Should Xbox End their Generation Early?

[Pemalite]

haxxiy said:

The main problem with an early release is that a generational leap (say, 4090-tier hardware at console prices) isn't there. If you're lucky, a 2025-2026 release will get you a very early and expensive N2 with no backside power delivery, but more likely just N3. The latter is only twice as efficient and dense as the N7 node used in the 9th gen consoles. Not good enough unless you want to rely heavily on FG/ML upscaling.

Mind, the problem doesn't get much better in 2028, but at least by then some of this stuff will get a bit cheaper, and people will be more likely to take what they can get if they've been gaming for 5+ years in the same hardware. Maybe it works if you just want a reboot of the brand to play the same games at higher resolutions and framerates - I don't know, that's sort of what the Pro consoles are for already.

It's not just fabrication process that brings improvements.
It's architecture. It's packaging. It's smart chip design and choices.

At the moment the Xbox Series X has a single chip which is monolithic... But running at a slow speed.
Comprising of 3328:208:80:52 (Shader:TMU:ROP:RT) running at 1825Mhz.

It's performance is roughly equivalent to a Radeon 6750XT... Which is comprised of 2560:160:64:40 functional units. (Shader:TMU:ROP:RT) running at 2321Mhz.
AMD traded die size for clockspeed and came out with the same answer.

The equivalent 7000 series would have offered a good 15-20% performance gain just by going to N5+N6 nodes and changes to architecture.

But it's not just Teraflops, bandwidth, geometry throughput and large chips that will dictate a chips capability going forth, A.I is now a thing, Ray Tracing is a thing, the console with the best Ray Tracing capabilities can have a substantial advantage.

***

What I am going to surmise is that Microsoft opted for a large chip, just to counteract what Sony may opt for... If Sony's hardware specifications ended up exceeding Microsoft's, Microsoft could have increased clockspeeds at the last minute to maintain their hardware advantage.

Just like what Microsoft did with the Xbox One's clockspeed increase at the last minute... Although there was no closing that massive gap.

***

As for what improvements AMD can bring to the table come 2026... We are already getting some rumors and movement on that front with the potential 3D V-Cache, TSMC N2 node getting denser libraries and gate-all-around-transistors ready for 2026, but N3 would be the sensible node to settle on.

Should be an interesting few years at any rate... Especially as we have entered the era of Ray Tracing and A.I, the standard cadence, expectancy and improvements are going to be very different going forth.