Discussion RDNA4 + CDNA3 Architectures Thread

Page 479 - 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
1,779
6,798
136
1655034287489.png
1655034259690.png

1655034485504.png

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 :grimacing:

This is nuts, MI100/200/300 cadence is impressive.

1655034362046.png

Previous thread on CDNA2 and RDNA3 here

 
Last edited:

Ranulf

Platinum Member
Jul 18, 2001
2,911
2,584
136
How much slower are these 9070 non xt models I'm seeing at NE? About $100 cheaper for some models compared to the XT version. Cursory glance says it is just gpu clock speed.
 
  • Like
Reactions: marees

poke01

Diamond Member
Mar 8, 2022
4,807
6,135
106
Knowing Apple they locked in supply years ago. Not that that will stop them from bumping prices a bit just to get a little bit more.
Lets see if they do, the memory pricing for the M5 Pro/Max next year should tell us a lot about where the market is heading . If it increases from $200 per 8GB, I don't think the market is going to get better at all in 2026.

At least companies like apple, sony sell in mass quantities, its DIY users that will get beaten the most
 

Ranulf

Platinum Member
Jul 18, 2001
2,911
2,584
136
Its Apple, even if they have a decent supply of ram stored up they will take advantage and raise prices if possible. It definately looks like it will be worse for 1-2 years than it was with the DDR4 shortage in 17-18.
 
  • Like
Reactions: Joe NYC and poke01

poke01

Diamond Member
Mar 8, 2022
4,807
6,135
106
Anyone with 9070 XT here and 3 monitor setups?

What’s the idle power consumption for the GPU? I get 38-40 watts which is stupidly high. The RTX 4070 Super same setup idles at 10 watts.
 

iLLusiveMan

Member
Dec 13, 2022
38
67
61
How ****** did AMD not fix this yet?

It’s been a problem since RDNA2 for me.
It's been a problem since forever on AMD GPUs. They only recently fixed idle power draw in RDNA4 (or maybe RDNA3 idk because I skipped that one) on single monitors that have non standard V.Blanking timings...so basically almost all monitors with high refresh rate.
So for multi monitor setups ? Probably in RDNA5/RDNA6 or who knows ? Soon™
 

SolidQ

Golden Member
Jul 13, 2023
1,542
2,547
106
Anyone know Xfx swift 9070XT is good? Looking for this card. Cheapest model for now
 

poke01

Diamond Member
Mar 8, 2022
4,807
6,135
106
Anyone know Xfx swift 9070XT is good? Looking for this card. Cheapest model for now
I have this model. It’s okay, stock fans are a jet engine.

I have it -20% power limited and a bit of undervolt. At 240 watts (it can go up to 304 stock and 374watts overclock) with a custom fan curve it’s very quiet.

But just a warning it’s get very loud at stock settings
 

soresu

Diamond Member
Dec 19, 2014
4,244
3,748
136

basix

Senior member
Oct 4, 2024
310
608
96
Never heard the name Andy Pomianowski before which seems odd considering he apparently led the gfx architecture development for 3 generations.
Interesting is also Laks Pappu, the second presenter:
- SoC architect of RDNA4
- SoC architect of RDNA5
- Currently SoC architect of Instinct GPUs

He seems to have done his job well, when he got moved to "more important" products like CDNA Instinct GPUs.
And it is also an indication (as it was already speculated), that RDNA5 and CDNA5 SoC, cache and memory architectures converge closer together.
 

poke01

Diamond Member
Mar 8, 2022
4,807
6,135
106
Interesting is also Laks Pappu, the second presenter:
- SoC architect of RDNA4
- SoC architect of RDNA5
- Currently SoC architect of Instinct GPUs

He seems to have done his job well, when he got moved to "more important" products like CDNA Instinct GPUs.
And it is also an indication (as it was already speculated), that RDNA5 and CDNA5 SoC, cache and memory architectures converge closer together.
RDNA4 is really good. I think RDNA5 will be even better
 

basix

Senior member
Oct 4, 2024
310
608
96
I mean the new RDNA5 features like Universal Compression, Neural Arrays and the revamped cache architecture will be useful for CDNA5 as well.
So bringing RDNA and CDNA closer together makes absolutely sense to me. And on top of that you get much better and easier software compatibility between those GPU branches, including performance optimizations.
 
  • Like
Reactions: Tlh97 and soresu

marees

Platinum Member
Apr 28, 2024
2,223
2,865
96
I mean the new RDNA5 features like Universal Compression, Neural Arrays and the revamped cache architecture will be useful for CDNA5 as well.
So bringing RDNA and CDNA closer together makes absolutely sense to me. And on top of that you get much better and easier software compatibility between those GPU branches, including performance optimizations.
You mean CDNA 6 (MI500 ?)

CDNA 5 (MI400) releases by jun-july next year. That is too early for RDNA 5