There is no mention at all that I’ve found of one being an early version, it has only been inferred that way because it sounds crazy.
And the fact the data still exists together in unison is unusual.
I’m not stating there are two, i’m saying the data is a mess and that can clearly be seen with Arden popping up under Oberon.
Reindeer
Member
Arden is Series X, was confirmed by way too many knowledgeable people. It also has BC for XB1 and XB1X.
Last edited:
Yea obviously, although I have to make two comments.
1. From Scarlett video, APU calculation (by using memory chips as referrence) seems to be bigger then 350. Closer to 360-370mm² (in line with Xbone and X)
2. Zen2 with 1/4th of L3 cache (as predicted by leaked Flute benchmark and now notebook 8 core Zen 2) is likely smaller then 40mm². So, less then Jaguar with its L3...
I think you can easily fit 40CUs + RT + Zen2 inside 320mm².
Navi 10 with 256bit bus = 251mm²
Zen2 1/4th of L3 cache = ~35-40mm²
RT ~10% for each CU = 10mm²
Audio 3D engine + bus DDR4 = rest
So, easily fits.
What is also interesting is that PCB leak from May not only fits timeline of when chip for dev kit could be released, but also fits in with when V dev kit was released (around June - July for devs).
Also, this leak specified chip size at 22.4 * 14.1mm.
You know what has ~14mm² width? Navi 10 chip. Funny because this was leaked before any Navi was shown or released.
BattleScar
Member
Which has been explained as being someone fucking up their Python code and calling up the wrong codename.
Frankly, nothing is certain. But dual GPU, just cause there are two identical chips - one of which is clearly stated in the file as having improvements to cache bandwidth compared to the other - seems rather ludicrous to me.
Why would you have one GPU with 448GB/s bandwidth and the second having >530GB/s bandwidth. Doesn't make sense.
bitbydeath
Gold Member
Yes, I agree with you.
The ‘AMD leak’ has a lot of holes in it, that’s just one of them.
You are lying. There have been mutliple sites reporting on this, but you are calling everyone dumb but yourself (who doesnt even have a file or care to take a look).
Fake
Gold Member
This is getting a bit tiring at this point. Lost the count of how many times someone need to put the link of the wired article.
Last edited:
Reindeer
Member
I think having BC for Xbox is pretty much conclusive that it's not a PS5 APU.
Last edited:
bitbydeath
Gold Member
Wrong code name but the filepath still states Oberon as well. It should be calling all that information in a block, not line by line.
It doesn’t make sense and that’s my point.
Dim
Banned
You wrong.
After the great succes of Sony in BC with PS4-PS3-Ps2-PS1 the BC team announce that PS5 is BC with Xbobx One too
xool
Member
2x 36CUs @ 1.55GHz that's ~180W just for GPU , ~ 300W including RAM and CPU
I don't know why I'm even considering this - and yet ...
But surely Crossfire never worked well
bitbydeath
Gold Member
Did you post the wrong image? That doesn’t state they are the same GPU. Again you’re inferring the results you desire.
bitbydeath
Gold Member
I believe the whole output is false, that’s just the tip of the iceberg for what’s wrong with it.
- Status