Although the Xbox Series X machine learning hardware is not on the same level as the PlayStation 5 Pro’s, it has some use cases, but Microsoft never showed any interest in using it.
Speaking during the latest episode of their Weekly podcast, the tech experts at Digital Foundry addressed a point of discussion that’s been circulating on social media in the past few days, suggesting that Sony, with the PlayStation 5 Pro, has finally caught up with Microsoft regarding machine learning, saying that having machine learning hardware in the console, and not doing anything with it doesn’t exactly mean catching up. The around 100 TOPS of the Xbox Series X are in reference to INT4, which is usually not used for machine learning upscaling. Still, there’s nothing preventing Microsoft from developing a machine learning model that could run on the Xbox Series X and Xbox Series S, but the company just doesn’t seem interested in doing so. The Series consoles’ machine learning hardware could have been used, for example, to run Auto SR for Xbox 360 backward-compatible games, as the company developed a model that can operate on 40 INT8 TOPS MPUs, something that would have been easily manageable by the current generation consoles’ machine learning hardware.
Given that the Xbox Series X|S architecture was developed back in 2016 when machine learning was on the horizon, it’s not surprising that it wasn’t used for anything. Things, however, are definitely set to change in the future console generation, especially after seeing the excellent results the first iteration of the PlayStation Spectral Super Resolution is capable of achieving.
