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

 

[reaperrr3]

if price is real

Not ruling out it might be real for the US, but MSRPs will certainly land a bit below Nvidia's.
Although the 9070s have considerably improved AMD's brand perception, at about the same price, many people would likely still choose 5060Ti over 9060XT, especially at 8GB vs. 8GB due to AMD's worse memory management, and AMD knows this.

These prices are either including tariffs + early adopter mark-up + "OC model tax", or these particular shop listings are simply not representative.
Remember how some early leaked shop listings for 9070s were suggesting something like 700-800, yet at least the official MSRPs ended up way below that.
Pretty sure the official MSRPs will be at least 30-50$ below the 5060Ti's, so I'd expect real market prices to settle in that area relative to 5060Ti as well, once factors like tariffs are included for both.

My guess would be around $379 MSRP for the 9060XT-16GB and $299 for the 8GB (which would pitch the 8GB against the 5060 rather than 5060Ti-8G, where AMD's worse memory management may not hurt them as badly).

...And then I hope AMD/AIBs won't make many 8GB cards, phase them out soonish and replace it with a 9060 with 28 CUs, 12GB@18Gbps@96bit, 130-150W at $299, because such a card would likely outsell all other cards below the 5060Ti-16GB and 96XT-16GB.

 

[jpiniero]

...And then I hope AMD/AIBs won't make many 8GB cards, phase them out soonish

That I could see. The 12 GB thing isn't happening though.

 

[Rigg]

Also GPU market is a disaster right now. As mentioned earlier/elsewhere stuff like 7600 is still going for $300.

Why price down in a market where people will spend that kind of money?

Tarrifs okay a role, but so does AI compute and Nvidia's complete lack of interest in the consumer market as well.

The GPU market is a dumpster fire. 9070XT cards are selling at Micro Center for like 40-50% over MSRP right now. MSRP is just a meaningless number on a marketing slide at this point. The smart play for them is to just set 9060 (XT) 8 GB MSRP low and hope for good reviews based on value in this crappy market. They can then sell whatever stock they have (including a few token MSRP models) quickly at launch, and quietly stop making the 8 GB cards all together. Nobody wants 8 GB cards at this performance tier anyway, and prominent reviewers are going to 💩 on them worse if the MSRP is too high. AMD and their partners will make more money on higher demand 16GB versions in the long run so they shouldn't tarnish them with 8 GB models that have the MSRP set too high.