Intel and Apple First to Adopt TSMC's 3nm Technology

[porina]

@LAwLz @leadeater Intel's 14nm evolution is a good example of optimising design choices. If we consider Skylake, Kaby Lake and Coffee Lake, they're all essentially the same microarchitecture but on evolutions of the 14nm process, while they still bothered with +'s. Intel intentionally reduced density to enable the better scaling at higher clocks.

 

Don't take the numbers too literally, but if we assume a 1.4v limited overlock, Skylake was officially 4.2 GHz turbo, would hit around 4.7 or 4.8 GHz OC'd. Kaby Lake was 4.5 GHz turbo, could attain 5.0 GHz OC on a good sample. Coffee Lake initially offered 4.7 GHz turbo with many reporting overclocks well above 5 GHz. Later it was the first Intel CPU to offer 5.0 GHz officially as a turbo, initially as the limited edition 8086k, later as 9900k. On again to Comet Lake as the last iteration of Skylake microarchitecture, turbo beyond 5 GHz is offered.

 

In AMD side, up to early Zen 2 the process didn't really allow higher clocks. For mixed workload stability the practical limit without extreme cooling seemed to be around 4.3 GHz no matter how much voltage you threw at it, although for specific workloads people got better results. I never owned/benched one myself, but I hear that later Zen 2 CPUs (a year or more after launch) did benefit some some improvement somewhere, with many overclockers reporting attaining higher stable clocks than on early samples. Even without a change in process name it doesn't mean the process stands still. If there are ways to get more out, they will find it and make use of it.