Next Generation Xbox vs PS6: RDNA5 Architecture, 68CU vs 54CU, and Performance Outlook

PS6 and next generation Xbox adopt RDNA5 with varying compute units, memory bandwidth, and feature implementation across consoles.

Hardware by Masaru Hoshino on Feb 18, 2026

Both the next Xbox and the upcoming PlayStation6 from Sony are expected to use RDNA5-based graphics IP. Reports say the PS6 may use an early version of GFX13 instead of the full desktop version.

That issue has caused some disagreement, but the bigger technological picture shows that specification choices and system-level design decisions will be more important than small differences in IP. Uncertainty about when the launch will occur could also affect how significant these architectural differences really are.

Next Generation, Xbox, PS6, RDNA5 Architecture, 68CU vs 54CU, Performance Outlook, NoobFeed

Real Performance, Architecture Shifts, and Compute Units

One of the biggest changes in RDNA5 concerns how compute units are defined. One RDNA2 workgroup processor was equivalent to two compute units. A compute unit and a workgroup processor are basically the same thing in RDNA5. This means that comparing raw compute units between generations is not easy.

The difference may not seem that big at first: the next Xbox may have 68 compute units, while the PS6 has 54. But in terms of architecture, the uplift relative to older RDNA2-era designs might be significant. Because the workgroup logic has changed, a 16CU difference in RDNA5 terminology does not indicate a straightforward increase from previous generations.

What the Early GFX13 Fork Really Means

There was more talk once people said that the PS6 used an early version of GFX13. Clarifications show that the console supports about 99% of RDNA5's desktop capabilities; it is legitimate to call it RDNA5 even if it is not the full PC version.

That difference may not have a big effect from a technical point of view. Because platform holders have to approve hardware years before launch, custom console silicon generally locks in functionality earlier than desktop GPUs. In console development cycles, it's not often possible to wait for every desktop-oriented feature to be fully developed.

Customization vs. Standardization

It seems that the two platforms have different philosophical views. There are rumors that Microsoft is working on better PC alignment, possibly by adopting a design more like off-the-shelf RDNA5 implementations. That plan could make development more equal and make it easier to optimize across platforms.

On the other hand, others think Sony is focusing on customisation. The PS6 is said to focus on features such as neural arrays, a radiance cache, and ray tracing accelerated by machine learning. Mark Cerny has said that machine learning and ray tracing efficiency are two of the most important things for the next generation.

This difference shows that there are two different ecosystem strategies: one that emphasizes making things work like Windows-based gaming, and another that focuses on tightly integrating improvements specific to consoles.

Things to think about while setting up memory and bandwidth

Memory configuration is just as important as computing units. People say the PS6 will have 30GB of memory on a 160-bit bus, with a clamshell design, meaning the memory chips will be on both sides of the PCB. The estimated bandwidth is about 640GB/s.

At first sight, that might not appear like a big step up from better hardware from the current generation. However, RDNA5 adds universal compression improvements that could significantly increase effective bandwidth. Raw bandwidth metrics alone don't tell the whole story.

On the other hand, some people think the next Xbox could have up to 48GB of RAM, which suggests it would be priced higher.

What We Can Learn from the Current Generation

If you look at the PS5 and Xbox Series systems, you can see how they fit into the bigger picture. The PS5 launched with 2304 shading units, while the Xbox launched with 3328. But the distinctions between the platforms in the actual world have often been small.

The results were affected by things like clock speeds, ROP counts, engine optimization, and workload allocation. For instance, the PS5 might have high pixel fill rates in some cases, even though it had less shading units. This was because it went faster. Sometimes, the differences in performance had more to do with the game engines and how well developers optimized them than with how computers worked in theory.

The same thing will probably happen with the following generation.

When to launch and possible delays

Timing is another thing that makes things more difficult. There have been talks of moving the PlayStation 6 launch window to 2028 or perhaps 2029 because memory prices are rising. Longer deadlines can allow for changes to specifications, but making major architectural changes late in development is hard and costly.

In the past, delays haven't always necessitated major hardware changes. There are reports that the Nintendo Switch 2's schedule changed without any major changes to its specifications. If Sony changes its timeframe, it might improve the silicon rather than completely redesign it.

Next Generation, Xbox, PS6, RDNA5 Architecture, 68CU vs 54CU, Performance Outlook, NoobFeed

Desktop RDNA5 Background

Even with 68 compute units, the next generation of console hardware would still be much less powerful than the flagship RDNA5 desktop GPUs, which are said to have more than 150 compute units. That background is vital. Consoles want to perform well within limited cost and power constraints, not to be the best PC on the market.

The most important question is not whether consoles have GPUs as good as high-end PCs, but how well they combine power, memory bandwidth, compression technologies, and software integration at a competitive price.

Strategic Consequences for the Future

As more first-party games move to PC more quickly, hardware differences may not be enough to determine a platform's value. Integration with the ecosystem, price, feature sets, and development tools will all be very important.

Small differences in how GFX is used are not likely to affect the success of a generation. We should instead look at how each company balances machine learning acceleration, compute resources, clock frequencies, and memory configurations.

Ultimately, both consoles are set up to make significant improvements to architecture. It may not matter as much whether there is a full desktop GFX13 implementation or an early RDNA5 fork as how well each platform holder carries out its overall hardware and ecosystem plan.

Also, check our other PS5 Pro articles: