robaba

Macintosh—>Mac 
it’s an abbreviation, not an acronym 

sorry to be “that guy” but it’s such a simple thing to get right, not doing so is a real sign of disrespect to the company and to the reader.

Intel used to be king of two inter-related fields, micro-processor design and manufacturing process.  Both these leads have evaporated as the industry stumbles around, seemingly incapable of extricating themselves from o long series of critical mistakes.  These have been well documented gambles that never paid off, but management has continued to double down on them regardless of the outcomes.  This is why we have how many derivations to their 14nm process node—yes, the 10nm process exists but has absolutely horrible number of rejected parts making it almost unusable thus the exploding levels of binned chips they offer at 10nm.  Have you noticed that they actually had to BACKPORT Rocketlake due to the failure to execute Icelake?  Likewise the ‘lakes series of designs are going nowhere—their primary focus of development being tweaks to allow higher frequency and over locking at the expense of ridiculous power draw and heat issues and integrated graphics to act as a bulwark against AMD.

is it possible to turn this ship around?  Possible, but not probable.  If it were just an issue of straightening out process it would be one thing, and they are actually starting this with the shift to EUV at huge cost of $$, time, and engineering know how.  The problem is they need to scrap virtually everything they’ve worked on the last 10 years or so...that’s huge and it means that most of their engineering staff also need to re-specialize in a hurry.  But at the same time as this massive switch in production is going on, the design team faces their own quandary—how to maintain backward compatibility while increasing performance without chasing the higher frequency will-o-the-wisps.  They need to find a way to get more operations done per clock-tick rather than squeeze more clock ticks out of the hardware.  If they can’t do this then each successfully smaller node will only give them minimal performance benefit.  They could look at designing specialty cores to run legacy code, but designing gates to redirect code sections might be impossible given how intertwined the legacy code is with modern calls.  They will also need MS to support this at the os level since that’s where so much of the cruft resides.

TL:dr...it can be done, but success is much more likely if they went all in on the number 13 at the roulette wheel.