Discussion RDNA4 + CDNA3 Architectures Thread

Page 139 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,684
6,227
136





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
Or Phoronix

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

 
Last edited:

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
What was the + going to add? LPDDR6 support?
RAM IO is largely disaggregated from core µArchs now since we had Zen2 and IODs - the monolithic APUs are probably doing the same thing on the silicon, just with shorter IF links.

Zen3+ in Rembrandt did have some tweaks to the µArch for mobile/power conservation beyond the new memory support for the SoC.

Edit: Zen3+ was also on partial EUV N6 process node vs the DUV based N7 for Zen3.

So possibly Zen5+ might have been planned for a better process node too.
 
Last edited:

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
The whole thing about RDNA4 hi end being axed was revealed far too early for my liking, especially given Lisa Su's propensity to troll everyone.
I mean you have to make this decision early before you start the really expensive work. besides that I'm not sure a CEO can make such a statement hat is a complete lie and get away with it as whatever they say has an impact on stock price.
 

Tuna-Fish

Golden Member
Mar 4, 2011
1,419
1,749
136
It's funny how it was once thought that GPUs don't need to clock higher. Just increase the number of shader units for more parallel processing power. I suppose at some point, a limit was reached where increasing the number of shader units became too unwieldy and hard to manage so they started focusing on increasing the clocks?

It's because workloads got more complex. When you are doing dead simple stuff with minimal internal state, just eating interpolated arguments and outputting to the framebuffer, there is little to no difference between "execute wider" and "execute faster". But when you do all the crazy complex stuff games are doing nowadays, where every shader needs to store so much state, "execute wider" means you have to put that internal state somewhere for as long as the shader runs, for every instance of the shader running. This means that residency goes down, and with it, ability to hide memory latency and efficiency.

GCN was a very good implementation of an "execute wider" arch, and Nvidia really pulled ahead from it with Maxwell when they started minimizing execution latency. IIRC SMM had a third less of on-paper throughput compared to SMX, but because it also drastically reduced instruction latencies, residency and with it the actually achieved throughput went up. Then they subsequently started increasing the clocks.
 
Last edited:

marees

Member
Apr 28, 2024
101
66
61
CEO can make such a statement hat is a complete lie and get away with it as whatever they say has an impact on stock price.
As per prev published roadmap RDNA 4 is scheduled for H2 2024

But why are board partners (powercolor has stated this as a fait accompli) saying that stock on shelves will happen only in 2025 🤔
 

leoneazzurro

Golden Member
Jul 26, 2016
1,005
1,597
136
As per prev published roadmap RDNA 4 is scheduled for H2 2024

But why are board partners (powercolor has stated this as a fait accompli) saying that stock on shelves will happen only in 2025 🤔
Actual plausible explanations:
- Launch at the end of 2024 but availability in early 2025
- Powercolor is referring to the custom versions (which come out usually later)
- Corollary of the above: AMD may only launch a FE edition of the card in H2 2024
 
Reactions: Mopetar

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
I mean you have to make this decision early before you start the really expensive work. besides that I'm not sure a CEO can make such a statement hat is a complete lie and get away with it as whatever they say has an impact on stock price.
That's just the point thought, where is the statement?

We got this whole thing through the rumor mill.

Considering how famously tight lipped AMD is before releases (post Koduri anyways) that seems sus.

We didn't know about Navi/RDNA1 being just a single chip until very soon before.
 
Reactions: Tlh97

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
As per prev published roadmap RDNA 4 is scheduled for H2 2024

But why are board partners (powercolor has stated this as a fait accompli) saying that stock on shelves will happen only in 2025 🤔
Well delays can happen. But telling we won't do product x and then actual do it is something entirely different.
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
Seems like somebody got very tired of waiting and decided to do something about it
I'm not seeing the positive side of this that emoji seems to communicate 🤔

Such data breaches have the potential to delay product releases further, to say nothing of putting thousands of jobs in jeopardy.
 

JustViewing

Member
Aug 17, 2022
159
268
96
Seriously, I think all big companies should limit the network speed for external access regardless who is trying to access. With reduced speed it will be very difficult to copy gigabytes of data without somebody noticing it. For people want to view files quickly, it should be done through remote desktop, by this IT can track who is accessing the files.
If the these files are stored in Public/ Private 3rd party Cloud, then they have only themselves to blame.
 
Reactions: TESKATLIPOKA

Aapje

Golden Member
Mar 21, 2022
1,467
2,031
106
Such data breaches have the potential to delay product releases further, to say nothing of putting thousands of jobs in jeopardy.

It can just as easily cause releases to be sped up, to get ahead of the leaks. And it's silly to argue that thousands of jobs would be lost. You just seem to dislike such leaks and are throwing arguments out there, even if they don't make much sense.
 

inquiss

Member
Oct 13, 2010
89
160
106
It can just as easily cause releases to be sped up, to get ahead of the leaks. And it's silly to argue that thousands of jobs would be lost. You just seem to dislike such leaks and are throwing arguments out there, even if they don't make much sense.
Of course any normal person would dislike those leaks. Those are company secrets not accessible by the public for a reason. You seem taking umbrage at a poster that is taking a totally normal view. Weird.
 

soresu

Platinum Member
Dec 19, 2014
2,941
2,164
136
You just seem to dislike such leaks
There is good reason to.

They aren't good for business fullstop.

It's one thing to discuss speculation and rumors about release dates and specs in all fun for the pre-release hype trains.



It's another thing entirely when a criminal enterprise has caused a massive data breach, compromising potentially years worth of trade secrets, tens of thousands of employees private data, and who knows what else.

It's not like AMD have an unassailable position and mountains of money to burn here.

I wouldn't wish this on either Intel or nVidia, and I haven't bought their products in 20 and 15 years respectively.
 
Last edited:

adroc_thurston

Diamond Member
Jul 2, 2023
3,311
4,774
96
My long shot on why these were scrapped is after they had to take out IF from them to put an ever larger NPU inside, they decided there was no point in bringing higher-compute RDNA4 iGPUs to SoCs before LPDDR6 becomes widely available.
No, they just had to re-focus their roadmap for execution reasons.
Radeon and Instinct went thru the same thing more or less.
 

Hitman928

Diamond Member
Apr 15, 2012
5,593
8,760
136
Seems the hack isn’t as bad as initially reported. Breach was actually at a 3rd party who apparently did assembly work for AMD. If so, nothing really valuable in terms of IP or even future products would have been taken.

 

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,228
1,665
136
Seems the hack isn’t as bad as initially reported. Breach was actually at a 3rd party who apparently did assembly work for AMD. If so, nothing really valuable in terms of IP or even future products would have been taken.

Because of this, I'm highly skeptical of the legitimacy or accuracy of the "leaked roadmap", but here it is - freshly circulating today.

 
Reactions: marees
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |