3 ms·

Which scenario is better for Intel? - The server and desktop markets move to ARM. The AArch64 instruction set becomes the default that people develop and deplo

Which scenario is better for Intel? - The server and desktop markets move to ARM. The AArch64 instruction set becomes the default that people develop and deploy on. Optimizing for x86_64 CPUs becomes an afterthought. Intel only sells CPUs to organizations that have legacy systems somewhere that cannot be emulated. Intel has to enter the highly commoditized ARM market to stay competitive. - The server and desktop markets largely stay on x86_64 as a result of Intel licensing x86_64 to three other vendors, increasing competition. The market is not commoditized, because Intel can decide who they license x86_86 to and also control extension to the instruction set. They lose market share due to the increased competition. But they can compete on their own architecture that they control. Of course, these are extremes. But given that the current fastest supercomputer uses ARM and Apple showed that ARM is fit for the desktop, I am pretty sure that Intel is (or should be) worried about the first scenario.

klelatti 6y ago
The one thing that Intel has at the moment is x86 and the historic effort that has gone into generating and optimising code for the x86 instruction set. Sharing that with three other vendors (think three other AMDs all competing with Intel) would catastrophically undermine their CPU business. In any event it's too late. They maybe had an chance of fending off Arm a few years ago - by the time any licensing now kicks in Arm will have established itself in the server / PC market. So the only sensible strategy for Intel is to make the most of x86 whilst they can.
bertjk 6y ago
I don't see the point. None of them except possibly Intel themselves would have the economies of scale to put the next M1-challenging super x86_64 iteration on a bleeding edge process node, which would be required if one actually wanted to challenge M1. Also have we forgotten that there actually is a third x86 vendor? Via / Centaur's existence and dormancy suggests that having more x86_64 licensees out floating around won't really make much of a difference.
vimy 6y ago
Afaik there are currently 4 x86 manufacturers. Intel, AMD, Via and I can’t remember the fourth one.

klelatti 6y ago

The one thing that Intel has at the moment is x86 and the historic effort that has gone into generating and optimising code for the x86 instruction set. Sharing that with three other vendors (think three other AMDs all competing with Intel) would catastrophically undermine their CPU business. In any event it's too late. They maybe had an chance of fending off Arm a few years ago - by the time any licensing now kicks in Arm will have established itself in the server / PC market. So the only sensible strategy for Intel is to make the most of x86 whilst they can.

bertjk 6y ago

I don't see the point. None of them except possibly Intel themselves would have the economies of scale to put the next M1-challenging super x86_64 iteration on a bleeding edge process node, which would be required if one actually wanted to challenge M1. Also have we forgotten that there actually is a third x86 vendor? Via / Centaur's existence and dormancy suggests that having more x86_64 licensees out floating around won't really make much of a difference.

vimy 6y ago

Afaik there are currently 4 x86 manufacturers. Intel, AMD, Via and I can’t remember the fourth one.