It’s hard to look at the current-generation gaming hardware and not feel like we’re in the quiet before the storm. Because while the PS5 Pro just came out last year, and Microsoft just had a vague tease of its next consoles, it still feels like we’re sitting at the end of a generation. But there’s a light at the end of the tunnel.
We got a chance to sit down with Mark Cerny, lead architect for the PS5 and PS5 Pro, and Jack Huynh, AMD’s SVP for Computing and Graphics, to talk about graphics tech. While they wouldn’t go so far as confirming the PS6 is on the way, they’re definitely thinking about it. Either way, we can’t talk about graphics technology in consoles without acknowledging where things are for graphics cards – because a lot has changed since the PS5 came out back in 2020, thanks to AI.
A Disappointing Graphics Generation?
It’s an interesting time to think about future graphics technology. We’re fresh off of both Nvidia Blackwell and AMD RDNA 4 cards hitting the market earlier this year, and while both of these graphics architectures are very powerful, raw performance improvements have been lackluster. For instance, even though the Nvidia RTX 5090 now consumes up to 580W, up from 450W from the RTX 4090, my testing only showed an improvement of around 10-25% in most games.
That’s definitely better, but not the substantial GPU performance improvement we’ve come to expect every couple of years. AMD fared a little better, with the RX 9070 XT being about 17% faster than the RX 7900 XT, but at a lower price – though that lower price didn’t last long before retailers jacked it up. I’ve talked to both AMD and Nvidia about this, and both manufacturers tell me the same thing: It’s getting harder to shrink the manufacturing process and add more transistors, so software is the way to improve performance.
Both GPU manufacturers are leaning heavily into their software suites. With the RTX 5090, Nvidia launched its DLSS 4 software suite, which was headlined by the controversial Multi-Frame Generation, or MFG. Likewise, AMD debuted the RX 9070 XT with FSR 4, which implemented AI-powered upscaling for the first time in an AMD GPU, along with improved frame generation. Love it or hate it, frame generation is now a standard feature for graphics hardware.
It’s not hard to see the appeal, after all. With the click of a button, you can get your graphics card to introduce AI-generated frames to provide a higher frame rate, and who doesn’t want more frames? Like most things, though, there’s a catch.
Frame Generation essentially works by holding a frame in the render queue while an interpolated frame is inserted between the “real” frames. This can introduce visual artifacts, but the bigger problem is that it necessarily adds latency. These are both huge issues, but they can both be improved by having a higher starting frame rate. As such, I wouldn’t recommend enabling frame generation unless you’re already getting 45-60fps before frame generation.
It’s easy to take that advice and run with it if you’re installing one of these GPUs into a gaming PC, but the next generation of consoles will likely be running on GPUs that will support frame generation, and gamers will likely have much less control over what features they’re going to use.
AI In Gaming Is Only Just Beginning
The PS5 Pro has already broken the seal on implementing AI in console hardware. That mid-generation console introduced PSSR, or Playstation Spectral Super Resolution, Sony’s version of AI-powered upscaling. That was just the beginning, though. Cerny has already come out and said that PSSR was the beginning of what would become Project Amethyst – a partnership with AMD to make game graphics better for everyone, which would of course include Microsoft.
You see, PSSR could only do so much on the limited hardware of the PS5 Pro. Because it needed to maintain compatibility with current-generation PS5 games, the GPU is still largely using RDNA 2 – though certain elements of RDNA 4 were added to assist with AI performance. Still, most of the work of PSSR is still done locally on the shader core rather than being handed off to a Tensor unit or other sort of AI accelerator. That won’t be the case on the next generation of consoles.
According to Cerny, AMD and Sony have co-developed a new algorithm for AI-based upscaling, which is reportedly what the upcoming FSR Redstone, announced at Computex 2025, is at least partially based on. We haven’t seen this new algorithm in person, but it sure sounds like when DLSS 4 changed Nvidia’s AI upscaling algorithm to a Transformer Model, instead of a CNN (Convolutional Neural Network). This new co-developed algorithm is intended to make upscaled games look sharper than they do now, and Team Green's Transformer model is actually pretty good at what it does.
What’s more, AMD’s Jack Huynh let slip that the “machine learning acceleration hardware [we’re] co-engineering on RDNA 5, our next generation, is a direct outcome of the collaboration we’re doing.” We didn’t even know the next AMD GPU generation would be called RDNA 5 until now, let alone that Sony’s collaboration helped to engineer it. Neither Cerny or Huynh would confirm that RDNA 5 “or whatever it will be called” would be what’s actually powering the next-generation PlayStation, but why wouldn’t it at this point?
Even if the PS6 is built on RDNA 4, it’s going to have access to the same AI accelerator cores that allow the RX 9070 XT to use AI upscaling. But Sony isn’t interested in limiting developers to just upscaling. Cerny himself said that “this will support ChatGPT, if that’s what the developers want,” even if that’s not what Sony is actually focusing on. Instead, Sony Interactive Entertainment (SIE) is focused on making graphics technology better, but of course that would include frame generation.
Frame Gen Is Probably Coming to Next-Generation Consoles
Right now, AI is inherently controversial, but when it comes to graphics technology it offers a tempting opportunity for platform makers. After all, it allows drastic improvements to image quality without the raw performance that would otherwise be necessary. Instead, you can just render at a lower resolution, and then use AI to speed up rendering. While AI models are computationally intense, it takes much less work to have your GPU upscale an image to a higher resolution than to render all those pixels the old-fashioned way.
Frame generation takes that concept to another level, adding entirely new frames to the render queue. You can argue that these frames are artificial, and you wouldn’t even be wrong, but that doesn’t mean this tech isn’t coming to the PlayStation 6 – if that is indeed what Sony’s next-generation console is called. But at least it sounds like it won’t be all-encompassing, at least not to start with.
When we asked Cerny about whether frame generation was coming to the next-gen hardware, despite needing a high frame rate to begin with, he told us that ultimately it comes down to player choice. “[SIE] can support a high frame rate by having a lower resolution render and more aggressive super resolution,” he said. “We can also support a high frame rate by using frame generation. And once that choice is out there, Sony can learn more about what gamers want.”
It’s not the best answer, but it does sound like Sony is going to ultimately leave the option of using frame generation up to the developers. “We provide tools for developers,” Cerny said.
Neither of the next-generation consoles have been announced yet – unless you count the Switch 2 – so we know very little about what they’re going to look like, let alone what AI features they’ll support. But if all of the controversial AI features are coming to consoles, then the console makers are going to need to implement some kind of guardrails to ensure a good experience.
Because, while Nvidia and AMD are shoving software features into a high-end graphics card and letting everyone else figure out what they want to enable, consoles by their very nature are much less customizable. If frame generation is going to be a major PS6 feature, it needs to be optional on the user-side. I’m worried that if developers can just enable it on the back-end, we’re going to get a lot of “60fps” games that feel a lot like 30fps when you sit down to play them.
Jackie Thomas is the Hardware and Buying Guides Editor at IGN and the PC components queen. You can follow her @Jackiecobra
Connexion
