
The collaboration has already borne fruit; that’s the PlayStation 5 Pro’s PSSR upscaler. That stands for PlayStation Spectral Super Resolution, and it’s the DLSS-like resolution upscaler that’s available on PS5 Pro. Early results with PSSR have been mixed, but when used well it is absolutely capable of excellent image quality results. In that sense, it’s the same as other AI-powered upscalers like NVIDIA’s DLSS or Intel’s XeSS, both of which can produce some gnarly results if used poorly.
Cerny describes Amethyst as being a long-term partnership with two goals in mind. First of all, the two companies want to collaborate on future hardware architectures optimized for machine learning and AI, specifically with an eye toward gaming. It’s not just graphics, either; Cerny says that Sony is aiming for “democratization of AI”, allowing developers to use machine learning for gameplay as well.
Secondly, the Amethyst partnership seeks to train convolutional neural networks for gaming. It can fairly be said that trained neural networks are some of the most important IP components in tech right now, as getting exactly the right mix of training data and training methodology (to avoid overtraining, for example) is very difficult and extremely important. Sony and AMD want to develop and share AIs that can not only perform upscaling, but also frame generation, ray-tracing denoising, and more.
Earlier in the clip, Cerny talks about how the PS5 Pro’s GPU is neither RDNA 2, RDNA 3, nor “roadmap RDNA” (meaning RDNA 4), but rather somewhere in between with a sprinkling of custom hardware on top. Despite that, we can’t help feeling like the PS5 Pro is a snapshot of what AMD is likely to reveal in a couple of weeks at CES. The whole circa-30-minute talk is interesting and worth watching. We’ve embedded it above so that you can check it out for yourself.