In a detailed interview coming soon on Digital Foundry, PlayStation 5 lead system architect Mark Cerny has shared more on the Project Amethyst collaboration with AMD, confirming that ML-based frame generation is coming to "PlayStation platforms" but advising us not to expect any further updates this year.
"Just to clarify a few things about the collaboration with AMD, the new PSSR uses the same core co-developed algorithm as FSR Redstone's Upscaling (to avoid confusion, I'll use the new names today rather than FSR4)," Cerny shared. "FSR Frame Generation is also based on co-developed technology (or as my good friend Jack Huynh puts it, 'co-engineered technology'). I’m very happy with how that work is progressing, and an equivalent frame generation library should be seen at some point on PlayStation platforms."
The question is, what PlayStation platforms? We asked how Sony would determine which machine learning-based innovations would come to PlayStation 5 Pro and which may be targeted at next generation hardware instead. Right now, FSR Redstone consists of upscaling, frame generation and ray regeneration components.
"Great questions, particularly considering that FSR Frame Generation is technology that was co-developed between SIE and AMD, we're intimately familiar with it," Cerny revealed. "All I can say is that we have no more releases planned for this year. And that I look forward to discussing this more in the future!"
The full interview goes into depth on the many facets of Sony's new PSSR roll-out, concerning bespoke patching for existing PS5 Pro games vs the system level "Enhance PSSR Image Quality" toggle, whether the new PSSR algorithm requires any further game inputs, and how the performance cost of the revised upscaler compares against the old.
Mark Cerny also responded to recent reports of a Sony patent, which speculated that a prospective "PSSR 2" might employ adaptive AI quality, minimising the need for dynamic resolution scaling in favour of adjusting upscaler precision instead. In DF Direct Weekly #253, we suggested that upscaling cost is already so slight that there'd be minimal improvements to frame-rate by using this technique.
"In February many noticed a patent regarding ML-based upscaling ("varying precision of weights and activations"), which led to dozens of articles, but that concept isn't part of the current PSSR/FSR work," Cerny stated. "We file many patents during the course of business, not all of which relate to released systems or future products."





Comments 5
Very interesting. Exciting that you got another interview with mr. Cerny, looking forward to it.
Looking forward to the whole interview. Cerny is a must watch/read for me his clarity of thought and distilling complex ideas down to their essence makes every talk engrossing.
Adjacently, after the initial fallout of DLSS 5s neural rendering I’ve been thinking about this a lot and I suspect after the initial shock of Gen-AI faces wears off and we actually see creative devs using it I am quite excited for the future of the technology. But as with all things Gen-AI also worried and cautious about the bad sides.
I do wonder if Xbox’s next gen “backwards compatibility” might feature some sort of neural processing and that’s why they’ve gone with such a monstrous APU with an NPU; and if neural processing is one of the reasons they felt confident calling it “the largest technological leap in a generation”. Food for thought.
@themightyant Agree with you. Plus I’ve been thinking the same about Project Helix backwards compatibility, that it uses AI to solve things like fixed resolutions and frame rates from old days and uses an ML upscale plus a frame rate interpolater to make 60fps … or more.
So it’s likely to be a PS6 or PS6 Pro feature
Is it just me or is AI, like 3D printing and NFTs before it, an answer in search of a question?
Like, I’m glad tech bros constantly shove it down our throats.
Show Comments
Leave A Comment
Hold on there, you need to login to post a comment...