Since Skylake? They are still using the Skylake core in Kaby and Coffee with a few fixes here and there (errata etc). They haven't had a "new" arch since Skylake. Icelake changes that as well as a die shrink for desktop. I have big expectations otherwise i'll be heavily disappointed
I don't know. While I understand the expectation that something big should come from the next architecture, especially considering we've been on Skylake with minor optimizations for while now, I also think it's important to remember how far we've come. From the original 8086 to the 286, 386, and 486 there were many architectural "breakthroughs." 32 bits with the 386, math coprocessors, etc.. Then came the first superscaler x86 processor with the Pentium. Further optimizations occurred through the pentium years in the 90's and clockspeeds soared from 25MHz to 1GHz+ by the end of the classic Pentium era which I consider to the PIII.
The clockspeed wars also brought us the P4 which was a step backward for Intel but it allowed AMD to catch it's breath, and part of the market. Intel stormed back with the next breakthrough, the Core architecture which is with us today. Then there were 4 generations of core, from C2D, 2xxx, 4xxx, and now 6 though 9xxx. In additions even the process shrinks included architectural optimization that were significant with these generations of core. Core count for the desktop went from 2 to 6 and soon to be 8 for these parts, with clockspeeds topping out at nearly 5GHz.
And now after 40 years of architectural breakthroughs, incremental advances, numerous memory subsystem redesigns with the accompanying faster memory, increase in clockspeed and decreases in process size to the limit of current physics, and core count going increasing 600% we EXPECT something big from Intel. Is this realistic?
Would 8 cores on 10nm with 10% IPC improvement over Coffee Lake be BIG? Well that's where subjectivity comes in. Perhaps for some (like me) but not for others.
The low hanging architectural fruit was long ago picked, even the stuff requiring very large ladders, cherry pickers and fire trucks! Process size has shrunk to the point where physics is becoming a serious roadblock as is thermal density. And all the while Intel knows that the future is in mobile low power devices. So how much R&D really need be spent on the desktop except for features which are useful for mobile?
For better or worse we are nearing the end of one era of CPU development. Will we simply stall at this point with minor changes and continue to use CPU's that are "good enough" until modern physics advances? Or will there be a massive shift in how compute is achieved? Or something else entirely unforeseen and not predicted currently?
It's been 20 years since I first loaded up winsock and found this site. It's been a great ride so far and I'm very interested to see what happens over the next few years as process size and architecture start to hit the wall. Intel must look back fondly on the days when a die shrink meant going from 90nm to 65nm, easy peazy, smaller dies, faster switching, less power, more profit... bang there you go! Now we will see if Intel is the 800lb Gorilla they have always been. They pulled the proverbial rabbit out of the hat with C2D right when they needed it. Can they do it again?