Report: AMD Built Navi for Sony PS5, Delayed Vega to Do It

This site may earn affiliate commissions from the links on this page. Terms of use.

AMD’s GPU roadmap has been substantially aligned to meet semi-custom needs ever since it won the PS4 and Xbox One contracts. In a piece earlier today, we actually explored some of those links and how they may have impacted AMD’s overall roadmap in the past five years. But the links may actually go deeper than we thought. There are reports that Navi isn’t just aligned to Sony’s roadmap — it’s a chip that AMD built explicitly for Sony, even when doing so pulled resources away from Vega.

This information, reported by Jason Evangelho at Forbes, would explain a few things. When AMD first talked about Vega at its Sonoma event in 2015, it stated Polaris would be a major GCN revamp, while Vega would be, if not a complete break with GCN, a full-scale, ground-up rearchitecting. While it’s true that Vega makes some significant changes to GCN and was obviously rebuilt to focus on hitting higher clock speeds, clock-for-clock, Vega isn’t much more efficient than GCN was in shipping titles.

Comparisons found gains of 1-5 percent for the newer architecture. That was much smaller than the gains we had expected, and it played a part in why Vega hasn’t been as effective a competitor against Nvidia’s GTX 1070 and 1080 GPUs as AMD might have liked. It also impacts the overall temperature and thermals of the AMD chips — remember, clock speed and power consumption don’t scale linearly, which is why the Radeon R9 Nano was vastly more power efficient than the Fury X while still being roughly 90 percent as fast. If Vega had gained more performance relative to Fury, AMD wouldn’t have had to run the chip’s clock as high, which directly translates to cooler, quieter cards.

AMD-GPU-Roadmap-2016

AMD’s original roadmap for Polaris, Vega, and Navi. The “Nexgen” memory is simply a typo and should read “Next gen.”

AMD’s original plan, again as communicated in Sonoma in 2015, was to launch a midrange Polaris in mid-2016, a “Fat” Polaris high-end GPU at the end of 2016, and Vega in mid-2017. Then, a few months later, the company changed this. Fat Polaris was canceled, with Vega pulled in to replace it. But this wound up not happening. Vega launched in Q3 2017 (more or less the same time period AMD had initially defined) and Fat Polaris never happened at all. This could be explained if AMD pulled resources away from Vega development to put towards other projects.

We’ve also heard rumors that AMD may have pulled resources from Vega to finish Ryzen, which makes a certain amount of sense given the critical importance of that part to AMD’s overall recovery. But pulling resources from Vega to finish Navi is also interesting, given the relative importance of Sony as an AMD customer, and the need to align roadmaps to make certain AMD would have a new part in the right position to meet Sony’s needs. The further implication is that Sony’s PS5 will target midrange PC market performance for its 2020 console, though we can also assume that the game system will pack more firepower overall than the PS4 Pro.

Again, is it bad that AMD aligns its midrange launches with Sony’s long-term needs? Not necessarily. In fact, given how the console and PC GPU markets have grown together, it’s probably one of the most cost-effective ways for AMD to meet the needs of various markets and, as we’ve stated, AMD remains focused on and concerned overall with the PC space. But the implication here is that it’ll be a little while longer before AMD separates its efforts to build midrange chips for the PC market and the cores it builds for Sony or Microsoft. And speaking of Microsoft, we still don’t know how the Xbox Next (the safest name for the thing, given the company’s nonstandard naming policy) will play into all this. Is it also based on AMD hardware, or is Microsoft going to take a different route?

Let’s block ads! (Why?)

ExtremeTechGaming – ExtremeTech

Related posts: