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

 

[Krteq]

- Nah guy its going to have 42 GCDs and the answer to everything.

Think bigger, 42 GCDs, 42 TCDs on a single card but AMD have perfected xfire so they can sell you a 42 card stacked unit which can go into a 42 cabinet cluster. 42's all the way down.

9 of 10 Vogons will buy that for sure 😁

 

[Timorous]

9 of 10 Vogons will buy that for sure 😁

Did they file their PO in triplicate with all the requisite authorisations?

 

[tajoh111]

I am so bored. Where is the false hype of a 5,000 GHz 1,024 WGP 128GB 7090 killer released q1 next year for $400 so we can do silly stuff like make up product stacks based on those outrageous claims and then be sad when AMD don't meet those expectations?

I think people have wisened up at this point....or it could be that AMD admitting they are bowing out of the high end for now have killed their fantasies and they are just as delusional(just holding it back for the time being). .

It was really ridiculous last time around. I remember people believing the rx7600 which was confirmed to use 6nm even during the rumor phase, to be be faster than a 6900xt lol.

 

[linkgoron]

I think people have wisened up at this point....or it could be that AMD admitting they are bowing out of the high end for now have killed their fantasies and they are just as delusional(just holding it back for the time being). .

It was really ridiculous last time around. I remember people believing the rx7600 which was confirmed to use 6nm even during the rumor phase, to be be faster than a 6900xt lol.

I believe it's two things

- you'd have to assume that the so-called insiders that hyped Zen5 to death have been exposed as jokes, so they'll probably wait this one out.

- How much can you hype a card that'll probably match Nvidia's third or fourth tier card (probably 5070 or 5060ti level)

 

[adroc_thurston]

I think people have wisened up at this point....or it could be that AMD admitting they are bowing out of the high end for now have killed their fantasies and they are just as delusional(just holding it back for the time being). .

It was really ridiculous last time around. I remember people believing the rx7600 which was confirmed to use 6nm even during the rumor phase, to be be faster than a 6900xt lol.

Because GPGPU mines await.

 

[Mopetar]

I am so bored. Where is the false hype of a 5,000 GHz 1,024 WGP 128GB 7090 killer released q1 next year for $400 so we can do silly stuff like make up product stacks based on those outrageous claims and then be sad when AMD don't meet those expectations?

If you take a random Chinese tweet and "translate" it as that before posting it here or on some other web forum, MLID will have an article about it out before you can blink.

The real fun is seeing how much you can get away with while still getting MLID to publish it. See if you can get away with calling it Navi 41 3/4 or something utterly ridiculous.

 

[coercitiv]

See if you can get away with calling it Navi 41 3/4 or something utterly ridiculous.

Or talk about the major similarities between UDNA and Larrabee, as we are basically circling back to x86 core swarms.

 

[marees]

Equivalent of 4060 ti 8gb for $300 ??

https://twitter.com/x/status/1862815835691106535

 

[SolidQ]

Equivalent of 4060 ti 8gb for $300 ??

isn't there on screen win by running points?
300$ too much for 8gb, while B580 is ~250-260 and 12GB

 

[gaav87]

Equivalent of 4060 ti 8gb for $300 ??

https://twitter.com/x/status/1862815835691106535

Price will be 229-259$ (for top models)
320xmx (on the B580 is Asrock mistake) B570 has it's correct.
PCI-E 4.0 is correct (rumor PCI-E 5.0 is bugged will get fixed in bios update)

 

[maddie]

Price will be 229-259$ (for top models)
320xmx (on the B580 is Asrock mistake) B570 has it's correct.
PCI-E 4.0 is correct (rumor PCI-E 5.0 is bugged will get fixed in bios update)

Those specs. 320 vs 144 XMX engines???

 

[gaav87]

Those specs. 320 vs 144 XMX engines???

Like i said Asrock mistake real is 160xmx.
G31 is 256xmx 256bit 16gb 19.5gbps
G10 (bmg - x3) was 24xe2 and 28xe2 cores - according to rumors on twitter.
But i also heard about 320xmx(40core) 320bit 20gb
The EU count on this manifest are for old alchemist 1core=16eu according to bionic on twitter.

 

[jpiniero]

isn't there on screen win by running points?
300$ too much for 8gb, while B580 is ~250-260 and 12GB

Intel is irrelevant. What is relevant is AMD's existing products and supply is there but not that amazing of an amount.

 

[del42sa]

https://wccftech.com/amd-radeon-rx-...-performance-7900-xtx-on-par-rtx-4080-raster/

AMD Radeon RX 8800 XT “RDNA 4” GPU Allegedly Offers 45% Faster RT Performance Than 7900 XTX, On-Par With RTX 4080 In Raster & Lower Power Too​

 

[igor_kavinski]

^^^ If true, wow! 7900 XTX could go under $600!

 

[SolidQ]

45% Faster in RE4 with RT? Other games? he didn't say what games.
Found this

RTX 4090 in Re4?
Something off, unless it's in RE4 Raster also much faster than XTX

 

[maddie]

^^^ If true, wow! 7900 XTX could go under $600!

If true, who cares about 7900XTX.

 

[igor_kavinski]

If true, who cares about 7900XTX.

People who don't care about RT and who want a decent affordable card for raster or AI compute.

 

[maddie]

People who don't care about RT and who want a decent affordable card for raster or AI compute.

From article.

- 25% power
+ 45% RT ( Bioharzard 4RE )
= Raster ( 4080/4080S )

 

[SolidQ]

From article.

- 25% power
+ 45% RT ( Bioharzard 4RE )
= Raster ( 4080/4080S )

You miss this

 

[maddie]

You miss this

????????

 

[igor_kavinski]

????????

He's saying it's not just RE4 that was tested.

 

[SolidQ]

????????

and 4080 not only Raster

Also add RT Re4 from TPU
45%??

 

[ToTTenTranz]

+45% in RE4 is indeed weird. It would be more believable if it was 7900XTX +45% in Alan Wake 2 or Cyberpunk Overdrive.

 

[vanplayer]

wrong post