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

 

[MrPickins]

It's the x60 series competitor, just like Polaris was the 1060 competitor.

Again, only because of the upward trend in the marketing brackets. Didn't you see where I said the 4060 should be a 4050? In any traditional GPU market, these would be low-midrange parts.

I said I'd be happy to see another Polaris-type release; the 7600 was definitely not that, so no sense in even bringing it up.

 

[adroc_thurston]

In any traditional GPU market, these would be low-midrange parts.

The current market is the 'traditional' one.
Exotica is still HPC GPU realm.

I said I'd be happy to see another Polaris-type release

You aren't getting one, AMD has gross margin targets to adhere to.

 

[MrPickins]

The current market is the 'traditional' one.

It's really not, though. And I'm far from the only one that's said so.

 

[adroc_thurston]

It's really not, though

It is.
All the exotica remains contained to HPC parts like H100 or MI300.

And I'm far from the only one that's said so.

Yea whatever, you're still gonna buy it.

 

[MrPickins]

It is.
All the exotica remains contained to HPC parts like H100 or MI300.

Who is even talking about HPC?

I'm talking about current gen midrange consumer GPUs, which by all accounts have had their branding bumped up to the next tier without a corresponding increase in performance.

Yea whatever, you're still gonna buy it.

Considering I'm still using an RX 580 while I wait for an upgrade sweet-spot from either company, I'd say you're wrong.

But if this is how you want to discuss things, I'm not going to bother to respond to you. Have a nice day.

 

[adroc_thurston]

I'm talking about current gen midrange consumer GPUs

They're fine.

which by all accounts have had their branding bumped up to the next tier without a corresponding increase in performance.

Look at BOM.

Considering I'm still using an RX 580 while I wait for an upgrade sweet-spot from either company, I'd say you're wrong.

You'll be using it for 10 more years then.

 

[moinmoin]

they are basically competing against themselves trying to convince previous owners to upgrade

That's what all the iterations of DLSS so far are all about. And people are still eating it up to this day. Essentially customers want to buy expensive halo cards to generate fake frames.

These aren't exclusive. 1,3 &4 are almost the same thing with 2 also sharing tech to a lesser degree.

If you have to do 1,2&3, then 4 is not that an expensive proposition compared to only 4. Marginal costs.

It might mean slowing the multi core die strategy and the maximum size single core product. Means competing in smaller GPU core markets only. This drops your expenses a lot for the desktop card market and still allows you to service the largest sales segments.

Of course they are all sharing the same tech. But not to the degree that the Zen cores are shared and scaled across the whole product range. And that's the profitability benchmark RTG at some point will have to adhere to.

To even have a chance at being competitive (margin wise, not price wise) in the dGPU market requires halo entries that are costly to create just for this market. So either huge monolithic dies which AMD no longer does, or significant investment into disaggregated GPU dies which in case of RDNA3 was misdirected as the base architecture doesn't have the competitiveness they targeted.

 

[adroc_thurston]

To even have a chance at being competitive (margin wise, not price wise) in the dGPU market requires halo entries that are costly to create just for this market.

Yea they're building those anyway.

 

[GodisanAtheist]

Wonder if all this means is that mid-range RDNA 4 is being pulled forward (rather than respinning RDNA3 etc) to come sit on top of the RDNA 3 stack for a mid-gen refresh.

I can potentially see a fixed RDNA 4 mid-range part sitting on top of N31 by 20-30% if stuff is all fixed up and performing smoothly. Might make a good pipe cleaner part (remember those) before the full RDNA4 (maybe they'll just call it RDNA 5) stack is released later.

 

[dr1337]

IMO, this is another nonsense rumor. Only going to about Polaris level ( about RTX 5060 competition) next generation makes no sense at all.

I think the, eh, up and coming, rumor mill crowd has picked their wording very carefully if they actually have any insider knowledge. Given how these leakers have spoken before this, both seemingly very uncertain until very very recently, someone they apparently both trust must have *just* told them something.

Polaris was originally pitched as high end performance at low end costs. That doesn't seem unreasonable from AMD again as a "new" strategy. But then for some people to extend that all the way to say this is AMD giving up on consumer graphics, that sounds like clickbait-esque hype farming.

 

[adroc_thurston]

Wonder if all this means is that mid-range RDNA 4 is being pulled forward

The basic parts were always earlier than the plebs expected.

to come sit on top of the RDNA 3 stack for a mid-gen refresh.

Not happening either.

before the full RDNA4

No such thing.
It's two parts left.

That doesn't seem unreasonable from AMD again as a "new" strategy

Their strategy is unchanged: build something NV can't reach while adhering to strict TTM schedule, and if you can't then burn it and wait for the next cycle.

 

[Heartbreaker]

Again, only because of the upward trend in the marketing brackets. Didn't you see where I said the 4060 should be a 4050? In any traditional GPU market, these would be low-midrange parts.

I said I'd be happy to see another Polaris-type release; the 7600 was definitely not that, so no sense in even bringing it up.

You saying that, doesn't make it so. Both AMD and NVidia had smaller gains across the board this generation, and they still both have corresponding x060 and x600 competing with each other.

 

[SteinFG]

Current rumor from RGT on Youtube and "All the watts" on twitter is that only Navi43 and Navi44 remain. 41 and 42 are canned. Bruh

 

[adroc_thurston]

Current rumor from RGT on Youtube and "All the watts" on twitter is that only Navi43 and Navi44 remain. 41 and 42 are canned. Bruh

Not even the right numbers mentioned but yes, only two midrange parts remain.
Everything else dead.

 

[Saylick]

Current rumor from RGT on Youtube and "All the watts" on twitter is that only Navi43 and Navi44 remain. 41 and 42 are canned. Bruh

Is RGT just regurgitating what he's read on Twitter and "All the watts", or is he independently corroborating?

I wouldn't be surprised if it's the latter, since that tends to be the case for most of his videos: just regurgitating the latest Twitter rumors.

Not even the right numbers mentioned but yes, only two midrange parts remain.
Everything else dead.

RIP to the desktop GPU market for the next 2 years.

 

[MrPickins]

You saying that, doesn't make it so. Both AMD and NVidia had smaller gains across the board this generation, and they still both have corresponding x060 and x600 competing with each other.

It's not just me saying it. It's the major consensus in the 4060 reviews thread, and the majority of the reviews/commentary I've seen on the 4060 and 7600 releases. Hell, they barely beat last gen's 6 series cards (in VRAM heavy games, the 4060 8GB can even be worse than the 3060 12GB...).

These are 5 series cards being passed off as 6 series. You may be fine with the marketing shenanigans from both vendors, but I'm not.

Regardless, you're not even addressing my original point. I'd be happy with a Polaris-type release (very competitive in the midrange, great bang for the buck, and enough VRAM to make it somewhat futureproof). The 7600 is clearly not that, so bringing it up is a moot point.

 

[adroc_thurston]

RIP to the desktop GPU market for the next 2 years.

Wdym next 2 years? Forever.
AMD will build an overkill part in client dGFX sooner or later and you'll get a "meet the new boss, same as the old one".

 

[Saylick]

Wdym next 2 years? Forever.
AMD will build an overkill part in client dGFX sooner or later and you'll get a "meet the new boss, same as the old one".

Yeah, I think that's where you jumped the shark. What gives you any kind of confidence that AMD will take the crown with RDNA5?

 

[adroc_thurston]

What gives you any kind of confidence that AMD will take the crown with RDNA5?

N40 was a MI300-lite style of part aka something NV physically can't touch.
Was killed due to TTM reasons so N50 or something will take its place.

 

[SolidQ]

only two midrange parts remain.

hope at least one with 7900XT perfomance, but with RDNA4 RT

 

[adroc_thurston]

hope at least one with 7900XT perfomance, but with RDNA4 RT

Don't think so.
Those should be cheapo $399 or so midrangers.

 

[SolidQ]

Don't think so.
Those should be cheapo $399 or so midrangers.

So then max 6800XT perfomance

 

[gdansk]

Wow, they decided to compete in the mid-range where Battle Mage may end up. Don't think that follows a margin increasing strategy.

 

[adroc_thurston]

Wow, they decided to compete in the mid-range where Battle Mage may end up

Intel stuff is stillborn.

Don't think that follows a margin increasing strategy.

Oh it really does if your die is tiny.

 

[Joe NYC]

It is.
All the exotica remains contained to HPC parts like H100 or MI300.

It is going to trickle down, it is just a question of when and in what form.

1/4 of Mi300 is a good basis for future client products, both CPU and GPU.

Intel is already biting a bullet, adding the cost of an interposer to all client products. So, Intel is going to have to increase pricing of MTL parts. Which gives AMD some room to maneuver.

But also some pressure, if Intel later adds cache to the active interposer die.

It could very well be that both Zen 6 and RDNA 5 will be Mi300 derivatives.