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

 

[coercitiv]

Digital Foundry presents a sneak peak at FSR4:

Yes, the DF reference is a joke.

 

[insertcarehere]

Now this is very interesting... N48 clearly boosts high for one, benchmarks look better than expected...

 

[Ranulf]

Digital Foundry presents a sneak peak at FSR4:

Yes, the DF reference is a joke.

Heh, you almost got me there. I had to do a double take.

 

[gaav87]

There new info about CoD bench.
Some guy testing own Bench with VRS on/off on XTX
with on - 108 fps
with off - 80 fps
if it's was 9070 non XT at 99fps still very good result, sitting between 7900XT and XTX

NVdone.

 

[adroc_thurston]

NVdone.

View attachment 114545

View attachment 114544

How close is this to 7900XTX?

 

[gaav87]

Looks good.

 

[gaav87]

How close is this to 7900XTX?

5% below and 1:1 in speedway

 

[Win2012R2]

So boost isn't much higher than 3 Ghz...

 

[adroc_thurston]

So boost isn't much higher than 3 Ghz...

Perf@CU is still up like 50% over RDNA3.

 

[Win2012R2]

Perf@CU is still up like 50% over RDNA3.

It's awesome (if true), still sad relatively small chip can't get to 3.5-3.6 ghz...

 

[Saylick]

Board power draw was 330W.

 

[SolidQ]

FSR4

Looks good.

Only question how it's in gaming
RT also is big question how it's work in heavy/PT mode

5% below and 1:1 in speedway

XTX also 1:1 in speed comapred to 4070tiS, but we know in heavy/PT is massive lose

 

[Saylick]

Hmm, maybe there’s more performance to squeeze out of the chip?

https://twitter.com/x/status/1877265204330819699

 

[SolidQ]

Hmm, maybe there’s more performance to squeeze out of the chip?

There was show pre alpha driver

NVdone.

Where did you get 14k on 4080?

Here is 13.7 on 4080S and 85% better result than other

I scored 13 735 in Time Spy Extreme

AMD Ryzen 9 7950X3Dâ, NVIDIA GeForce RTX 4080 SUPER x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

 

[gaav87]

There was show pre alpha driver


Where did you get 14k on 4080?

Here is 13.7 on 4080S and 85% better result than other

I scored 13 735 in Time Spy Extreme

AMD Ryzen 9 7950X3Dâ, NVIDIA GeForce RTX 4080 SUPER x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

3dmark avg score

 

[SolidQ]

score

if it's avg do you know, that like between EXTREME OC and normal. You can't compare like this
You need normal comparision
it's like your salary 1 mil $, mine 3k - avg person is 501k? Lol

 

[insertcarehere]

if it's avg do you know, that like between EXTREME OC and normal. You can't compare like this
You need normal comparision

Well the GPU that's been tested is also a non-ref model from Asus with a 330w TBP, it is almost certainly expected to be on a higher range than 'normal'.

 

[adroc_thurston]

3dmark avg score

That's a suckass metric.
I suggest picking any random 4080S review instead.

 

[SolidQ]

Well the GPU that's been tested is also a non-ref model from Asus with a 330w TBP

Still there alot will be non ref, as default speed

I suggest picking any random 4080S review instead.

I already give him link to normal 4080S result. There is 13.7k

 

[adroc_thurston]

It's all preliminary numbers.
Wait a bit and have fun meanwhile.

 

[gaav87]

This is not final performance... True dual issue ?

 

[eek2121]

A new driver is going to bump those numbers up.

 

[gaav87]

Still there alot will be non ref, as default speed


I already give him link to normal 4080S result. There is 13.7k

There is still room man its going too the fukin moon 4090 ?

 

[adroc_thurston]

True dual issue ?

Caches.

There is still room man its going too the fukin moon 4090 ?

No but it's fast nuff.

 

[poke01]

Now this is very interesting... N48 clearly boosts high for one, benchmarks look better than expected...
View attachment 114539
View attachment 114540
View attachment 114541

Seriously find me the person that stopped RDNA4 halo. This is looking like AMD cooked. Very optimistic now