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

 

[Saylick]

Temps look good thou at 3.25g >330w
I imagine h++ will do 3.4 with golden 3.5

View attachment 118028

What... no way it's only sitting at 57C with that much juice. The benchmark must not be stressing the GPU all that much. I'll have to see it in a third-party review to believe it.

 

[adroc_thurston]

What... no way it's only sitting at 57C with that much juice

330W isn't really all that much for modern 3 slot chunguses.

 

[MrTeal]

What... no way it's only sitting at 57C with that much juice. The benchmark must not be stressing the GPU all that much. I'll have to see it in a third-party review to believe it.

From the 5070 Ti TPU reviews

Doesn't seem crazy out of line for a good air cooler.

 

[gaav87]

yeah, maybe AMD surprpise us with GDDR7 version

Heard rumors about this today from chinese but not for 9070 not on chiphell other forum. But idk if true

What... no way it's only sitting at 57C with that much juice. The benchmark must not be stressing the GPU all that much. I'll have to see it in a third-party review to believe it.

Its heaven benchmark on linux mesa hack and slash drivers so take it as it is.

 

[SolidQ]

Heard rumors about this today from chinese but not for 9070 not on chiphell other forum. But idk if true

you mean this? About 9080?

 

[Saylick]

330W isn't really all that much for modern 3 slot chunguses.

Shows how out of touch I am with the modern state of GPUs. I am still expecting 75C+ temps from 7-8 years ago when coolers weren’t ridiculously oversized.

 

[adroc_thurston]

Shows how out of touch I am with the modern state of GPUs. I am still expecting 75C+ temps from 7-8 years ago when coolers weren’t ridiculously oversized.

It's kinda good tho. Lower temps mean less leakage.

 

[Bryo4321]

you mean this? About 9080?

I’m not going to count on it but gosh I’d love to see it. Seems like it could be a high margin product. If it could compete with a 5080 it seems like an opportunity worth taking, no?

 

[Mopetar]

If it does support GDDR 7 memory you'd think they'd at least let AIBs who were willing source their own memory to pair with a top-end OC model. AMD would do well by both consumers and their board partners to let that happen.

Then again maybe it turns out similar to Blackwell where the added memory bandwidth doesn't do as much as one would typically assume and no one wants to pay for more expensive memory that adds very little performance.

 

[linkgoron]

If it does support GDDR 7 memory you'd think they'd at least let AIBs who were willing source their own memory to pair with a top-end OC model. AMD would do well by both consumers and their board partners to let that happen.

Then again maybe it turns out similar to Blackwell where the added memory bandwidth doesn't do as much as one would typically assume and no one wants to pay for more expensive memory that adds very little performance.

is there ANY reason to believe that RDNA4 supports GDDR7? It's unlike AMD to waste space and engineering on something like GDDR7 support.

 

[adroc_thurston]

If it does support GDDR 7 memory you'd think they'd at least let AIBs who were willing source their own memory to pair with a top-end OC model

Supports != they validated or designed a PCB for it.

It's unlike AMD to waste space and engineering on something like GDDR7 support.

wdym space and engineering.
memory IP is licensed from EDA vendors.

 

[RnR_au]

is there ANY reason to believe that RDNA4 supports GDDR7? It's unlike AMD to waste space and engineering on something like GDDR7 support.

DDR7 makes sense on a 32GB Pro card.