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

 

[Josh128]

Navi4C was cancelled in late January 2023.

Which logically would have taken place once they determined from developing RDNA3 that chiplet GPUs still have too many challenges. Cancelling chiplet 4C was likely a great call because of that. Chiplet= no bueno vs mono.

 

[Kepler_L2]

Which logically would have taken place once they determined from developing RDNA3 that chiplet GPUs still have too many challenges. Cancelling chiplet 4C was likely a great call because of that. Chiplet= no bueno vs mono.

There is no relationship at all between Navi4C and Navi31/32.

 

[gaav87]

Anyway cant post screenshots for some reason.
Anyone with 9070xt if you want to see how broken the overclocking is
go to heaven benchmark near the baloon (free camera)
Set -25mV or -30mv or -35mV and +275mhz clock offset (leave the power and memory stock)
U will boost to 3550mhz then try -40mV or 0mV BAM 3300mhz
I crash at 3650mhz at -30mV or -35mV any other mV settings im at either 3250 or 3300mhz... Broken as hell...

 

[gaav87]

- Same thing in helldivers 2...
I can fiddle with the offset and mV and somehow trick "force" the card to run at 3550mhz stable for few min...
Sometimes the card is stuck at 3150mhz and 320W even with +10% power limit... should be up to 410W (highest i saw in furmark)
hotspot at 75C (forced fans at 80%)... .
- I force reset drivers and same spot on the ship in helldivers bam 374W 3550mhz and crash after few min xD
FPS with stuck boost clocks 144 at 3150mhz and 157-158 with ~3550mhz "unstuck"

- steel-nomad 3dmark seem to be broken... You can have 2950mhz at 290W or 378W+ with exact same perf. Its like the clock is stuck at 2950mhz and 890mV !!!
Like what the heck is this. Sometimes im able to "unstuck" it and gain like 5% perf its boosting to 3150mhz netting me low top100 leaderboard...

-Same thing i did in furmark when i somehow unstuck the clock and did 5% faster than rtx 4090D performance
furmark: Stock 343fps, normal overclock 373fps, unstuck oc 400fps
This is madness hope they fix this...

- I have also noticed that the voltage is also stuck... Card is doing 1050mV at 3150mhz and u can go 270W or 330W and its still stuck. So its not powerlimited. Something is sreaming broken... Why doesnt it always go 1.175V and 3550mhz ? (talking about helldivers) its driving me nuts... Its like the driver is forcing the card to undervolt itself.
When the clock and voltage is stuck at for example 1050mV at 3150mhz i can go -100mV w/o crash and the voltage does not budge... but if i go -105/-110mV the clocks and the voltage decrease, a little with crash in next few sec if i dont increase the voltage again...

 

[Josh128]

There is no relationship at all between Navi4C and Navi31/32.

How so?