Discussion RDNA4 + CDNA3 Architectures Thread

DisEnchantment · Mar 23, 2022

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

marees · May 24, 2025

inquiss said:
Right, supply and demand. I agreed. However costs also have been going up too, so if you were to establish supply, msrps would be going up because costs are going up too. That's part of the supply bit.

Costs are going up because TSMC is the only game in town

Plus I believe Nvidia has miscalculated/misjudged or something like that

So, What's Next...? The rate of advancement in gaming performance... (Part 2) Is PC gaming dying?!

A couple of years ago, I looked at the rate of advancement of GPUs over time . It's the only way of tracking how the user experiences techno...

hole-in-my-head.blogspot.com

Intel never fabbed a GPU in their fabs
Samsung's contract for AMD chromebook APU sonoma valley seems to have been scrapped

Hopefully things change
Intel in discussion with Microsoft
Samsung in discussion with Sony for PlayStation handheld (could be typical Korean FUD though)

inquiss · May 24, 2025

marees said:
Costs are going up because TSMC is the only game in town

Plus I believe Nvidia has miscalculated/misjudged or something like that

So, What's Next...? The rate of advancement in gaming performance... (Part 2) Is PC gaming dying?!

A couple of years ago, I looked at the rate of advancement of GPUs over time . It's the only way of tracking how the user experiences techno...

hole-in-my-head.blogspot.com

Intel never fabbed a GPU in their fabs
Samsung's contract for AMD chromebook APU sonoma valley seems to have been scrapped

Hopefully things change
Intel in discussion with Microsoft
Samsung in discussion with Sony for PlayStation handheld (could be typical Korean FUD though)

Yes, competition in foundry is another layer of supply and demand. Take that out and you're still left with the fundamental that cost per transistor is going up because it's harder to make them smaller. Chiplets in CPUs have caused a one time drop in cost structure. Need more tricks like that to reset costs for making things.

Chiplets can do that for GPUs if you can solve the workload. But it's much more fun to use them to make things you couldn't make monolithic.

Kepler_L2 · May 24, 2025

marees said:
Samsung in discussion with Sony for PlayStation handheld (could be typical Korean FUD though)

It's TSMC N3P

jpiniero · May 24, 2025

Kepler_L2 said:
It's TSMC N3P

That would be more realistic. The price of it is going to be interesting.

Thunder 57 · May 27, 2025

I was just looking at a comparison of AMD chipsets again. Seems it was a smart move for AMD to include PCIe x16 on the 9060 series as only the highest end chipsets support PCIe 5 from what I saw. Who knows how much one would lose by going down to PCIe 4, probably not much. But for the end user it would be free performance. And for those still rocking PCIe 3, I imagine the benefit would be quite a bit.

jpiniero · May 28, 2025

Thunder 57 said:
I was just looking at a comparison of AMD chipsets again. Seems it was a smart move for AMD to include PCIe x16 on the 9060 series as only the highest end chipsets support PCIe 5 from what I saw. Who knows how much one would lose by going down to PCIe 4, probably not much. But for the end user it would be free performance. And for those still rocking PCIe 3, I imagine the benefit would be quite a bit.

The 5060 Ti 16 GB didn't lose all that much dropping from 5.0 to 4.0.

marees · May 28, 2025

Vulkan & open cl scores for 9060xt

https://twitter.com/x/status/1927685691183423788

https://twitter.com/x/status/1927685642428817797

Thunder 57 · May 28, 2025

jpiniero said:
The 5060 Ti 16 GB didn't lose all that much dropping from 5.0 to 4.0.

The 8GB sure did.

SolidQ · May 28, 2025

marees said:
Vulkan & open cl scores for 9060xt

as expected ~7700XT/5060ti lvl

Shmee · May 29, 2025

SolidQ said:
as expected ~7700XT/5060ti lvl

Only being on par with 7700XT is a bit disappointing, for both these new cards IMO. Though I will be waiting for full reviews to pass judgement of course, and also depending on actual pricing they are available for. If I can really get a 9060XT 16GB for around $350, that could be an option to upgrade my 5700XT in my second comp for this year's Quakecon.

marees · May 29, 2025

Shmee said:
Only being on par with 7700XT is a bit disappointing, for both these new cards IMO. Though I will be waiting for full reviews to pass judgement of course, and also depending on actual pricing they are available for. If I can really get a 9060XT 16GB for around $350, that could be an option to upgrade my 5700XT in my second comp for this year's Quakecon.

Raster = 7700xt
RT = 7800xt

Hopefully

adroc_thurston · May 29, 2025

Shmee said:
Only being on par with 7700XT is a bit disappointing

N44 is 66% of N32 chop bandwidth.
Getting anywhere close is great.

Timorous · May 29, 2025

Shmee said:
Only being on par with 7700XT is a bit disappointing, for both these new cards IMO. Though I will be waiting for full reviews to pass judgement of course, and also depending on actual pricing they are available for. If I can really get a 9060XT 16GB for around $350, that could be an option to upgrade my 5700XT in my second comp for this year's Quakecon.

It is about what I would expect.

9070XT is ~45% faster than the 7800XT with 4 more CUs and a bit more memory bandwidth.

9060XT has the same number of CU's as the 7600XT and also has a bit more memory bandwidth so a 40% or so uplift is right in line with how the 7700XT performs.

That is obviously raster performance. In RT the 9060XT will perform ahead of the 7800XT most likely.

Vikv1918 · May 29, 2025

The most important thing for 9060 XT is that MSRP is actually followed by AIBs and retailers. Right now in the US, the 5060 Ti 8GB is $420 and 16GB is $470. If 9060 XT 16GB is really $350, it would be a complete no-brainer purchase in the entire $250-$450 price range.

jpiniero · May 29, 2025

Vikv1918 said:
The most important thing for 9060 XT is that MSRP is actually followed by AIBs and retailers. Right now in the US, the 5060 Ti 8GB is $420 and 16GB is $470. If 9060 XT 16GB is really $350, it would be a complete no-brainer purchase in the entire $250-$450 price range.

No reason to think it will be, other than maybe a couple units on Day 1. Price comparable to the 5060 Ti 8 GB most likely.

reb0rn · May 29, 2025

Not even kids waiting santa claus believe the AMD msrp prices

Mopetar · May 29, 2025

The problem with a $350 card trading blows with (or beating out) a $450 competitor's alternative is that scalpers will exploit that difference by buying up and flipping as many cards as they can manage.

gdansk · May 29, 2025

N44 8GB most likely has a lower BoM than the RTX 5060 due to the cheapo memory.
If the MSRP is fictional, it would be because Radeon (or their partners) are somehow taking more margin than Nvidia. Highly impressive if it turns out that way.

jpiniero · May 29, 2025

gdansk said:
N44 8GB most likely has a lower BoM than the RTX 5060 due to the cheapo memory.
If the MSRP is fictional, it would be because Radeon (or their partners) are somehow taking more margin than Nvidia. Highly impressive if it turns out that way.

Yeah but I am assuming the 9060 XT 16 GB is going to be comparable in price to the 5060 Ti 8 GB and not the 16.

Is the price of GDDR7 more than double than GDDR6? It's possible.

Plus it's also a bit bigger die.

SolidQ · May 29, 2025

AMD need redstone. Because RTX 5xxx with RR which add FPS

MrTeal · May 29, 2025

Shmee said:
Only being on par with 7700XT is a bit disappointing, for both these new cards IMO. Though I will be waiting for full reviews to pass judgement of course, and also depending on actual pricing they are available for. If I can really get a 9060XT 16GB for around $350, that could be an option to upgrade my 5700XT in my second comp for this year's Quakecon.

The 5060 Ti is basically half a 5070 Ti and the 9060XT is half a 9070XT, so slotting into that same performance class as the 7700XT and 5060 Ti is kind of expected.

reb0rn · May 29, 2025

The 4060 already had issue with slow 128bit gddr6, the 9060xt will have a lot more issue as its faster chip then 4060, so defo slower then 5060ti

adroc_thurston · May 29, 2025

reb0rn said:
The 4060 already had issue with slow 128bit gddr6, the 9060xt will have a lot more issue as its faster chip then 4060, so defo slower then 5060ti

You're comparing different microarchitectures with vastly different caching schemes. Don't.

reb0rn · May 29, 2025

I still stand what I said it range will be 5% more so in intensive memory use even more in ray trace but we know already know that
AMD if it was same or 1% faster would already brag hard, more so NV chip had better power/perf rate also

itsmydamnation · May 29, 2025

reb0rn said:
I still stand what I said it range will be 5% more so in intensive memory use even more in ray trace but we know already know that
AMD if it was same or 1% faster would already brag hard, more so NV chip had better power/perf rate also

So vega56 is better then everything sub 9070?

Discussion RDNA4 + CDNA3 Architectures Thread

Golden Member

Golden Member

Senior member

Senior member

Lifer

Diamond Member

Lifer

Golden Member

Diamond Member

Golden Member

Memory & Storage, Graphics Cards Mod Elite Member

Golden Member

Diamond Member

Golden Member

Junior Member

Lifer

Senior member

Diamond Member

Diamond Member

Lifer

Golden Member

Diamond Member

Senior member

Diamond Member

Senior member

Diamond Member