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

 

[gaav87]

Maybe they waiting RTX 5080/5090 reviews

I think its coz of this.

NVIDIA RTX Neural Rendering Introduces Next Era of AI-Powered Graphics Innovation | NVIDIA Technical Blog

NVIDIA today unveiled next-generation hardware for gamers, creators, and developers—the GeForce RTX 50 Series desktop and laptop GPUs. Alongside these GPUs, NVIDIA introduced NVIDIA RTX Kit…

developer.nvidia.com

 

[soresu]

I think its coz of this.

NVIDIA RTX Neural Rendering Introduces Next Era of AI-Powered Graphics Innovation | NVIDIA Technical Blog

NVIDIA today unveiled next-generation hardware for gamers, creators, and developers—the GeForce RTX 50 Series desktop and laptop GPUs. Alongside these GPUs, NVIDIA introduced NVIDIA RTX Kit…

developer.nvidia.com

Nothing in there is new for anyone paying attention to the research/academic side of gfx.

AMD may not know exactly what is going from the research level to implementation push with nVidia, but they can infer it from language, and the Neural Radiance Cache paper made it pretty obvious that their intent was to put it to use soon.

 

[adroc_thurston]

I think its coz of this.

NVIDIA RTX Neural Rendering Introduces Next Era of AI-Powered Graphics Innovation | NVIDIA Technical Blog

NVIDIA today unveiled next-generation hardware for gamers, creators, and developers—the GeForce RTX 50 Series desktop and laptop GPUs. Alongside these GPUs, NVIDIA introduced NVIDIA RTX Kit…

developer.nvidia.com

Irrelevant

 

[GTracing]

Crockpot theory time. What if the reason for the delayed launch is AMD is planning to severely undercut Nvidia.

If AMD announces prices before Blackwell launches, Nvidia will lower their prices. But if AMD announces prices after Blackwell launches, Nvidia can't lower prices without bigger consequences. Early adopters would not be happy, quite a few will return them just to buy again at the new lower price.

Probably not the case, but one can hope.

 

[adroc_thurston]

What if the reason for the delayed launch is AMD is planning to severely undercut Nvidia.

It's the opposite.

 

[GTracing]

It's the opposite.

Just for curiosity's sake, how would you rate your confidence in that compared to, say, your confidence that Zen5 is a 40% ipc uplift?

 

[adroc_thurston]

how would you rate your confidence in that compared to, say, your confidence that Zen5 is a 40% ipc uplift?

Pretty high, Turin was a 40-50% perf bump indeed.

 

[coercitiv]

But if AMD announces prices after Blackwell launches, Nvidia can't lower prices without bigger consequences.

We're talking about a company that unlaunched a card to change it's name and MSRP. What bigger consequences?!

 

[adroc_thurston]

We're talking about a company that unlaunched a card to change it's name and MSRP. What bigger consequences?!

yeah, GPU wars are normal.

 

[GTracing]

We're talking about a company that unlaunched a card to change it's name and MSRP. What bigger consequences?!

Changing the price after people have actually bought it is far more disruptive.

 

[adroc_thurston]

Changing the price after people have actually bought it is far more disruptive.

Hell no rapid price cuts are the GPU war stable. But NV prefers vomiting new SKUs instead.

 

[CastleBravo]

Crockpot theory time. What if the reason for the delayed launch is AMD is planning to severely undercut Nvidia.

If AMD announces prices before Blackwell launches, Nvidia will lower their prices. But if AMD announces prices after Blackwell launches, Nvidia can't lower prices without bigger consequences. Early adopters would not be happy, quite a few will return them just to buy again at the new lower price.

Probably not the case, but one can hope.

AMD would have to wait for the 5070 and 5070 Ti to launch as those are the only announced Nvidia GPUs AMD is trying to compete with. AMD could give the 9070 XT away for free and it wouldn't affect 5090 pricing.

 

[Keller_TT]

If the performance is indeed 4080 in raster and 4070 Ti Super in RT, and the price is 600$ or less, I'm buying immediately. That's solid value IMO. That's plenty of card for 1440p w/o upscaling or 4K with upscaling.

The MBA XT can be at max stretch $549 if it's well within 10% from 4080S for 260W.. Else they loose their play. The mindshare and market share to overcome is so enormous. The decision is about whether they have to come in at $479 or $499 or more.

In the real world, kids & dads & moms won't even bat an eyelid if AMD is priced above Gold Standard™ Nvidia with "amazing 50 series features" for a parallel class card.

AMD would have to wait for the 5070 and 5070 Ti to launch as those are the only announced Nvidia GPUs AMD is trying to compete with. AMD could give the 9070 XT away for free and it wouldn't affect 5090 pricing.

They don't have to. All they need is to feed an Excel sheet (or an ML model) and their engineers can work out all comparative details to a <3% margin of error based on 5080/90 and 40 series.
They would already have an idea and can get down to SM level, along with PPA, and finer details.

 

[adroc_thurston]

The MBA XT can be at max stretch $549 if it's well within 10% from 4080S for 260W

It's just 7900XTX (with more stuff).

 

[Keller_TT]

It's just 7900XTX (with more stuff).

Looking at the spread from MBA to AIB tiers, I presume they'll be power limited for the XT, and priced accordingly.
The reference will be the baseline and not far from 4080S, and the AIBs could push it to XTX level for 320-330W.
The 9070 is even more power limited looking at the clocks. So, a 4070Ti for 220-230W is what I think.

 

[adroc_thurston]

Looking at the spread from MBA to AIB tiers, I presume they'll be power limited for the XT, and priced accordingly.

What

The reference will be the baseline and not far from 4080S, and the AIBs could push it to XTX level for 320-330W.

The reference is 7900XTX perf. That's it.

The 9070 is even more power limited looking at the clocks. So, a 4070Ti for 220-230W is what I think.

I've no idea what a 4070ti is but 9070 non-XT is 7900XT perf.

 

[Keller_TT]

What

The reference is 7900XTX perf. That's it.

I've no idea what a 4070ti is but 9070 non-XT is 7900XT perf.

At the moment, it all sounds too good to be true because the specs and base power for the XT is more or less known.
If it is indeed the case, this might be the best hardware AMD has ever launched. But for it to be a Ryzen moment, they have to come in and wow the market like they did with 8 cores and 92% single core performance of 7700K at the same price along with a meaty cooler packed in.

 

[adroc_thurston]

At the moment, it all sounds too good to be true because the specs and base power

no it's normal.
more like n31 is ass.

If it is indeed the case, this might be the best hardware AMD has ever launched

hell no, ATi did bigger, better and meaner things before.

But for it to be a Ryzen moment

they have to ship Navi40. Or any chungus Navi5c.

they have to come in and wow the market like they did with 8 cores and 92% single core performance of 7700K at the same price along with a meaty cooler packed in.

Ryzen wasn't a thing until 3950X, and that had no cooler.
Frankly Ryzen wasn't a thing until 5950X won it all.

 

[techjunkie123]

The MBA XT can be at max stretch $549 if it's well within 10% from 4080S for 260W.. Else they loose their play. The mindshare and market share to overcome is so enormous. The decision is about whether they have to come in at $479 or $499 or more.

Bruh we're talking raster performance that previously cost 900$+ and RT performance that cost at least 750$+. Now being sold for 600$ (??), using a similar die size. How is that unreasonable??

I don't get the hype that expects AMD to price the 9070XT at 500$ or less. (If performance is indeed 4080 raster and 4070 Ti Super in RT)

 

[Mopetar]

Although JHH would never do it, I'd wet myself laughing if NVidia also delayed to force AMD to launch their mid-tier first or delay even further.

 

[PJVol]

The reference is 7900XTX perf. That's it.

And the perf. with RT is +10% at best?

 

[adroc_thurston]

And the perf. with RT is +10% at best?

What does "+10% at best" mean.

 

[Novacius]

Bruh we're talking raster performance that previously cost 900$+ and RT performance that cost at least 750$+. Now being sold for 600$ (??), using a similar die size. How is that unreasonable??

I don't get the hype that expects AMD to price the 9070XT at 500$ or less. (If performance is indeed 4080 raster and 4070 Ti Super in RT)

Stop comparing it to the previous generation and look at the 5070 instead. It's very possible the 9070 XT is 20-30% faster in raster and just about as fast in RT. Nvidia on the other hand has a feature advantage, especialy with DLSS. With DLSS 4 it will be possible to override the DLSS version via driver in every game that has at least DLSS 2.0, so a lot of games from the past five years, which is massive. Also Nvidia has mindshare and is often viewed as the default choice. To overcome that and get more marketshare, AMD can't price the 9070 XT even the same as the 5070. $50 difference is the minimum and gets AMD below the $500 barrier, which is psychologically important I think, even if the difference isn't that big.

 

[adroc_thurston]

the other hand has a feature advantage, especialy with DLSS. With DLSS 4 it will be possible to override the DLSS version via driver in every game that has at least DLSS 2.0

You can do the same with FSR4. Snooze.

 

[Novacius]

You can do the same with FSR4. Snooze.

How many games are there with FSR 3.1 again versus games with DLSS 2.0 at least?