Kaby Lake Review

[SteelCrysis]

https://www.extremetech.com/extreme/2419...ts-desktop

Last week, a rumor spread that Intel was working on a new x86 architecture. I have no inside information on whether this is true, but Intel’s CPU performance improvements have been limited to small year-on-year gains since Sandy Bridge launched in 2011. Much of this is due to physics being a great deal less cooperative, and if you compare Intel’s performance in the 15W – 35W space the company has delivered much larger gains. That’s not much comfort to desktop die-hards who remember when you could count on a new CPU delivering 2x the performance of your last CPU within 24-36 months, and it wouldn’t surprise me at this point if Intel was working on a new clean-sheet design.

https://www.extremetech.com/computing/24...ake-family

I’m a bit torn over Kaby Lake. On the one hand, it’s a fine update as far as it goes. There’s not much reason for Skylake owners to upgrade, but review results from the 7700K show that it generally outperforms Devil’s Canyon from 2014, to say nothing of Sandy Bridge or even earlier chips. Squeezing an extra 8-10% out of Skylake in equivalent power envelopes is no small accomplishment given how hard it’s been to move the ball on x86 performance.

At the same time, however, Intel has been working with Sandy Bridge-derived architectures since 2011 and doesn’t have much to show for it — at least, not compared with previous rates of performance improvements. A great deal of this is due to simple physics and the intrinsic difficulty of designing a core that is more efficient, draws the same amount of power (or less, ideally) and provides increased performance without resorting to clock speed gains to deliver it. In a talk several years ago, former Intel Chief Architect Bob Colwell estimated that modern chips are 50-60x more efficient than the original 8086 — but they clock 1,000x higher than that core (4GHz compared with 4MHz). For all the gains we’ve gotten from building better cores, the gains from clock speed are more than an order of magnitude higher.

I mentioned this in the Core i7-7700K review, but it bears repeating: I have absolutely no inside knowledge that Intel is contemplating a new uarch and am not claiming it is. But it wouldn’t surprise me if the company does go this route, especially if Zen proves competitive against the 7th Generation core family. With Apple breathing down its neck with the iPad Pro and its old enemy preparing to return to the arena, it’s a good time to revisit old assumptions and see if new tricks can be found to boost perf, cut power, and deliver a superior product.