Discussion RDNA4 + CDNA3 Architectures Thread

[DisEnchantment]

With the GFX940 patches in full swing since first week of March, it is looking like MI300 is not far in the distant future!
Usually AMD takes around 3Qs to get the support in LLVM and amdgpu. Lately, since RDNA2 the window they push to add support for new devices is much reduced to prevent leaks.
But looking at the flurry of code in LLVM, it is a lot of commits. Maybe because US Govt is starting to prepare the SW environment for El Capitan (Maybe to avoid slow bring up situation like Frontier for example)

See here for the GFX940 specific commits

History for llvm/lib/Target/AMDGPU - llvm/llvm-project

The LLVM Project is a collection of modular and reusable compiler and toolchain technologies. - History for llvm/lib/Target/AMDGPU - llvm/llvm-project

github.com

Or Phoronix

More AMD "GFX940" Enablement Work Landing In LLVM - Phoronix

www.phoronix.com

There is a lot more if you know whom to follow in LLVM review chains (before getting merged to github), but I am not going to link AMD employees.

I am starting to think MI300 will launch around the same time like Hopper probably only a couple of months later!
Although I believe Hopper had problems not having a host CPU capable of doing PCIe 5 in the very near future therefore it might have gotten pushed back a bit until SPR and Genoa arrives later in 2022.
If PVC slips again I believe MI300 could launch before it

This is nuts, MI100/200/300 cadence is impressive.



Previous thread on CDNA2 and RDNA3 here

Question - Speculation: RDNA3 + CDNA2 Architectures Thread

Man I have been dying to make this one for a while now. First rumours for RDNA3 are here so new thread time! Just going to start off with this one for now: kopite7kimi on Twitter: "@VideoCardz Ah, I mean a simple mcm design with 10240 cores is not enough. Because the lift from RDNA2 to RDNA3...

forums.anandtech.com

 

[adroc_thurston]

I am just asking for model numbering, be it RX7000 or RX8000?

Why would they insert wholly different microarchitecture into an existing lineup?

Is that so hard for you to answer?

Numbers don't matter.

 

[Tigerick]

Why would they insert wholly different microarchitecture into an existing lineup?

Numbers don't matter.

Cause as I explained, the architecture difference is not as big as you said, it is just die shrink of N32 with integrated IF and GDDR6. If AMD able to launch new card with GDDR7, why the delay of RDNA5???

It is alright to have different opinions, I am inclined to believe RDNA4 cards to be called RX7000 series. And you seem to believe it is going to be RX8000 series like others. Time will tell...but next time don't simply call people dumb just because we have different opinions cause I am pretty confident about AMD's namings haha

 

[adroc_thurston]

Cause as I explained, the architecture difference is not as big as you said

Ughhhh what.

it is just die shrink of N32 with integrated IF and GDDR6

What makes you say that?

why the delay of RDNA5???

delay of what

It is alright to have different opinions, I am inclined to believe RDNA4 cards to be called RX7000 series. And you seem to believe it is going to be RX8000 series like others. Time will tell...but next time don't simply call people dumb just because we have different opinions cause I am pretty confident about AMD's namings haha

This really sounds like an unmedicated rambling.

 

[Ajay]

I am talking about a completely different topic

Oops, sorry - guess I lost track of the thread within the thread.

 

[TESKATLIPOKA]

RT is more memory constrained vs raster.

I don't think simply increasing clocks 50% would be a guaranteed 50% win for RT perf without memory improvements.

Of course, It would need more BW. Let's welcome GDDR7.

 

[dr1337]

This really sounds like an unmedicated rambling.

If AMD markets zen 2 and zen 3 along side zen 4 as 7xxx series in laptops, they could start doing the same thing in GPUs. Fact of the matter is their speculation is as baseless as yours. And frankly you shouldn't be one to judge with your endless stream of vague and unhelpful single sentence quips.

I am just asking for model numbering, be it RX7000 or RX8000? Is that so hard for you to answer?

He doesn't know anything really, and nobody will until we're close to the announcement. AMD almost branded the 5700xt as the rx 690, sloppily leaving traces of this in their E3 2019 slides. I really doubt they're going to do the laptop thing to their GPU branding but things are technically still up in the air.

I definitely don't believe the MLID rumor millers that RDNA 4 is an eco round. But doing another rx 480 type launch isn't out of AMDs cards and with such an idea would also potentially come a branding reshuffle.

 

[adroc_thurston]

If AMD markets zen 2 and zen 3 along side zen 4 as 7xxx series in laptops, they could start doing the same thing in GPU

GPU marketing has absolutely nothing to do with the CPU marketing.
No cross-pollination there.

He doesn't know anything really, and nobody will until we're close to the announcement

someone's very upset.
need help?

Fact of the matter is their speculation is as baseless as yours

Very upset.
No need to cry.

Trolling and personal attacks are not allowed. You need to stop being so toxic. -AT Moderator Shmee

 

[Joe NYC]

I am just asking for model numbering, be it RX7000 or RX8000? Is that so hard for you to answer? Or simply I don't know also matter...

Marketing still has to burn a few 100s hours on this one before even they decide...

 

[linkgoron]

Cause as I explained, the architecture difference is not as big as you said, it is just die shrink of N32 with integrated IF and GDDR6. If AMD able to launch new card with GDDR7, why the delay of RDNA5???

It is alright to have different opinions, I am inclined to believe RDNA4 cards to be called RX7000 series. And you seem to believe it is going to be RX8000 series like others. Time will tell...but next time don't simply call people dumb just because we have different opinions cause I am pretty confident about AMD's namings haha

Marketing-wise it's always better to increase the numbers, even if technically the changes are not large. Just look at Intel's 14th gen, or all of the Skylake gens.

 

[soresu]

If AMD markets zen 2 and zen 3 along side zen 4 as 7xxx series in laptops, they could start doing the same thing in GPUs

Wer u bin bruh? 😅

AMD have been doing that for years and years at the lower end of gfx SKUs.

Not sure if nV do it to, but it wouldn't surprise me.

 

[SolidQ]

Interesting patent for RDNA4 or RDNA5?

DISTRIBUTED GEOMETRY

Complete Patent Searching Database and Patent Data Analytics Services.

www.freepatentsonline.com

 

[soresu]

Interesting patent for RDNA4 or RDNA5?

DISTRIBUTED GEOMETRY

Complete Patent Searching Database and Patent Data Analytics Services.

www.freepatentsonline.com

Took less than 4 months from filing to publishing so it clearly didn't have much in the way of competition for originality.

 

[branch_suggestion]

N10 and N22 used the same shader config and the same node family, that is where the similarities end.
Just because the config looks similar doesn't mean the uArch is similar.

 

[SolidQ]

Just because the config looks similar doesn't mean the uArch is similar.

looking at this. Seems cyberpunk benefit from RDNA3 uarch, compare 6800XT vs 7800XT

 

[TESKATLIPOKA]

looking at this. Seems cyberpunk benefit from RDNA3 uarch, compare 6800XT vs 7800XT

Yeah, there are some games, which profit from this dual-issue CU. CB2077 is one of such games If not the one with highest gains.
The problem is that there are only a few such games.

 

[Ajay]

Yeah, there are some games, which profit from this dual-issue CU. CB2077 is one of such games If not the one with highest gains.
The problem is that there are only a few such games.

I hope AMD can improve its dual issue use - it's a bit restricted at present. Nvidia, it seems to me, did a bit better with this. Of course, this may have only been in games optimized to maximize NV's dual issue abilities.

 

[Glo.]

It is alright to have different opinions, I am inclined to believe RDNA4 cards to be called RX7000 series. And you seem to believe it is going to be RX8000 series like others. Time will tell...but next time don't simply call people dumb just because we have different opinions cause I am pretty confident about AMD's namings haha

Architecture of 7000 series is called RDNA3.

We are in the thread of RDNA4, and you suggest its going to be called 7000 series... again?

 

[Glo.]

Interesting patent for RDNA4 or RDNA5?

DISTRIBUTED GEOMETRY

Complete Patent Searching Database and Patent Data Analytics Services.

www.freepatentsonline.com

It is interesting because it allows not only to create multidie dGPUs, but also... gaming APUs.

 

[igor_kavinski]

It is interesting because it allows not only to create multidie dGPUs, but also... gaming APUs.

What do you say to that, Mr. @adroc_thurston ?

 

[GodisanAtheist]

What do you say to that, Mr. @adroc_thurston ?

-He says "short, vague, non-committal response that sounds confident but can be interpreted 12 ways from Sunday based on your own personal biases".

 

[Glo.]

What do you say to that, Mr. @adroc_thurston ?

Well you do not have to be adroc to understand the patent, and its consequences.

Its architectural patent, not just physical design patent. So it can be used in multiple ways, as long as you have compatible IP.

So if this is about RDNA GPU architectures - it will be possible to be used in both dGPUs, and APUs.

And now we can have technical discussion about what is required for multidie GPU chiplet to work, and how can it be implemented.

 

[Tuna-Fish]

RX7700XT?

There is just no way, no how they are going to launch a new set of cards whose names are so hard to distinguish from the old ones. Plenty of people in this thread are very misguided about what drives naming. It's not architecture, mm², process node, common sense, or any of that. It's that AIBs and especially laptop OEMs want to release a new lineup of gear every year, with new marketing names on it that are clearly visually "better" than what was sold last year. What they absolutely do not want, is to have sold boxes "7700" printed on the front this year and have to sell ones that have "7700 XT" next year. "8700" is just so much better, see, it has a bigger number!

If AMD keeps some RDNA3 cards in the lineup, the solution isn't naming the new cards to fit the old lineup, it's to rebadge old cards to fit the new lineup. That way, there's no trouble fitting anything anywhere, you can just decide what models you keep and give them appropriate numbers. No connection or logic to the old lineup is required, see for example Pitcairn, which was sold as the: { HD 7870 XT, HD 8870, R9 270X, R9 370X }, and that's just the top cuts of it in each stack.

And yes, they absolutely can throw out "7800 XT" from their active lineup after not even a year, either in reality or just in name, and it wouldn't even be their shortest-lived model.

To be clear, I have no specific information about RDNA4, but I can honestly tell you that marketing execs would recoil from your proposal like a vampire recoils from sunlight.

 

[beginner99]

Architecture of 7000 series is called RDNA3.

We are in the thread of RDNA4, and you suggest its going to be called 7000 series... again?

Why shouldn't it be? It would not be the first time different architectures appear under the same series and especially AMD on the CPU side is doing this all over the place. I will actually say that it would also be better for the consumer to use a 7700 series name IF the chips slot in somewhere between n32 and n33. a 7700 and 7600XT. this would be very clear for consumers how they fit into the stack in terms of performance. Since there will be not other RDNA4 part, why make it a new series? naming them 8600 and 8700 for example would be way more confusing as then it's not clear a 7700xt is faster as a 8700.

And given what we do know about die sizes and the fact n32 was launched and these should launch Q1 2024, I can simply not see they actually outperform N32.

 

[PJVol]

Took less than 4 months from filing to publishing so it clearly didn't have much in the way of competition for originality.

Not quite

US11755336B2 - Distributed geometry - Google Patents

Systems, apparatuses, and methods for performing geometry work in parallel on multiple chiplets are disclosed. A system includes a chiplet processor with multiple chiplets for performing graphics work in parallel. Instead of having a central distributor to distribute work to the individual...

patents.google.com

 

[TESKATLIPOKA]

Why shouldn't it be? It would not be the first time different architectures appear under the same series and especially AMD on the CPU side is doing this all over the place. I will actually say that it would also be better for the consumer to use a 7700 series name IF the chips slot in somewhere between n32 and n33. a 7700 and 7600XT. this would be very clear for consumers how they fit into the stack in terms of performance. Since there will be not other RDNA4 part, why make it a new series? naming them 8600 and 8700 for example would be way more confusing as then it's not clear a 7700xt is faster as a 8700.

And given what we do know about die sizes and the fact n32 was launched and these should launch Q1 2024, I can simply not see they actually outperform N32.

When did a new architecture use the naming of the previous generation?