Support NeoGAF

[Loxus]

lol Loxus making a statement with YouTube video

???

 

[Elios83]

So basically RDNA4 = PS5 Pro tech with almost simultaneous launches.
Don't know if FSR4 will be ready to take advantage of these new AI capabilities that Sony seems to be using with their own PSSR trained algorithm.

 

[SolidQ]

So basically RDNA4 = PS5 Pro tech with almost simultaneous launches.

PS5pro is RDNA 3.5WGP, RDNA4 is difference WGP

 

[Elios83]

PS5pro is RDNA 3.5WGP, RDNA4 is difference WGP

I meant it has what matters of RDNA4, the new ray tracing blocks and AI stuff.
Surely some building blocks are taken from different iterations.

 

[SolidQ]

AMD FidelityFX SDK update brings FSR 3.1.2 and native Microsoft GDK compilation for Xbox Series X/S - VideoCardz.com

FidelityFX SDK update enables native compilation for consoles AMD continuing its open-source efforts by sharing an updated Software Development Kit (SDK) for its FidelityFX platform. We should remind readers that FidelityFX is not just the upscaling components, it is a suite of technologies...

videocardz.com

I meant it has what matters of RDNA4, the new ray tracing blocks and AI stuff.

That need more detail analyze between PS5pro RT(on RDNA 3.5 WGP) and RDNA4 RT(on R4 WGP)

 

[deeptech]

I was planning a 4070 soonish perhaps, but I'll wait to see what AMD does with this. If they manage to up RT perf to somewhat competitive levels and keep prices lower than NV, then that's great I think.

 

[Corporal.Hicks]

I hope they can come through with massive RT performance gains. If this generation has taught us anything, it’s that RT is only really noticeable and worth the cost when you use gobs and gobs of it or just full path tracing. And in games with heavy RT/path tracing AMD gets destroyed.

If you have a basic knowledge of light, you will easily notice RT effects. Even RT reflections make a big difference, and the performance cost is relatively small on high-end GPUs.

RT off

RT reflections + RT shadows

278fps vs 220fps is not a big relative difference and the graphics look much better.

RT off

RT reflections + Shadows

I recently upgraded my PC and although I bought an AMD CPU, I did not even consider AMD GPUs because of their poor RT performance. If AMD want to win customers like me, they are going to have to close the gap on RT performance and upscaling features.

 

[llien]

Don’t RDNA2/3 cards lose more performance than Nividia cards once you turn on Raytracing? How do you account for the relatively larger loss in performance if not for a comparable inefficiency somewhere

The same way I account for 4090 losing more fps than 3050, when going from 7800X3D to Celeron n3450.

Although I might bee too old to realize that 4090 is less efficient than 3050 as it would drop more FPS, cough.

 

[llien]

I'll wait to see what AMD does with this. If they manage to up RT perf to somewhat competitive levels

7900GRE is about 10% behind 4070 and about 25% behind 4070sup in tpu benchmarks.
You can always find games where the gap is bigger, or where AMD wins at RT.
This is normally 'countered' by inventing new metrics to judge "RT efficiency"...

 

[Corporal.Hicks]

7900GRE is about 10% behind 4070 and about 25% behind 4070sup in tpu benchmarks.
You can always find games where the gap is bigger, or where AMD wins at RT.
This is normally 'countered' by inventing new metrics to judge "RT efficiency"...

If the game is built to take advantage of all the RT performance saving features (like OMM) built into the RTX 40-series card, the RT performance gap is massive compared to RDNA3 cards.

At 4K + quality upacaling + full RT the 7900XTX has 8fps while the RTX4090 42fps (my card has 35fps). With some tweaks (very high settings instead of cinematic + DLSS FG) the game will run at 80-100fps on the RTX4090, but if you have RDNA3 card there's nothing you can do to play this game with RT.

 

[llien]

If the game is built to take advantage of all the RT performance saving features (like OMM) built into the RTX 40-series card

It is an elegant way to refer to a green sponsored game.

4090 running something at 41 fps at fake 4k is, possibly, a sign of particularly awesome coolness which I am too old to grasp.

Things are mighty impressive even without "hardwah RT" it seems:

Black Myth: Wukong - Ray Tracing & DLSS 3 Benchmarks

In this article, we're going to share our impressions, comparison screenshots and benchmarks for its Ray Tracing effects and NVIDIA DLSS 3.

www.dsogaming.com

18 fps without upscaling... was it developed for 5000 series?
Also, the bazinga-efficiency-metric seems to be pretty bad, going from 38 to 18 fps, right?

Any particular reason you've picked up this game as an example?

 

[winjer]

BTW, this is from the new video from AMD:

We are also partnering deeply with our close friends at Activision
to deliver the absolute best experience on Call of Duty: Black Ops 6, with game optimizations and integration of FSR 3.1. We're also working very hard to enablethe next generation ML-Based FSR on Call of Duty: Black Ops 6.

FSR with AI might be coming sooner than we thought.

 

[Corporal.Hicks]

It is an elegant way to refer to a green sponsored game.

4090 running something at 41 fps at fake 4k is, possibly, a sign of particularly awesome coolness which I am too old to grasp.

Things are mighty impressive even without "hardwah RT" it seems:

Black Myth: Wukong - Ray Tracing & DLSS 3 Benchmarks

In this article, we're going to share our impressions, comparison screenshots and benchmarks for its Ray Tracing effects and NVIDIA DLSS 3.

www.dsogaming.com

18 fps without upscaling... was it developed for 5000 series?
Also, the bazinga-efficiency-metric seems to be pretty bad, going from 38 to 18 fps, right?

Any particular reason you've picked up this game as an example?

If it was an AMD sponsored title, Black Myth Wukong would not even use Path Tracing, becasue RDNA3 cards dont have necessary features to run extremely demanding PT at reasonable framerate. Even the developer made a video about that. Read about "opacity micro maps" (OMM) build into RTX40 series. This feature alone can speed up RT performance by up to 10x when there's a lot of vegetation in the scene, which is why even RTX30 series cards perform much worse than the RTX40 series. RDNA3 also does not have this feature and that's the reason why we such drastic performance difference in this game.

Black Myth Wuking is pushing RT technology to the max. The game is very demanding at maxed out settings with PT, but it's also very scalabe. If you only have RTX40 series card, you can play this game with RT and have an amazing experience.

Below you can see my screenshots and benchmark results (4K DLSS performance + very high settings + fullRT). The game is perfectly smooth and responsive at 80-90fps and I don't care that it's not running at 4K native because the image quality still looks like 4K to my eyes. If I would not tell you that these are 4K DLSS performance screenshots even you would think these are 4K TAA native screenshots.

If I would choose 7900XTX instead of 4080S, I would have a much worse RT experience in PT games like black myth wukong, or even RT games like cyberpunk or the witcher 3.

 

[deeptech]

BTW, this is from the new video from AMD:

..
FSR with AI might be coming sooner than we thought.

Will AI FSR will require latest hardware?

 

[winjer]

Will AI FSR will require latest hardware?

We still don't know.

 

[llien]

"opacity micro maps" (OMM) build into RTX40 series

Thanks. I've got to discussion titled "RT still not worth it", which is probably not what you meant.
Could you link to the developer talking about AMD missing the needed bits to make a performant game.

Black Myth Wuking is pushing RT technology to the max. The game is very demanding at maxed out settings with PT

I'd say a game that does 38 frames per second without any "hardwahr RT" enabled is simply pretty demanding.

the image quality still looks like 4K to my eyes

Might be compression artefacts, but the face/hair look quite blurred.

 

[Corporal.Hicks]

Thanks. I've got to discussion titled "RT still not worth it", which is probably not what you meant.
Could you link to the developer talking about AMD missing the needed bits to make a performant game.


I'd say a game that does 38 frames per second without any "hardwahr RT" enabled is simply pretty demanding.


Might be compression artefacts, but the face/hair look quite blurred.

You need to enable english subtitles. The raytracing and OMM segment starts at 7:23.



As for hair quality, here's a comparison. It seems that the hair looks more noisy with DLSS performance. It's not something I notice during gameplay (because the character and his hair move all the time), however I can definitely notice a much higher frame rate (83fps vs. 33fps) when I use DLSS performance. People who want to play with more detailed hair can always buy $2000 RTX4090 and use DLSSQ instead of performance .

I still think the 4K DLSS looks incredible in this game, especially in comparison to the PS5 version.


PC 4K DLSS performance (1080p internally) + mod that removes excessive sharpening this game has.

Playstation 5 (performance mode)

 

[SolidQ]

TensorWave plans world’s largest AMD GPU cluster with gigawatt capacity by 2025 | igor´sLAB

TensorWave, a cloud service provider focused on AMD’s AI technology, has announced that it is developing GPU clusters based on AMD’s Instinct MI300X, MI325X and MI350X accelerators. The aim is to…

www.igorslab.de

AMD UDNA "Radeon" Gaming GPUs Rumored To Enter Mass Production In Q2 2026, Sony PS6 Also Expected To Utilize Next-Gen Architecture

AMD's UDNA architecture-based Radeon gaming GPUs will allegedly enter mass production in Q2 2026 & also be integrated into Sony's PS6.

wccftech.com

• There is no RDNA5 code name. After RDNA4, it will be UDNA.
• MI400 and RX9000 use the same UDNA, and the architecture uses an ALU design similar to GCN.
• UDNA Gaming GPU is tentatively scheduled for mass production in Q2 26.
• Sony's PS6 will use UDNA, and the CPU has not yet been determined to be ZEN4 or ZEN5. Sony's handheld will also use AMD hardware.
• I heard that Microsoft's handheld will choose between Qualcomm and AMD? I don't know about this
• The above information comes from the supply chain, and I don't know the specifications and performance.

 

[MikeM]

BTW, this is from the new video from AMD:




FSR with AI might be coming sooner than we thought.

I hope it works on RDNA3

 

[SolidQ]

DNA has wide matrix cores and other wide compute workload improvements, which AMD wants to bring to UDNA. It also has multi-chip scaling.

Rumors tell that RDNA4 will finally have matrix cores in consumer space. Seems that AMD is integrating matrix cores early to RDNA lineup.

My expectation is that UDNA compute unit will be RDNA4 descendant instead of CDNA3 descendant. They definitely need 1 cycle low latency scheduling in consumer space, and Nvidia does well with it in AI space too. I don't see them going back to GCN-style design for UDNA.

Integrating matrix cores to RDNA4 compute unit would be an iterative step towards UDNA. They could iterate that compute unit further to meet AI workload.

I would expect UDNA to borrow lot of CDNA3 tech for caches, memory controllers and connectivity with chiplets.

There's rumors that RDNA4 will be a small iterative improvement as they are ditching RDNA arch one year later. But i'd say it's entirely possible that RDNA4 is a an iterative step towards UDNA. First time they merge new RDNA ALU pipes with CDNA matrix units. Could be a big step.

It's important to consider the importance of AI for AMD. RDNA architecture was initially designed for AMDs most important market segment (gaming), while UDNA got the old GCN design. Now AI is most important segment for them. They want newest arch for professional AI chips too.

One might argue that simple GCN-style scheduling is still good for AI workloads, but lower latency scheduling can operate better on register pressure. And it's often better for caches to avoid running too wide workload on every compute unit. RDNA/Nvidia style arch is just better.

UDNA makes perfect sense. AMD today wants matrix units for AI based upscaling (and other client AI workloads) in consumer space. And they want their latest compute unit architecture for AI chips. AI is now a priority. GCN-style arch is outdated. Nvidia/RDNA-style is better.