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

GaiaHunter · Mar 19, 2025

Have you tried to reset the sdrenaline app?

I've seen the app going funny before when you play too much with settings.

Shmee · Mar 19, 2025

I will say that recently, I have noticed the latest drivers crash a bit in some games on my 7900XTX. Now I was playing ETQW, but I had the frame rate limited to 270FPS. Anyway, I thought it was my undervolt, but then the game started crashing at default settings. So I set to default, then reduced max boost clock to 2800. Worked for a bit, but then it crashed again in WoT, so I backed down to 2700. So far so good, but I wonder if anyone has had instability with the latest drivers on RDNA3 or older cards?

itsmydamnation · Mar 20, 2025

Shmee said:
I will say that recently, I have noticed the latest drivers crash a bit in some games on my 7900XTX. Now I was playing ETQW, but I had the frame rate limited to 270FPS. Anyway, I thought it was my undervolt, but then the game started crashing at default settings. So I set to default, then reduced max boost clock to 2800. Worked for a bit, but then it crashed again in WoT, so I backed down to 2700. So far so good, but I wonder if anyone has had instability with the latest drivers on RDNA3 or older cards?

Yeah I feel latest driver does something, things that were stable now aren't. But no issues at stock.

marees · Mar 20, 2025

If AMD had launched Navi 44 last june then they would have had a one year lead !!!

https://twitter.com/x/status/1902630465741205719

tsamolotoff · Mar 20, 2025

Shmee said:
ETQW

Isn't it like ages old and also opengl ? You might try replacing opengl libs with dlls from pre 24.5.1 versions of the driver, some time after that they've changed something with ogl and now it doesn't work in fullscreen mode for me for more than half a year.

Shmee said:
Worked for a bit, but then it crashed again in WoT, so I backed down to 2700

Due to how power limited n3x cards are, you basically have to setup undervolt for each game (like it was on Vega56/64), with the generic rule of 'lighter the load, lesser the undervolt' (as GPU boosts higher if load is high enough (aka not CPU limited scenario) and requires more voltage)

PJVol · Mar 20, 2025

Shmee said:
I will say that recently, I have noticed the latest drivers crash a bit in some games on my 7900XTX. Now I was playing ETQW, but I had the frame rate limited to 270FPS. Anyway, I thought it was my undervolt, but then the game started crashing at default settings. So I set to default, then reduced max boost clock to 2800. Worked for a bit, but then it crashed again in WoT, so I backed down to 2700. So far so good, but I wonder if anyone has had instability with the latest drivers on RDNA3 or older cards?

Synopsis:
I experienced the same thing when I tried to play Callisto Protocol recently on my Navi 21 XT. The game started crashing, sometimes with a PC hard lock, and today it crashed at the cutscene without any pop-ups or gpu OD reset. Previously, going back to default VddGfx offset helped, but now it's crashing regardless.
FIX: )
Reverted back to the 24.10.1 with cleanup and so far so good, no hiccups with my daily OD and PP settings (Fmax +200, UV'ed Gfx via overdrive and Vsoc / VddMem / VddCI via powerplay). Heard a lot of complaints about stutters from people playing UE games with 25.3 driver.

tsamolotoff said:
with the generic rule of 'lighter the load, lesser the undervolt' (as GPU boosts higher if load is high enough (aka not CPU limited scenario) and requires more voltage)

actually it's just the opposite

Hans Gruber · Mar 20, 2025

If you are running Windows 11 24H2, that will screw up your gaming experience with Ryzen CPU's. Microsoft has not fixed this issue. AMD GPU's drivers were buttery smooth before 24H2.

Shmee · Mar 20, 2025

Interesting. Well so far 2700MHz core is working well, no crashes so far. Also I am on Windows 10.

But yeah sometimes I play older games like ETQW or Doom, though this driver is the first one that has given me any issues at stock in a while.

tsamolotoff · Mar 20, 2025

PJVol said:
actually it's just the opposite

Okay, if it is opposite, just run Vega with huge undervolt and some sort of light load (some old 2009 game, for example) and report your results. It is the same story as with CO "overclock" (it aint, really) with Ryzens - CPU is stable at high load (like prime95 or linx or w/e) with very negative CO, but crashes in idle because of overboost and insufficient voltage.

Josh128 · Mar 20, 2025

FAFO with OC/UV/driver updates is the moral of the story, I guess? I've learned that when I get a rock-solid stable system, I dont even upgrade drivers unless they are absolutely required for a new feature or support.

GodisanAtheist · Mar 20, 2025

Josh128 said:
FAFO with OC/UV/driver updates is the moral of the story, I guess? I've learned that when I get a rock-solid stable system, I dont even upgrade drivers unless they are absolutely required for a new feature or support.

-100%. Stability is the dragon we all chase. Once we catch it, no point in letting it go.

gaav87 · Mar 20, 2025

Shmee said:
I will say that recently, I have noticed the latest drivers crash a bit in some games on my 7900XTX. Now I was playing ETQW, but I had the frame rate limited to 270FPS. Anyway, I thought it was my undervolt, but then the game started crashing at default settings. So I set to default, then reduced max boost clock to 2800. Worked for a bit, but then it crashed again in WoT, so I backed down to 2700. So far so good, but I wonder if anyone has had instability with the latest drivers on RDNA3 or older cards?

Hm... interesting check what clock and voltage it wants/reporting at stock voltage settings in drivers and if undervolting actualy lowers the reported voltage or it behave same as on rdna4 ?

gdansk · Mar 20, 2025

9070 XT is the most "bimodal" graphics card I have used.
In some games it performs like a super 7800XT and in others like a 7900 XTX.
Really wish AMD had enough volume/money to get Samsung to make that vaporware memory. Or to get Hynix to actually do it right.

Abwx · Mar 20, 2025

UV and OC at Computerbase, basicaly UV alone indeed overclock the card.

OC und UV auf der AMD Radeon RX 9070 XT im Test

Was ist bei der AMD Radeon RX 9070 XT mit Navi-48-GPU mittels Overclocking und Undervolting herauszuholen? ComputerBase hat es getestet.

www.computerbase.de

Thunder 57 · Mar 20, 2025

Abwx said:
UV and OC at Computerbase, basicaly UV alone indeed overclock the card.

OC und UV auf der AMD Radeon RX 9070 XT im Test

Was ist bei der AMD Radeon RX 9070 XT mit Navi-48-GPU mittels Overclocking und Undervolting herauszuholen? ComputerBase hat es getestet.

www.computerbase.de

No surprise. AMD has been juicing their cards since Polaris. I'm not sure why they can't figure out how to give them proper V/F curves.

gdansk · Mar 20, 2025

Thunder 57 said:
No surprise. AMD has been juicing their cards since Polaris. I'm not sure why they can't figure out how to give them proper V/F curves.

It's been stated already in this thread. There's gamer stable undervolts and then there's what AMD can actually ship if they want it to work.

Thunder 57 · Mar 20, 2025

gdansk said:
It's been stated already in this thread. There's gamer stable undervolts and then there's what AMD can actually ship if they want it to work.

I'm not talking about excessive undervolts, rather mild ones. "Gamer stable" is what some might be trying but when I had Polaris mine was rock steady with a mild undervolt.

PJVol · Mar 20, 2025

Thunder 57 said:
I'm not sure why they can't figure out how to give them proper V/F curves.

Because PVT variance is still a thing?

adroc_thurston · Mar 20, 2025

Thunder 57 said:
I'm not sure why they can't figure out how to give them proper V/F curves.

Oh they can.
But they like it cheap.

HuesToo4 · Mar 20, 2025

Announcing DirectX Raytracing 1.2, PIX, Neural Rendering and more at GDC 2025! - DirectX Developer Blog

Today in the DirectX State of the Union at GDC 2025, we proudly showcased the next evolution in graphics with the announcement of DirectX Raytracing (DXR) 1.2. This update promises groundbreaking performance improvements and breathtaking visual fidelity, marking another milestone in our mission...

devblogs.microsoft.com

SolidQ · Mar 21, 2025

Interesting on slide, every company mentioned, about Omm, ser, but not AMD.
If AMD gonna update Omm, that should boost Wukong performance

adroc_thurston · Mar 21, 2025

SolidQ said:
Interesting on slide, every company mentioned, about Omm, ser, but not AMD.

AMD isn't interested in doing FF slop.

SolidQ · Mar 21, 2025

adroc_thurston said:
AMD isn't interested in doing FF slop.

And they gonna be drastically slower in some games

adroc_thurston · Mar 21, 2025

SolidQ said:
And they gonna be drastically slower in some games

you can just do things (make better shader cores).

branch_suggestion · Mar 21, 2025

HuesToo4 said:
Announcing DirectX Raytracing 1.2, PIX, Neural Rendering and more at GDC 2025! - DirectX Developer Blog

Today in the DirectX State of the Union at GDC 2025, we proudly showcased the next evolution in graphics with the announcement of DirectX Raytracing (DXR) 1.2. This update promises groundbreaking performance improvements and breathtaking visual fidelity, marking another milestone in our mission...

devblogs.microsoft.com

Very meh, just codifying last gen NV stuff, not much forward looking.
Gotta wait for 10th gen stuff with GFX13 for the really big DX updates.
Still theoretically CV would allow for enough access to the matrix hw that all ML upscalers should work across all vendors.

Work Graphs is still the best new feature in a while, reducing the CPU bottleneck will be necessary for beyond reticle MCM GPUs to be viable in gaming.

Discussion RDNA4 + CDNA3 Architectures Thread

Golden Member

Diamond Member

Memory & Storage, Graphics Cards Mod Elite Member

Diamond Member

Senior member

Senior member

Senior member

Platinum Member

Memory & Storage, Graphics Cards Mod Elite Member

Senior member

Senior member

Diamond Member

Senior member

Diamond Member

Lifer

Diamond Member

Diamond Member

Diamond Member

Senior member

Diamond Member

Member

Golden Member

Diamond Member

Golden Member

Diamond Member

Senior member