Project Amethyst’s Collaboration with AMD Is a Step Change, Though Its Impact is Hard to Measure Until the Developer Actually Implements the Features

Sony Interactive Entertainment and AMD released a major update to their Project Amethyst collaboration (originally launching in December 2024) last week in a video featuring PlayStation Principal Architect Mark Cerny and AMD SVP and GM of the Computing and Graphics Group Jack Huynh. The duo brings three key new features, Neural Arrays, Radiance Cores, and Universal Compression, which will be coming to future AMD GPUs and consoles like the Sony PlayStation 6.

However, the video is only about nine minutes long, and leaves viewers with a lot of questions. Luckily, the guys from Digital Foundry got some answers from Mark Cerny via an email Q&A and posted them in their latest video, where they also discuss the features mentioned above.

First, when asked about changing Sony’s hardware and software strategy from the PlayStation 5 Pro to the next PlayStation console, Mark Cerny said Project Amethyst would bring big changes:

We have shifted our focus substantially. In the past, we mostly created custom technology just for the PlayStation platform. But now with Project Amethyst, we’re putting more focus on co-engineering and co-development with AMD on their hardware and library roadmap. These changes will have a huge impact, because when developers can create their technology with the understanding that it will work across multiple platforms such as desktops, laptops, consoles, etc., there will be a much greater increase in new features. So I’m sure we’ll see a huge impact from the technology announced last week!

That said, the PlayStation Lead Architect also acknowledged that any kind of quantification of the benefits brought by technologies like Universal Compression and Radiance Cores will have to wait until game developers actually implement them on their machines (if it concerns the former) or have access to new hardware (if it concerns the latter).

With this technology, we know that great things will happen but they are difficult (or impossible) to measure. For Universal Compression, I have high hopes for synergies with ML, for example FSR and PSSR are recurrent neural networks that write feature maps to system memory – how well do they compress? Radiance Cores are similar, they are definitely a great way to do processing, but I think we need to get prototypes into the hands of game developers to understand how far they can “level up their engines.

It’s true, in his video announcement, Mark Cerny has emphasized that all of these features are ‘currently only in simulation’, meaning it will be some time before developers can play with them.

PlayStation Principal Architect also shared some other interesting information in an email exchange with Digital Foundry. First, he agrees that Universal Compression is critical as it becomes increasingly complex to achieve significant increases in memory bandwidth over generations. Second, he confirms that Sony (and, by extension, AMD) is very interested in improving machine learning and ray tracing for its next hardware compared to plain rasterization, which will get some improvements ‘where possible’ but is clearly no longer a focus.

This fits perfectly with rumored PlayStation 6 specs that call for ‘only’ a 2.5x to 3x rasterization improvement over the PlayStation 5, while ray tracing will be 6x to 12x faster on the new console (which according to leaker Kepler, Sony plans to launch in 2027).

Follow Wccftech on Google or add us as a preferred source, to get our news coverage and reviews in your feed.