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

 

[Heartbreaker]

Even if we had a perfect snapshot of DIY sales in all major markets around the world, that would not be enough to extrapolate into market share which is primarily driven by OEM sales. The same applies in reverse.

Not an exact number, but a clear direction could be discerned, or at least enough to know that it isn't reversed like it is at Mindfactory.

Also I expect OEM choice to a large degree is driven by their experience offering consumers a choice, and seeing what they chose. It's not like they care about anything more than the bottom line. So they are going to offer what sells best.

 

[Keller_TT]

You didn't even catch the point I was trying to make, not to mention the Strawman part.

The Amazon screenshot was about challenging our preconceptions regarding pricing and SKU of choice among consumers. It said little if anything about AMD market share. Going back to @Jan Olšan 's post, he pointed out the discrepancies between prebuilt and DIY market, in the sense that one should not use data from one of them to extrapolate the results in the other.

Even if we had a perfect snapshot of DIY sales in all major markets around the world, that would not be enough to extrapolate into market share which is primarily driven by OEM sales. The same applies in reverse.

Ok. My impression was that you were providing a contra to the other post that NV still is King in retail too, i.e., DIY card purchases, and that the data from one German vendor is irrelevant.

Regarding the DIY vs prebuilt market, I don't see it like the OP mentioned. NV owns desktop. It used to be Intel+Nvidia for ages. Now desktop CPU has tilted more towards AMD and AMD+Nvidia is just the norm, and also the one that brings margins to OEMs. AMD had poor OEM rollout even for RDNA2 when it had the advantage. It should have been much more, but the RTX brand sells!

 

[Jan Olšan]

You can't assume that one shop in Germany has any correlation with, say, another shop in China. I can check top ten most sold GPUs in one local shop: 10 GeForces, one Radeon (same thing in another shop). It's pretty obvious that at least here in Finland it's very, very unlikely that AMD is selling better or even 50/50 split. Totally different story when it comes to CPUs - only AMD in top ten.

Correlation? There absolutely should be *some sort* of it. Direct correspondence probably not since markets differ.

The problem with bestseller charts of individual items is that they are like a majority election system. A company that has a 55 % market share when all unit sales are counted can take 100% of top-20 or even top-50 best selling cards chart's spots under the right circumstances. Similar reason behind the popular "Radeon cards don't even appear in Steam's survey" idea (since there is that cutoff that influences what appears in the stats, and cars with lower overall share are much more likely to stay under it).

 

[dacostafilipe]

I don't think it's faire to compare both.

One is specialised pc parts shop, mostly used to people that have a higher understanding of computers and the other has everything and is more accessible by people that are less tech-savvy. Who do you think will make the best compromise about buying a GPU and who will just get the Nvidia?

I got 4 Ryzen CPUs and 2 Radeon GPUs last year from Mindfactory, so I'm partly to blame ...

 

[Thunder 57]

To offer even more context to the post above, a screenshot from Amazon:

View attachment 115104

So strong is the cult of Nvidia that even today 3060's at near $300 are the best sellers.

 

[MarutTheSlayer]

So strong is the cult of Nvidia that even today 3060's at near $300 are the best sellers.

It has been the only RTX card in the 300 bucks range with 12GB for a long time. Could be the peer influence too, with many entry level card buyers and online gamers using it and hence a safe buy. It's still a very decent card for 1080P.
Those buying want the 3060 12GB. That's it. No comparisons needed.

 

[GodisanAtheist]

Good lord AMD needs to put this farce of a launch out of its misery.

Physical cards are at physical retailers and we still have essentially *nothing*.

 

[Saylick]

Good lord AMD needs to put this farce of a launch out of its misery.

Physical cards are at physical retailers and we still have essentially *nothing*.

I'm currently surprised a whole slew of gaming benchmark leaks from AIBs haven't come out yet, premature BIOS or not. What we seen so far is only Timespy, Port Royal, and COD BO6?

 

[BurnItDwn]

Could they be intentionally waiting for return windows to close for all the folks who bought up the 7900xt and xtx cards in December? Or is that foolish ramblings of a lunatic?

 

[Heartbreaker]

So strong is the cult of Nvidia that even today 3060's at near $300 are the best sellers.

The Amazon UK Best Sellers has the NVidia GT 710, as the #6 best seller.

Granted, it's super cheap, but exactly how old would your IGPU have to be, to actually be worse than that?

 

[RnR_au]

Could they be intentionally waiting for return windows to close for all the folks who bought up the 7900xt and xtx cards in December? Or is that foolish ramblings of a lunatic?

Thats handled by retailers. Not AMD.

 

[GTracing]

Thats handled by retailers. Not AMD.

It's still bad for AMD if their GPUs get returned en masse.

 

[jpiniero]

Could they be intentionally waiting for return windows to close for all the folks who bought up the 7900xt and xtx cards in December? Or is that foolish ramblings of a lunatic?

Still say they are reconsidering the price, and need a better idea on how they compare to the 5070 and Ti.

Having two models so close in performance isn't helping things.

Edit: The Wildcard might be if DLSS4Q got a performance boost. I see some reviewers like to use the Quality Mode instead of pure raster.

 

[MarutTheSlayer]

I'm currently surprised a whole slew of gaming benchmark leaks from AIBs haven't come out yet, premature BIOS or not. What we seen so far is only Timespy, Port Royal, and COD BO6?

You can find it if you look for it on YT in the last couple of days. Could easily have been fudged though but interesting buildup.

 

[MarutTheSlayer]

Still say they are reconsidering the price, and need a better idea on how they compare to the 5070 and Ti.

Having two models so close in performance isn't helping things.

Edit: The Wildcard might be if DLSS4Q got a performance boost. I see some reviewers like to use the Quality Mode instead of pure raster.

It's probably bcos of FSR4 not being ready and reviewers giving it a rough time when DLSS4 is ready to roll, and brings substantial image quality improvements to 40 series as well.

 

[Thunder 57]

It's probably bcos of FSR4 not being ready and reviewers giving it a rough time when DLSS4 is ready to roll, and brings substantial image quality improvements to 40 series as well.

If DLSS was already "better than native" what does that make DLSS4? "Double better than native"?

 

[GodisanAtheist]

If DLSS was already "better than native" what does that make DLSS4? "Double better than native"?

-Obviously 4x as good as already-better-than-native. Duh.

 

[MarutTheSlayer]

If DLSS was already "better than native" what does that make DLSS4? "Double better than native"?

Nothing is ever perfect. Ray Reconstruction noise elimination, ironing out ghosting/smearing, and being able to resolve more details with their more refined ML models, all add up.
All this, while being more performant even for RTX 40 series, that's another step.

 

[MarutTheSlayer]

-Obviously 4x as good as already-better-than-native. Duh.

Was this joke made using FSR?

 

[Thunder 57]

Nothing is ever perfect. Ray Reconstruction noise elimination, ironing out ghosting/smearing, and being able to resolve more details with their more refined ML models, all add up.
All this, while being more performant even for RTX 40 series, that's another step.

It has been my understanding that DLSS with the new Transformer model improves image quality at the cost of performance.

 

[Thunder 57]

Was this joke made using FSR?

I have never once read that FSR was "better than native". I've read that it is crap compared to DLSS though.

 

[MarutTheSlayer]

It has been my understanding that DLSS with the new Transformer model improves image quality at the cost of performance.

Idk, but DLSS has improved with every step. I think it'll be fine. Lets see how it is.

 

[Mopetar]

Well hopefully if nothing else this means that we'll have a GPU launch with good availability so that the people who want one can easily get a card without having be awake and available at a specific time or to run a bot and that the scalpers can't buy them all up to flip on eBay.

It is unusual for the cards to have been shipped already without a launch date for them. It's even more unusual that none of them have been sold either because it doesn't matter what price AMD settled on for the MSRP when someone will pay $1,000 just to have one now and earlier than anyone else. A tech review site could get around embargo dates that way while being able to write it off as a business expense.

Soon (TM)

 

[MarutTheSlayer]

I have never once read that FSR was "better than native". I've read that it is crap compared to DLSS though.

DLSS Q is remarkably better than native 4K with TAA. Issues arise when FG is used, but without that it's fine mostly.
It's also great to reduce power consumption and run cool. Something Sony is now trying to perfect with PSSR.

 

[Thunder 57]

DLSS Q is remarkably better than native 4K with TAA. Issues arise when FG is used, but without that it's fine mostly.
It's also great to reduce power consumption and run cool. Something Sony is now trying to perfect with PSSR.

That's because TAA is crap, not because DLSS makes it magically better than native.