The PS5 Pro will have a new scaling technology called PSSR, which stands for PlayStation Spectral Super Resolution. This technology will mean an important innovation compared to the resizing that Sony used in the PS4 Pro, known as checkerboard, and will also be better than the AMD FSR version that the PS5 used in some games.

This technology has not yet been confirmed by Sony as the PS5 Pro was still officially unveiled, but the video that caused it to leak has been taken down due to Sony’s claims and all the information contained in said video has been verified by Tom Henderson. , who is one of the most reliable insiders in the sector, so despite everything It is very believable.

What is PSSR and how will it work on PS5 Pro

It is an advanced scaling technology that will be supported by artificial intelligence Hardware acceleration for better performance and image quality. Unlike other previous technologies, PSSR should work with spatial and temporal elements, it will have access to game motion vectors, and to improve the final scaling process, AI will be used to combine all available information to create the best image possible.

I’m sure this description will sound familiar to you and it makes sense because in the end The concept is the same which we saw in NVIDIA DLSS and Intel XeSS, two technologies that also perform image reconstruction and scaling process supported by AI accelerated by specialized hardware.

However, in the case of PSSR, there is an important difference, namely that the hardware acceleration would not be done through the specialized cores present in the GPU, but through NPU specific that could be integrated into a dedicated and independent chip. The mentioned NPU would have a performance of up to 300 TOPsand would be key to minimizing the cost of this rescaling at image rendering time.

For comparison purposes, the NPU of the new Ryzen AI 300 APUs has a maximum performance of 50 TOP, which means that the PS5 Pro solution He would be six times stronger. This specialized hardware will completely free the CPU and GPU from the burden that PSSR will carry, but that doesn’t mean the technology won’t cost.

In theory, the process of scaling and reconstruction using PSSR it will only cost 2 milliseconds, the value that applies when scaling from 1080p to 2160p. This is equivalent to NVIDIA DLSS performance mode and involves rendering only half the pixels needed to achieve native resolution. The other half is generated through this intelligent reconstruction and scaling process.

The confluence of spatial, temporal and motion vector elements translates into a large amount of data that PSSR and AI can use to estimate not only pixel position generate but also their movement. To make everything work, a buffer is created with the data and features necessary for the scaling and image reconstruction process, and in this case it would only take a few 250 MB memory.

All this enables their generation in a much more accurate wayit improves the final quality of the image, its sharpness and image stability, because it finally makes the pixels generated by artificial intelligence what they should be and where they should be.

The rendering cost would be very low if these 2 milliseconds were asserted. When the game is rendered frames are generated at a certain rate. By calculating the average that is generated in one second, we can get the universal value that we all know when measuring performance in games, frames per second.

A game that moves at 30 frames per second has a render time per frame 33.33 milliseconds, while a game running at 60 FPS has a render time per frame 16.66 milliseconds. With this clear data, it’s easy to see why I say that the cost of 2 milliseconds in exchange for halving the number of rendered pixels is a very positive number.

In summary, this means that the power increase that occurs when PSSR is activated will not be “clean” and that it will be seen lightly penalized for that rendering time cost. However, as I already told you, this value is very good, and that is something completely normal which happens with all current scaling techniques.

Why will this technology be so important?

This will be very important as it will allow to significantly improve the frame rate in very demanding games, even with applied ray tracing. When you apply scaling, you reduce the number of pixels by a certain amount, which release the GPU a workload that can be overwhelming.

The higher the target resolution and the lower the total number of rendered pixels, the greater the performance improvement. In cases like the one I mentioned above, when scaling from 1080p to 2160p it is easy to achieve a 2-3x performance improvement vs. native resolution.

By this I mean that if the game is running at 30 FPS in native 4K upscaling from 1080p, this performance can be doubled or even tripled, so the average would rise to 60 FPS or even 90 FPS. We’ve already seen this in many games thanks to NVIDIA DLSS, which in configured performance mode would match the PSSR of the previous example.

In terms of performance, the improvement we could achieve is very clear, but what about image quality? This is a great question and another key point of PSSR on PS5 Pro There are many types of image resizing and reconstruction. not everyone uses the same technologies and they don’t have hardware-accelerated AI, so the results they achieve can vary greatly.

PSSR will theoretically be an advanced solution that will use different types of information to generate pixels, combine elements of intermediate images and use hardware-accelerated AI to achieve the best possible result. That is, in theory you should be able to get good image quality despite rendering at 1080p and having to scale to 2160p.

Sharpness will not be the only thing that will determine the final quality of each image. Image and ghost stability They are also two very important issues, although both can only be seen in motion and can depend a lot on each specific game. Those with more small and distant features can be more challenging to maintain good image stability than others that are more closed and have more geometry.

If everything we’ve seen so far is true, then it’s the PS5 Pro’s PSSR technology should be clearly better than AMD FSR 3.1and it could be at a level similar to that of NVIDIA DLSS, as the NPU 300 TOP would be above the 242 TOPs achieved by the tensor cores of the NVIDIA GeForce RTX 4060 and below the 353 TOPs of the GeForce RTX 4060 Ti .

This performance improvement as well as a reduction in rendered pixels will help PS5 Pro with ray tracing, something we already saw when NVIDIA’s DLSS technology debuted was exactly the green giant’s answer to the high cost of ray tracing, and the same will happen with PSSR.

I strongly believe that this scaling technology will be necessary for the PS5 Pro to use ray tracing at a much higher level than what we saw on the PS5, but it will also present a significant challenge in reducing the noise generated by expanding ray tracing and scaling, which NVIDIA has solved with ray reconstruction.

Finally, we also have to consider that PSSR could enable the PS5 Pro improve existing PS5 games in a simple and straightforward wayand its implementation should be very simple. We’ll see what comes of all this when Sony unveils the PS5 Pro, a console that will almost certainly be revealed during this year’s Tokyo Game Show.

AI generated images.