Who else can't wait for Intel to crash and burn? 5% improvement per generation is just disgusting, honestly.
Problem is it will only matter at the ultra high end space, and even that is being negated by intel's custom Xeons with on package FPGA's
Once they integrate the FPGA on the same Die as the CPU, they will have much lowered latency and can take advantage of the FPGA further assuming they have a decent FPGA programming library.
Power9 looks legit, but intel isn't out of it.
>>56257854
>on package FPGA's
>grate the FPGA on the same Die
I know some of these words
>>56257854
I don't necessarily mean Power9 specifically, just general trend. Power9 is, at least according to IBM, about 60% faster than Power8 in worst case scenario. That's still more than Intel even if you account for the longer generations. Zen should be 40% IPC uplift over Excavator(which is still pretty good even if the starting point was fairly low). ARM is also improving fairly quickly.
>>56257976
>about 60% faster than Power8 in worst case scenario
in SMT8, and that's just because Power8 SMT8 was garbage and never worked correctly.
Don't get me wrong, it's still hugely impressive from IBM if its correct. Good working SMT8 is fantastic, but it's a bit of a cherrypicked figure to say it's straight up 60% better.
>>56257976
>That's still more than Intel
That's because Intel's core arch has reached the plateau since it's still a souped up P6 on steroids.
The core has nowhere else to go, Intel needs a new design.