AMD to Skip RDNA 5: UDNA Takes the Spotlight After RDNA 4

[AleksandarK]

While the current generation of AMD graphics cards employs RDNA 3 at its core, and the upcoming RX 8000 series will feature RDNA 4, the latest leaks suggest RDNA 5 is not in development. Instead, UDNA will succeed RDNA 4, simplifying AMD's GPU roadmap. A credible source on the Chiphell forums, zhangzhonghao, reports that the UDNA-based RX 9000 series and Instinct MI400 AI accelerator will incorporate the same advanced Arithmetic Logic Unit (ALU) designs in both products, reminiscent of AMD's earlier GCN architectures before the CDNA and RDNA split. Sony's next-generation PlayStation 6 is also rumored to adopt UDNA technology. The PS5 and PS5 Pro currently utilize RDNA 2, while the Pro variant integrates elements of RDNA 4 for enhanced ray tracing. The PS6's CPU configuration remains unclear, but speculation revolves around Zen 4 or Zen 5 architectures.

The first UDNA gaming GPUs are expected to enter production by Q2 2026. Interestingly, AMD's RDNA 4 GPUs are anticipated to focus on entry-level to mid-range markets, potentially leaving high-end offerings until the UDNA generation. This strategic pause may allow AMD to refine AI-accelerated technologies like FidelityFX Super Resolution (FSR) 4, aiming to compete with NVIDIA's DLSS. This unification is inspired by NVIDIA's CUDA ecosystem, which supports cross-platform compatibility from laptops to high-performance servers. As AMD sees it, the decision addresses the challenges posed by maintaining separate architectures, which complicate memory subsystem optimizations and hinder forward and backward compatibility. Putting developer resources into RDNA 5 is not economically or strategically wise, given that UDNA is about to take over. Additionally, the company is enabling ROCm software support across all products ranging from consumer Radeon to enterprise Instinct MI. Accelerating software for one platform will translate to the entire product stack.



View at TechPowerUp Main Site | Source

 

[wNotyarD]

While (re)unifying the architecture may be a step forward, what AMD needs to do is unify the software stack and get ROCm from the highest Instinct to the lowest Radeon.

 

[AleksandarK]

While (re)unifying the architecture may be a step forward, what AMD needs to do is unify the software stack and get ROCm from the highest Instinct to the lowest Radeon.

Already in the works!

 

[Dawora]

World will be sad place if there is no fast GPUs at all.

 

[Redwoodz]

While (re)unifying the architecture may be a step forward, what AMD needs to do is unify the software stack and get ROCm from the highest Instinct to the lowest Radeon.

"Additionally, the company is enabling ROCm software support across all products ranging from consumer Radeon to enterprise Instinct MI. Accelerating software for one platform will translate to the entire product stack."

 

[wNotyarD]

"Additionally, the company is enabling ROCm software support across all products ranging from consumer Radeon to enterprise Instinct MI. Accelerating software for one platform will translate to the entire product stack."

I sure hope they deliver on the promise. Support for ROCm on consumer Radeon products should be widespread, instead of locked to the top-of-the-range cards. (But I admit missing the quoted text when I read the article earlier)

 

[AleksandarK]

(But I admit missing the quoted text when I read the article earlier)

Ive added that after. People should know, so thanks!

 

[GoldenX]

Wonder how much did Sony influence this decision.

 

[3valatzy]

I don't think "skipping" is the right word. The forth generation RDNA is the last RDNA, the next architecture will be considerably different.
There have been rumours circulating about this for quite a while...

AMD RDNA 5 a "Clean Sheet" Graphics Architecture, RDNA 4 Merely Corrects a Bug Over RDNA 3

AMD's future RDNA 5 graphics architecture will bear a "clean sheet" design, and may probably not even have the RDNA branding, says WJM47196, a source of AMD leaks on ChipHell. Two generations ahead of the current RDNA 3 architecture powering the Radeon RX 7000 series discrete GPUs, RDNA 5 could...

www.techpowerup.com

 

[lilhasselhoffer]

So...what was said earlier. AMD is only focusing on the volume production (read, lower price) GPUs for the next generation and getting their crap together behind the scenes. Cool. It's almost like someone there viewed the Nvidia and AMD slap fight for marginally profitable markets in premium pricing brackets and finally said "I'm in a downward spiral and need to make better decisions." Meanwhile, a cocaine fueled bear named Nvidia is tearing through the forest and searching for honey like Winnie the Pooh was a horror story.


God, I wish that stuff like this wasn't just a Scfy channel original abomination. The image of a bear foaming at the mouth, in a leather coat, is making me happy all of the sudden.

 

[3valatzy]

So...what was said earlier. AMD is only focusing on the volume production (read, lower price) GPUs for the next generation and getting their crap together behind the scenes.

Let's see:

Navi 10 Radeon RX 5600 XT 2020 -> +12% to Navi 23 Radeon RX 6600 (2021) -> +15% to Navi 23 Radeon RX 660(5)0 XT (2022) -> +1% to Navi 33 Radeon RX 7600 (2023) -> ??? empty space in 2024.

What a great execution with volume production.

Meanwhile, Steam hardware survey shows 0% market share for the above mentioned.

 

[Daven]

Let's see:

Navi 10 Radeon RX 5600 XT 2020 -> +12% to Navi 23 Radeon RX 6600 (2021) -> +15% to Navi 23 Radeon RX 660(5)0 XT (2022) -> +1% to Navi 33 Radeon RX 7600 (2023) -> ??? empty space in 2024.

What a great execution with volume production.

Meanwhile, Steam hardware survey shows 0% market share for the above mentioned.

Every GPU costs AMD the same at volume. AMD has 15% on steam. AMD took 15% of the TAM revenue. It’s not a lot but it’s something. Definitely room for improvement.

 

[Lew Zealand]

Let's see:

Navi 10 Radeon RX 5600 XT 2020 -> +12% to Navi 23 Radeon RX 6600 (2021) -> +15% to Navi 23 Radeon RX 660(5)0 XT (2022) -> +1% to Navi 33 Radeon RX 7600 (2023) -> ??? empty space in 2024.

What a great execution with volume production.

Meanwhile, Steam hardware survey shows 0% market share for the above mentioned.

7600 XT is 48% faster than the 5600 XT.
4060 Ti is 53% faster than the 2060 Super.

Both are advancing equally slowly at the bottom of the market but if AMD wants to increase market share, they need to do better than just match Nvidia.

 

[tpa-pr]

I've only gotten back into computer hardware relatively recently, so I'm not familiar with AMD's GCN hardware. Could someone please give me a layman answer on why the unification will be beneficial?

 

[Bigbigrick]

I don't think using RDNA2+RDNA4 is a feasible solution. It's clear that even after significantly increasing the cu in the PS5 Pro, its size remains almost unchanged—not to mention the additional cu added for PSSR. Why would they modify RDNA2 to 4/5nm when there's already an RDNA3 design available on 4/5nm?

 

[eidairaman1]

7600 XT is 48% faster than the 5600 XT.
4060 Ti is 53% faster than the 2060 Super.

Both are advancing equally slowly at the bottom of the market but if AMD wants to increase market share, they need to do better than just match Nvidia.

Do sponsoring and have better advertising

I've only gotten back into computer hardware relatively recently, so I'm not familiar with AMD's GCN hardware. Could someone please give me a layman answer on why the unification will be beneficial?

RDNA is not GCN, GCN Ended with Vega.

Use this link here and look up GCN architecture.

TechPowerUp

Graphics card and GPU database with specifications for products launched in recent years. Includes clocks, photos, and technical details.

www.techpowerup.com

 

[Guwapo77]

The only thing I'm worried about are the drivers... GCN had some really dark days with their drivers.

 

[Craptacular]

I've only gotten back into computer hardware relatively recently, so I'm not familiar with AMD's GCN hardware. Could someone please give me a layman answer on why the unification will be beneficial?

As AMD sees it, the decision addresses the challenges posed by maintaining separate architectures, which complicate memory subsystem optimizations and hinder forward and backward compatibility.

 

[Daven]

The good news from all the comments I’m reading is that we really want AMD to do well in the GPU space and the concern of not having good competition is real.

But a fair and free market is a two way street. If AMD makes a good product, consumers should be able to see that and reward them. Buying only from one company based on name only no matter what will never end well for any of us.

 

[evernessince]

The good news from all the comments I’m reading is that we really want AMD to do well in the GPU space and the concern of not having good competition is real.

But a fair and free market is a two way street. If AMD makes a good product, consumers should be able to see that and reward them. Buying only from one company based on name only no matter what will never end well for any of us.

The GPU market isn't free and fair. Software lock-in is extremely prevalent and their competitor has both horizontal and vertical monopoly leverage. Nvidia's share of the market exceeds the share that Bell Systems had before it was broken up for being a monopoly and that's before you consider that Nvidia is integrated into adjacent technologies to GPUs up and down the stack in both the hardware and software realms.

It's not necessarily the customer's fault if Nvidia twists the market so that certain proprietary features essentially require you to forgo even considering AMD (CUDA and AI for example) or heavily incentivizes it (games, VR). It's designed to be an inherently uneven choice. How good AMD's next gen architecture is almost irrelevant because these software gates that Nvidia has placed around industries that use GPUs will take a long long time for AMD to do anything about.

I very much doubt there will be any such rewarding for AMD "giving it a good shot" either. The average consumer purchases based on mindshare first followed by word of mouth, hype, and probably last actual specs / reviews. I also don't seem AMD being able to do anything about the software disadvantage they have. Things like that takes years to develop. Even in the CPU market it took 3 good gens just to put AMD on the map and that ecosystem has minimal software lock-in. I'd anticipate that it would take at least 4 good AMD GPU generations to put them back into the fight. Unless AMD has something revolutionary for gaming it makes sense for them to focus on enterprise / server / AI. Much less barriers, mindshare isn't nearly as big an issue, and higher margins to boot.

 

[Luminescent]

I saw a clip with someone from AMD who questioned why Nvidia "waste" silicon space for tensor cores and other stuff i don't understand.
Now look who isn't laughing anymore and wants some AI money, always late, always behind.
For the benefit of my wallet i really hope they succeed in the gpu market but right now i don't see it, no dlss equivalent, no ray tracing equivalent, they are so far behind it's questionable if they can even compete with a 3090 from 2020 in the latest games, i saw a detailed comparison of PSSR/FSR4 vs DLSS on digital foundry and they are years behind.

 

[StimpsonJCat]

Thanks for destroying Radeon, Raja. Anyone with a brain knew RDNA was all smoke and mirrors, and an architectural dead-end.

 

[Quicks]

World will be sad place if there is no fast GPUs at all.

World is already a sad place where you have the fastest GPU, but still run like crap because of unoptimized software.

Maybe they can finally try and optimise software again.

 

[Gameslove]

Cross finger for release a Radeon RX 9900 XTX 48 GB VRAM for maximum 1000 $(€)...I think will be great choice...

 

[TheinsanegamerN]

The GPU market isn't free and fair. Software lock-in is extremely prevalent and their competitor has both horizontal and vertical monopoly leverage. Nvidia's share of the market exceeds the share that Bell Systems had before it was broken up for being a monopoly and that's before you consider that Nvidia is integrated into adjacent technologies to GPUs up and down the stack in both the hardware and software realms.

It's not necessarily the customer's fault if Nvidia twists the market so that certain proprietary features essentially require you to forgo even considering AMD (CUDA and AI for example) or heavily incentivizes it (games, VR). It's designed to be an inherently uneven choice. How good AMD's next gen architecture is almost irrelevant because these software gates that Nvidia has placed around industries that use GPUs will take a long long time for AMD to do anything about.

I very much doubt there will be any such rewarding for AMD "giving it a good shot" either. The average consumer purchases based on mindshare first followed by word of mouth, hyper, and probably last actual specs / reviews. I also don't seem AMD being able to do anything about the software disadvantage they have. Things like that takes years to develop. Even in the CPU market it took 3 good gens just to put AMD on the map and that ecosystem has minimal software lock-in. I'd anticipate that it would take at least 4 good AMD GPU generations to put them back into the fight.

consumers are not zombies. If AMD continues to offer an inferior product, they will continue to not sell. This has been blatantly obvious for decades yet AMD still struggles with this concept. Nothing is stopping AMD from making their own alternatives. Sometimes these are successful (hair FX, mantle to vulkan) and others are flops (FSR, RT).

Ironically, you hit the real problem on the head. It will take 4+ good GPU generations to get consumers back. As in, 4+ generations without major screwups or missing products. The last time AMD had that was......2012. 2013 if you count hawaii. Over a decade ago. THAT is why consumers dont buy radeon. A major part of the mystical "mindshare" people like to blame is optics, being unable to consistently put out products makes consumers less confident in a brand. Guess what AMD has a problem with?

Look at me, I've got a 6800xt. 7900xtx was good but not worth the upgrade price, especially given AMD restricted 6000 series inventory and as a result that 6800xt was WAY too expensive. Now, AMD has no answer for me with the RX 8000s. So I have to wait yet another 2 years. If I want an upgrade, my only option is nvidia. This is how you lose consumers. You know why high end radeon doesnt sell? Because us high end radeon buyers are tired of waiting 8+ years for a proper upgrade, then being left in the cold again.

Unless AMD has something revolutionary for gaming it makes sense for them to focus on enterprise / server / AI. Much less barriers, mindshare isn't nearly as big an issue, and higher margins to boot.

The irony of saying the corporate server market has "much less barriers and mindshare" if friggin HILARIOUS. That market is the SLOWEST to adopt new tech, hence why so many still use xeons. You also need dedicated software support (EG, CUDA) which AMD has always struggled with.