drkohler said:
Just because he was lucky with some of his past predictions doesn't mean he's right for the future. It is pretty obvious he is wrong with his 2027 prediction. The reasons are pretty obvious: Cerny explained that everything they're talking about is just early simulations. There's a plan, but there is no hardware. There probably are several ideas on how and where to implement the new functions into the gpu hardware and what minimum of memory would be required. There is, very obviously, not a trace of toolbox software to access these future hardware assets for developers. So here are the steps Sony/AMD has to undergo: a) Find the best software algorithms for all the new stuff b) Find out which of those ideas can best be implemented in hardware, fully or partially. c) Find out what the costs are for those ideas (power draw, die space required, memory draw, data contention, etc, etc). d) Make a prototype of the future SoC. e) Test the prototype. If it doesn't have the wished/planned performance gains, go back to b) or shoot yourself. f) Start developing some primitive toolboxes for the developers and send a prototype PS6 to chosen developers. g) Find out if developers can actually do something with the tools. h) If they can't, go back to f) or shoot yourself As you see, there is a lot of shooting involved. This is a radical departure for hardware design so you better make it correct, at best right from the start. All this a) to h) takes time, a lot of time. Really, a lot of time. My best guess would be four years. This is not the usual "Put some more CUs into the SoC to make it faster and add a little more memory" paradigm. This whole "Put more CUs into the socket" has reached the end of the line (unless you want a 800W console). Any performance progress has to come from somewhere else (hint: fake frames, ray tracing in hardware etc). My second guess would be there might be a handheld "PS6 Lite" sometimes in 2027, offering something close to PS5 in performance with maybe some new feature(s) unlocked. Basically buying time for "The Real Thing" sometimes in 2029. What I'm curious about is what actually is in this rumoured next XBox to appear in 2026 (2027?). Obviously, there is nothing in it from those new Sony/AMD ideas as these are just ideas as of now. My guess is it will mostly be an old-fashioned brute force "More CUs more memory" design. Who knows how expensive wafers will be by the time these new consoles hit manufacturing. TSMC has been increasing wafer prices by the week it seems. We are now at $25-27k for 3nm and $30-33k for 2nm (with possibly around 100-120 working SoCs given all the new stuff goes into the SoC - they are not getting smaller) so I can even imagine a switch to Samsung. They are currently around a third cheaper than TSMC and that can translate to an easy $100 savings in single chip costs. |
They're likely well into the process you outline already; companies plan out this stuff years in advance, R&D on the next gen starts practically as soon as the current one is out, and it's 5 years into the PS5 gen now.
Given how long AAA games take to make now, even if PS6 is a 2028 or 2029 release, its games will already be in development, so they will have a pretty clear idea of what the final target hardware will be.
