What to expect with a PS5 pro... if they even make one.

 Wild Speculation

I’ve seen a lot of speculation about what specs the PS5 Pro will have. Most of that speculation has been people asking themselves “What specs would I like to see in a much more powerful console” and for the most part, are describing what would amount to a new generation. I will try to bring the speculation back to some reality by instead asking “what makes sense given everything we know about Playstation development, what reasons have there been for mid-generation spec improvements in the past, and what reasons might motivate a drive for a mid-generation upgrade this generation?”

Before we get into what’s possible for the PS5 Pro, we need to establish some things.

Development difference between PS5 and PC

This point will be a recurring theme, but it’s very important for this to be understood early. Playstation is *not* a PC, and the API that it uses to send commands from the game engines to the hardware is *not* similar to DX12 or Vulkan, which are designed primarily to scale well across hardware not only of very different performance levels, but even from different chip designers.  Playstation’s API is a lot lower level than DX12; developers are working much closer to hardware. This means changes in hardware have a greater impact on game compatibility than they do on PC. For example, among the many things we heard about from mark Cerny in his famous Road to PS5 event, he discussed backwards compatibility (BC). While the PS3 achieved PS1/2 BC through having a literal PS2 SoC inside which ended up getting removed in further iterations after a cost-down, the PS5 was designed to support PS4 backwards compatibility natively in hardware. However, due to the nature of the low-level programming of Playstation’s APIs, perfect backwards compatibility was not possible, and there were a handful of PS4 games that are officially unsupported on the PS5, despite both consoles being x86 with an AMD Radeon GPU. Among the hardware decisions to support BC was the choice to use the same number of compute Units as the PS4 Pro, which notably was exactly double that of the base PS4. The PS4 had 18, the PS4 Pro had 36, and the PS5 has 36 Compute Units. This is not by chance but by design. Further evidence of the lack of scalability in PS’s API is the fact that while PS4 games can generally run on the PS5, the far more powerful hardware of the PS5 does not improve performance over the PS4. It’s important to remember that the Playstation is not a PC, and you cannot simply throw faster hardware at it and get better performance. This is also why Xbox has so many more titles that are enhanced by Series X/S than the pS5 has; Xbox uses DX12; it’s trivial to get XBOne games enhanced on the XSS/S compared to the PS5.

I have to acknowledge that it’s possible that when they designed the API for PS5, it was a higher-level APU that could allow them to be a bit more lenient with PC hardware for future upgrades. However, with a hypothetical PS5 Pro, they’re still restrained by the need to support PS4, which was a major driver in the architecture in the PS5. Mark Cerny explained that while the PS5 uses RDNA2, it’s not simply off-the-shelf RDNA2. It’s a customization designed to emulate the PS4 in hardware. Any PS5 Pro would have to go to similar lengths to ensure PS4 compatibility as dropping this backward compatibility is no longer an option given the strategy uncovered with PS+ offering PS4 games run locally as part of the service. This means the cost of developing a PS5 Pro would be very high as it’s not just slapping together some AMD IP with the PS5 ASICS onto a chip. They’d have to collaborate to co-develop a GPU architecture for their needs much the same way they did with the PS5.

Past ‘Pro’ versions

To learn about what to expect in the future, we have to examine the past. Not only what was done, but why it was done. There’s 3 relevant examples we need to look at.

·         PS4 vs PS4 Pro

·         Xbox One vs Xbox One X

·         Xbox Series S vs Xbox Series X

There’s something interesting we can learn from each of these.

The following 3 examples, the pS4 Pro, Xbox One X, and Xbox Series X, all expanded on the ‘base’ console by offering a more power GPU for 4k over 1080p rendering. Let’s take a deeper look.

The PS4 Pro only existed at all because when the PS4 came out, 4k TVs were not only expensive, but it was very difficult for GPU tech to push that kind of resolution. A few years in however and a couple of GCN generations later, and 4k was not only common and affordable for TVs, but the GPU tech to push the resolution was accessible. Sony, however, made it extremely clear that this Pro version remained distinctly within the console generation. This meant the set of games that played on the PS4 had to be identical to the set of games that played on the PS4 Pro. Not only was it fully backwards compatible with PS4 games, but no games that ran on the PS4 Pro could be unplayable on the base PS4, so in a sense the PS4 had to remain forward compatible with the PS4 pro. To support this, you’ll notice the hardware changes are measured. While the Pro had a very different GPU, as mentioned before it *exactly* doubled the Compute Units (intentionally, for compatibility) and it also maintained the exact same old CPU architecture. It didn’t upgrade to Bulldozer or anything else. It kept the same amount of RAM, and it kept the same old spinning rust hard drive. The RAM and CPU were overclocked, but nothing was fundamentally changed. This was done to keep game development as easy as possible. Remember, this isn’t a PC where new things run the same APIs but faster. Playstation APIs are very low-level, so compatibility and ease of development is a concern. The intention with the PS4 Pro was that game developers should have to do as little as possible to ensure the game worked on both systems. They could make it for the PS4 alone, or if they wanted, they could support a higher render resolution on the Pro, but otherwise they ran effectively the same. Inter-compatibility was very easy.

The Xbox One X, like the PS4 Pro, existed to support the new wave of 4k TVs everyone was buying. However, it too like the PS4 Pro upgraded only the GPU while keeping the majority of the rest of its internals intact. Why didn’t Xbox work with AMD to produce an early Zen CPU (I realize XOX came out before Zen, but it may have been possible for Microsoft to have worked with AMD to create a custom solution using components from Zen similar to what they did with RDNA2) or use an SSD, or significantly change the RAM? Basically, it comes down to how Xbox still will basically want parity between their consoles. Even though they’re basically PCs and they could scale visual features and performance between them far easier than with Playstation, Xbox doesn’t WANT one console to be a significantly different experience from the other. While they want people to have an upgraded experience on the more powerful version, if the differences are too great, it could impact sales as they still want people buying the higher-volume smaller console because they make a lot of money off software sales on their platform; 100% of their own software and 30% off third party sales. They also have to consider the presence of hardware generations. A mid-generation upgrade is distinctly *not* a new generation. Imagine how awkward it would have been if the One X had an SSD, 16GB of RAM and a Zen CPU, and absolutely *could* run early Xbox Series games just as well if not better than the Series S but since it was last gen, it got thrown by the wayside. It would feel awkward and arbitrary, and the Series S would have felt like a worse One X that arbitrarily get supported longer… in fact many people already feel that way, though this sentiment is one borne of ignorance.

The Xbox Series X and S honestly serve mainly to reinforce most of the ideas presented with the One X. The S and X are intended to offer exactly the same experience except the X at a higher resolution. This is why they have identical CPUs and SSD architectures. It’s important to note that the Series S, despite on paper having a slower GPU than the One X, is a true 9^th gen console, while the One X is a mere 8^th gen. This is because the far more advanced CPU, SSD, and GPU architecture will allow the Series S to support game development techniques that the One X simply does not, with its lethargic CPU and spinning rust HDD holding it back. One difference with the Series S compared to the X though is its not only small but incredibly slow RAM compared to the Series X. This flies in contrast to the points I’ve made about parity… however, despite this difference, Xbox still claims it should offer the same gaming experience but at a lower resolution. But the single most significant complaint from developers this generation is about specifically the RAM situation with the Series S.

A Pro version can correct Base console missteps?

One thing I’ve heard several times that I need to address right away is that people like to argue that a Pro needs to happen because of specific shortcomings of the PS5 base console. Things like poor Ray Tracing performance, or an opportunity to increase RAM, or things like this. People need to clearly remember that if a Pro *does* happen, it is *not* an opportunity to correct things people perceive as missteps or mistakes. If they massively improve Ray Tracing performance for example, it does not make development easier because now they’re free to use more Ray Tracing effects or are less constrained by RT performance. It makes it harder because they now have to develop two separate RT implementations, and still have to deal with making RT work on the old hardware. It fixes absolutely nothing. Furthermore, they can’t just stop supporting RT on the version for the old consoles because RT was a key selling feature. There have been cases of developers skipping Ray Tracing support on the Xbox Series S, but that has more to do with the anemic RAM (they can drop the layer of textures required to support RT reflections and such saving some space) than it does the console’s RT perf, and you won’t see RT dropped by first-party.

But PC GPU tech is moving so fast!

This one is pretty obvious. The PS5 and Xbox One X are around 10TFLOPS. The 6800XT is 20TFLOPS. The 3090ti is ~40 (though not much faster than the 6800XT; remember TFLOPS aren’t everything, but it’s a *very* rough guide) The next generation from AMD and Nvidia are supposed to be in the vicinity of 70-90TFLOPS, so the generation after that, coming 2023-2024 should be well over 100TFLOPS. Given this sudden rapid evolution of GPU tech, wouldn’t it be expected that Sony and Microsoft will look at that with envy for their consoles? Maybe, maybe not. Let’s explore. First of all, despite an expected doubling of performance with RDNA3 over RDNA2 (and something similar with Nvidia’s Lovelace) it doesn’t mean the consoles will be able to see a doubling in performance with that tech; Consoles are limited by their form factor; they don’t like to go over a certain heat threshold because large heat syncs inflate size, noise, and production cost. On PC, GPU size and TDP are inflating massively. What used to be a soft barrier of ~250W has been pushed to 350W then 450W. It seems likely we’ll see 600W before this year is out, but this type of growth is not sustainable on consoles. RDNA3 is likely to be closer to 50% performance per watt increase over RDNA2, so they could maybe see 50% faster. Another 50% increase with RDNA4 would get them to a mere ~2.25x increase in performance over the current generation by 2023-2024, which is around the timeframe one might expect a Pro version to come out. While this is more or less in line with what we saw out of the GPU performance improvement with the PS4 Pro over the PS4, it’s nowhere near as flashy as the massive TFLOPS numbers make it out to be, sounding like >10x perf could be possible. That would be a pipe dream.

What about a 40CU PS5 Pro?

There’s an idea floating about that a PS5 Pro could use a die shrink on the same GPU arch, overclock it more, and use the full 40CU die since the current GPU is 40CU, but cut down to 36. Well, as previously mentioned, this is simply not possible due to the nature of the PS4 and PS5 APIs being so low-level. The PS5’s 36CU architecture was intentionally designed for backwards compatibility, being exactly double that of the base PS4 and the same count as the PS4 Pro. Any theories about the PS5 Pro not being an even multiple of 36 should be immediately dismissed. This could present a reasonable way of evaluating rumors or claimed ‘leaks’ in the next couple years as well. If someone claims an insider ‘leak’ with something other than an even multiple of 36 for a PS5 Pro, call them on their BS.

Is Sony working on a new PS5?

Yes, absolutely. There are rumours and supposed leaks that they’re working on one, but we don’t even need that; it’s obvious that they must be. Many consoles, including every Playstation generation to date, has released a later, smaller version with significant changes, later on even making use of newer fab tech to reduce the size, cost, and heat/power required by the design. At the same time as the PS4 Pro was launched, so too was the PS4 Slim which, using the same advanced process node, was able to make a chip that produced less heat and consumed less power, so a thinner version was released. Given the massive size of the PS5 and need for liquid metal to help dissipate the heat, I would consider it an extremely controversial claim that Sony isn’t working on a redesign including a die shrink that would reduce production and shipping costs. However this is not evidence that playstation is releasing a Pro, though it’s also not evidence that they are not.

Why would Sony make a PS5 Pro?

Look back to the previous generation; why did they do a Pro then? They effectively released the same console but with a more powerful GPU because they needed to target 4k. The PS5 is already advertised to hit 4k. What would adding a more powerful GPU to the PS5 accomplish? Higher frame rates? The issue with frame rates is not only are they hard to advertise because they’re so inconsistent, and again, the PS5 is already advertised as supporting 120Hz refresh rates. Sure most games never hit that, but the point is Sony advertises it. This makes it extremely difficult for them to market a more powerful console. They can claim “it’s more powerful” but that’s not flashy, people will want to know how without having to think about it. With the PS4, claiming 4k was a very simple way to promote it. With the PS5 what can they claim without admitting that the base console doesn’t accomplish what it’s advertised to do? “Better Ray Tracing!” will bring the question “I don’t understand, you told me the base PS5 was capable of Ray tracing, what’s the difference?”

And what would a PS5 Pro look like?

There is one way I can see Playstation releasing a PS5 Pro though. As previously mentioned, it’s far easier for Xbox to release another console generation. While I also have reasons for suspecting Xbox may not be eager to release one, they’re *far* more likely to pull the trigger because of how much easier performance scales, making that performance easier to market. While I don’t think Playstation wants to release a PS5 Pro, it’s possible that they’re preparing one in case Microsoft forces them to release one. If I had to guess, the most likely situation we get for a PS5 Pro would do a fundamentally RDNA2 but with features and IP blocks taken from RDNA3, but being careful to keep the PS4 hardware emulation, and maybe adding a few features. Something along the lines of minimum effort to get 2-2.5x performance, but they probably wouldn’t even start taping it out until they knew the Xbox upgrade was coming.

Conclusion

Ultimately, I think people tend to be assuming that since something happened last gen, it must be happening this gen. They’re asking themselves “what do I want to see in a PS5 Pro” as opposed to “why would Sony make one, does it actually make sense, and why did they make a PS4 Pro?” They speculate throwing specs and numbers at it as if they were building a PC, not understanding that Playstations are *not* PCs. They operate under a different set of rules, and you can’t just throw new hardware at them the same way you can a PC. The biggest issue for me when considering a PS5 Pro though, is the pure lack of motivation I see to make one. The pS4 and Xbox Series X existed to support the new 4k standard. The PS5 is already marketed for 4k. They can’t advertise 60 or 120fps, because they can’t make significant changes to the CPU. If a game is already CPU bound at 30 or 60fps, throwing more GPU at it won’t get it to 60 or 120fps. The only thing they could really do is “Well if you’re playing a game that happens to support the option of quality vs performance, you can now play with quality visuals at performance framerates, if either the game’s engine happens to support it or the devs go back and patch it. Remember, despite the PS5 supporting backwards compatibility, and the PS5 having much faster hardware, Playing PS4 games on the PS5 does *not* offer better performance; it emulates a PS4 Pro through hardware. This is because, and I really must stress this, the PS5 is not a PC; the APIs don’t work like DX12 and Vulkan, and are not designed to scale well with a variety of hardware. You can’t just throw specs at it and PS5 go brrrr.

While I don’t believe a PS5 Pro is likely, and I believe they’re certainly working on a significant revision that may even add features, I don’t think a PS5 Pro is impossible. It’s more likely to me however, that if Sony and Microsoft are looking at PC specs with envy, it’s more likely that the next generation may just come a year or two early.