RDNA 2 supports hardware-based ray-tracing, a first for AMD. If Sony's using AMD's RT solution then of course it's going to be RDNA 2, just like it is for XSX and PC.Okay, that wasn't in the slides. Is the quote transcribed somewhere? I want to read it verbatim.
I knew that Sony was using AMD's raytracing but I didn't think they would be using RDNA2.
No the dude is being completely serious. Pretty sure it's an alt account for TimDog.
RDNA 2 supports hardware-based ray-tracing, a first for AMD. If Sony's using AMD's RT solution then of course it's going to be RDNA 2, just like it is for XSX and PC.
Interesting.What does the news change on the data in Github?
Speculation:
You could easily assume Oberon was already RDNA2 at the testing (just like Arden already was). You could also assume Ariel was actually a RDNA1 part which would make a lot of sense that we had Ariel data in there too for comparison. The data doesn't give us the answer and the news of today doesn't change anything for me. It still stated 36 CUs full count which is independent from architecture.
Oberon A0 is called Navi10 Lite. Lite just means APU. So Oberon is Navi10 which is contradiction with RDNA2.
As the doctor has said, probably an old chip Sony used to test/develop BC. Klee and Matt been knew.
That is just not true. If we just use linear logic then any calculation of ppw increase between GCN and RDNA1 doesn't come out to 50% gains.PPW is PPW. IPC is IPC. There is absolutely no reason to thin tat they said PPW but instead meant or arrived at it by what you are sayng instead.
Dude, do you seriously think that anyone with the name Tim is TimDog? <rolls eyes>
That guy is so obnoxious. I am on the west coast right now, that dude is on the East coast.
We are two different people.
Ms has collaborated with Nvidia (and likely AMD too) to create a RT api that can be used on Vulkan. Likely that Is the unified api AMD talked about. (And it means we should see the same RT apis across all platforms)So AMD said today that they co-architected the DirectX Ray Tracing API with Microsoft to take full advantage of their RDNA2 ray tracing solution. Does this mean that Sony will be rolling their own API to leverage the RDNA2 architecture for ray tracing since I don't think that they will be using DXR?
Aahhh...I think you're right.Doesn't Klee's game info strongly suggest the PS5 V dev kit has some form of RDNA 2 as the game had very impressive RT in June last year?
A lot of people on this hardcore gaming forum for sure. 95% of the people buying these things won't have any idea what kind of chip is in there.
What does the news change on the data in Github?
Speculation:
You could easily assume Oberon was already RDNA2 at the testing (just like Arden already was). You could also assume Ariel was actually a RDNA1 part which would make a lot of sense that we had Ariel data in there too for comparison. The data doesn't give us the answer and the news of today doesn't change anything for me. It still stated 36 CUs full count which is independent from architecture.
Yeah, it may have been expensive for them to do but investing that money to make sure every person that bought a PS4 can bring their library with them is huge and clearly Sony sees it as important.I originally said that this was most likely the case as well. This really isn't that surprising to be honest.
We should probably be viewing it through the lens of Sony investing in their hardware-based bc solution going forward.
Do we know when Arden was tested?What does the news change on the data in Github?
Speculation:
You could easily assume Oberon was already RDNA2 at the testing (just like Arden already was). You could also assume Ariel was actually a RDNA1 part which would make a lot of sense that we had Ariel data in there too for comparison. The data doesn't give us the answer and the news of today doesn't change anything for me. It still stated 36 CUs full count which is independent from architecture.
The 50% GCN bump was bogus anyway because they were counting process gains too. RDNA 1 to RDNA 2 is 50% on the same process. IPC is baked into both numbers, per AMD's breakdown of said gains.That is just not true. If we just use linear logic then any calculation of ppw increase between GCN and RDNA1 doesn't come out to 50% gains.
If we look at it from a Teraflop per watt the RX 5700 XT (RDNA1) has barely better power per watt than the Vega 64 (GCN). the 5700 XT runs at 225w to create 9.75 TFLOPS, or 0.043 TF per watt. The Vega 64 runs at 295W to create 12.66 TFLOPS, or 0.0429 TF per watt.
Of course that makes no sense because it doesn't take into account IPC gains from one architecture to the next. As shown by almost all benchmarks 5700 XT performance in games is 10-15% better even though it has a large TF "deficit".
Maybe we could calculate the CU power per watt? Still doesn't work out to 50% because there is no perfect metric for us to compare two different architectures.....
Here are the two most powerful GPUs from AMD GCN and RDNA architectures respectively:
Radeon RX Vega 64
Next-Gen CUs 64
Base Clock Rate 1247 MHz
Boost Clock Rate 1546 MHz
Peak FP32 Perf. 12.66 TFLOPS
Board Power 295W
AMD Radeon™ RX 5700 XT
Next-Gen CUs 40
Base Clock Rate 1605 MHz
Boost Clock Rate 1905 Mhz
Peak FP32 Perf. 9.75 TFLOPS
Board Power 225W
I'm really curious how you would calculate these in some linear method to come up with 50% ppw gains and completely ignore IPC......
I can't believe a developmental APU unearthed over a year ago isn't representative of final hardware to release 2 years later. Just flabbergasted.
What does the news change on the data in Github?
Speculation:
You could easily assume Oberon was already RDNA2 at the testing (just like Arden already was). You could also assume Ariel was actually a RDNA1 part which would make a lot of sense that we had Ariel data in there too for comparison. The data doesn't give us the answer and the news of today doesn't change anything for me. It still stated 36 CUs full count which is independent from architecture.
Now, I don't speak fanboy, so please indulge me... Why would they be referring to a console that isn't even officially confirmed to exist?Not yet. PlayStation doesn't use DXR 1.1, only Microsoft related platforms. Consoles mean Xbox Series X/S.
We do not have any measured Arden test results.
Matt told us months ago to move past the GitHub leak, but some people can't really leave the past behind lol
Next stop, 11-12TF
Everyone needs to realize, we all win. :) There is no need for fighting imo.
Yup, a certain poster went as far as tell me I had no idea what I was talking about and how my contributions to these threads were average at best. Because I was saying over 200W consoles.Of course everything there was call crazy, stupid etc etc by people who though they knew better .
Man the last 2 years have been very funny in retrospect if you look at what certain people were saying .
Operon far ahead of Arden is a hot take I actually had to chuckle about.I think it's fairly obvious to everyone now that Oberon is part of the RDNA ariel/gonzalo/flute line.
The only hope for team GitHub was that PS5 was RDNA1 based.
It isn't. So pack it up folks, Github is dead. Long live GitHub.
Cling to your GitHub all you want, you still haven't answered my questions. So I'll ask them again.
Why did Sony develop a RDNA1 line of chips in gonzalo/flute/ariel if PS5 is RDNA2 based?
Why do you assume Oberon is not part of this RDNA1 line, despite so many similarities and literally no confirmation of a single RDNA2 feature?
Why is Sony's RDNA2 development so far ahead of MS's since Oberon was far and away more developed than Arden at the same point in time when they're using the same tech from the same company? Logically if they were both using RDNA2 and releasing at the same time they would be in very similar states of development. They weren't. Almost as if Oberon was a RDNA1 based evolution of ariel/flute/gonzalo...
Data is useful Colbert, but you also have to interpret it properly and look at it critically. GitHub's Oberon as final PS5 silicon hasn't made sense for a variety of reasons for some time. As time goes on it makes less and less sense. In fact the truth is that anyone claiming it is either RDNA2 or final PS5 silicon are basing those claims on precisely nothing. So if you're as data driven as you claim you should be very open to the likelihood that Oberon isn't RDNA 2, which all evidence points to, which means it isn't PS5's final APU.
AMD clarified the 7nm thing. They intentionally used a generic "7nm" without any regards to which version
In order to avoid confusion, AMD is dropping the '+' from its roadmaps. In speaking with AMD, the company confirmed that its next generations of 7nm products are likely to use process enhancements and the best high-performance libraries for the target market, however it is not explicity stating whether this would be N7P or N7+, just that it will be 'better' than the base N7 used in its first 7nm line.
Matt never said move past it, he said it's not confirmation of anything, which is true. Nothing can be confirmed until it happens.
" I'm glad to see we've moved pass thinking the Github leak was confirmation of anything. "
Pertaining to 1.) An AMD link was posted here over last weekend showing that in AMD terminology, letter changes are model revisions with new model numbers. The numbers are the steppings.I'm confused why people think RDNA2 being confirmed for PS5 changes much of anything. For final PS5 to be something other than the Oberon chip show in the GitHub data, one of the following has to be true:
- Oberon E0 vs Oberon B0 has very substantial architectural changes atypical for steppings.
- Oberon B0 shown running the back compat tests wasn't the full chip configuration (disabled SE/CU hypothesis, this is where I am).
- There's another, secret chip that we've never even heard a code name for that represents the final PS5, and Oberon was an entirely different line of development used only for BC (Doctor Avatar this is you, right?)
I don't see that PS5 being confirmed as RDNA2 makes any of those three possibilities more or less likely.
I was pretty sure it was not.
you haven't read the last two pages then...
I do agree, AMD is counting a huge amount of things to come up with their 50% number from GCN to RDNA1. They are counting process gains from 14nm to 7nm, IPC gains, and increased efficiencies in power required to push up clock speeds. If they really have a 50% PPW increase from RDNA1 to RDNA2 with a much more limited process gain (maybe they are talking about moving from 7nm to 7nm+?) then they must have some pretty great efficiency gains or they have found ways to massively increase clock speed with minimal power increase. Maybe both? Or all of the above? Can't wait for some benchmarks on actual RDNA2 products......The 50% GCN bump was bogus anyway because they were counting process gains too. RDNA 1 to RDNA 2 is 50% on the same process. IPC is baked into both numbers, per AMD's breakdown of said gains.
Sony will talk when they want to talk. Their silence I believe is not indicative of anything, just that as they said in the ps5 website they are just not quite ready to give more info right now, for whatever reasons.The obvious question is why is Sony not providing the answers here. If the PS5 is RDNA2, then why isn't anyone from Sony posting this out to the public?
Why do we need to hear this from a third party?
The marketing department from Sony such has changed a lot from 2013 and not in a good way.
You are being too nice. I say sometime today.
We were told, some didn't listen...Next-gen PS5 and next Xbox speculation launch thread |OT8| - The Dark Tower (See Staff Post)
I’m glad to see we’ve moved pass thinking the Github leak was confirmation of anything.www.resetera.com
Matt never said move past it, he said it's not confirmation of anything, which is true. Nothing can be confirmed until it happens.
I am 50/50 on Github stuff but it's been fascinating to see knowledgeable people argue about what it means