A new set of rumors have leaked regarding the next-generation Xbox and PS5 and the GPUs both consoles will bring to market. We’ve known the broad specs of both platforms for a bit — both use AMD GPUs and CPUs, with the GPU based on AMD’s most recent RDNA architecture, while the CPU is derived from the same 7nm Ryzen CPU cores that launched earlier this summer. What we’ve lacked is specific details on the GPU cores themselves.
Eurogamer has gotten their hands on some leak data they feel is fairly legit, and the website’s track record with this kind of information is solid. There have been some rumored APU configurations that leaked earlier this year, but this new data implies the Sony PS5 will feature 36 GPU clusters clocked at up to 2GHz. Supposedly the silicon, codenamed Oberon, is designed to operate in three different modes (Gen 0, 1, and 2) with clocks of 800MHz, 911MHz, and 2GHz respectively. Supposedly memory bandwidth is 448GB/s in Gen 2 mode (though 512GB/s is an alternate possibility) and the GPU can reportedly also be variably configured in terms of ROP and core counts. Eurogamer states:
While a 2.0GHz GPU clock is used for what is described as the fully unlocked ‘native’ or ‘Gen2’ mode, the processor is also tested in what is referred to as Gen1 and Gen0 modes. The former is explicitly stated as running with 36 compute units, a 911MHz core clock, 218GB/s of memory bandwidth and 64 ROPs – the exact specifications of PlayStation 4 Pro. The latter Gen0 mode cuts the CU and ROP counts in half and runs at 800MHz, a match for the base PS4. The indications are that back-compat is an integral part of the silicon, which in turn raises some interesting questions about the makeup of the Navi GPU and the extent to which older GCN compatibility may be baked into the design.
The implication here is that the PS5 SoC contains multiple GPU clusters, just like the PS4 Pro did. Using multiple GPU clusters in the same SoC would give Sony the same ability to turn the clusters on or off depending on which mode the GPU was running in. Alternately, the GPU cluster could be physically unified but designed to allow for this kind of fine-grained power gating. Stamping out identical clusters would be simpler, designing a unified cluster with fine-grained gating is probably more complex but saves on die space.
As for the Xbox Series X, Eurogamer is implying this console packs serious firepower. Here’s the rumored configuration:
If this rumor proves true — always something to keep in mind — the Xbox Series X will launch packing the equivalent of a high-end PC GPU. The largest GPU AMD has ever built are cards like the R9 Fury X and Vega 64, with 4096 cores. A 56-cluster Navi GPU would pack 1.4x more GPU cores than the 5700 XT, which already competes in the high-end PC GPU segment at the ~$ 400 price point. While AMD is expected to launch Navi 20 before the Xbox Series X debuts, we haven’t seen any indication that the company intends to dramatically expand the number of GPU cores it offers — Navi improved on GCN’s performance by making the individual cores more efficient as opposed to simply throwing more cores at the problem. It’s highly unlikely, in other words, that AMD would build a 56 CU for Microsoft and then ship a 128 CU design into the PC market.
If this rumor proves true, Microsoft is playing a far more aggressive game than it did last generation. Let’s assume, for the sake of argument, that AMD ships an 80 CU version of Navi 20, which comes out to 2x Navi 10. That would give the Xbox Series X 3,584 GPU cores compared to 5,120 for Navi 20, or about 70 percent as many.
In 2013, the Xbox One shipped with 768 GPU cores. The month before, AMD had shipped the R9 290X, with 2,816 cores. The PS4, at debut, had 1,152 cores. The Xbox had 27 percent as many GPU cores as the R9 290X, while the PS4 had 41 percent. While we can’t draw linear comparisons between console and PC performance strictly on the basis of GPU core count, the PC GPU was obviously far larger, with significantly more compute and graphics resources.
If — again, if — these rumors are true, the gaps are going to be a lot narrower this time around. The 1.7GHz clock speed on the Xbox Series X’s GPU is required to hit a supposed target of 12TFLOPS, but Eurogamer didn’t get that clock speed leak directly. The gap in GPU performance between the PS5 and XSX would be partially offset by faster clocks on the PS5, but only partially.
Frankly, the spec gap between the PS5 and XSX is large enough that you could argue the Xbox specs are less likely to be true. It’s also possible Microsoft decided to pull out all the stops after the disaster of the Xbox One. Doubling down on beating Sony in raw performance from Day 1 might represent Microsoft’s big idea for preventing a repeat of what happened last generation.
If the Xbox rumors are accurate there doesn’t seem to be a way for MS to sell the console at $ 400 without losing money — and I’ve got doubts about $ 500 as well, given that the system is expected to also use a high-speed NVMe-attached SSD and GDDR6. Hard drives might be slow, but they’re still cheaper than the equivalent amount of solid state storage. That doesn’t mean MS can’t pursue a loss-making strategy, but both MS and Sony opted not to do that with the initial Xbox One / PS4 after taking heavier-than-expected losses on X360 and PS3 (particularly in Sony’s case).
This kind of configuration would make a lot more sense if Microsoft is serious about a lower-end version of the console and intends to debut both. The PS5’s smaller GPU looks more like what we’d expect from a generational update. On the other hand, if this points to an upper-end Xbox Series X, it means that version of the console is going to pack high-end* PC-equivalent performance. With a 56-CU Navi, 8-core Ryzen 7nm CPU and 560GB/s of system memory bandwidth there’s no way it could perform like anything else.