Not well at all. It was fine for PowerPC->x86 because the x86 was the faster processor; an x86 at twice the speed runs PowerPC code written for a processor of half the speed just fine. If you moved from 1.4 GHz G4 single core to 1.83 GHz Core Duo your PowerPC code ran about as fast as before. Moving to ARM would be moving to a much slower processor; your x86 code would run quite badly.
The market for ARM processors is low power, not-too-high performance devices, and it does very well at what it is designed to do. First I think ARM chips are running nowhere near the speed limit for a mostly unchanged device, but increasing the clock speed would mean much higher power consumption (double clock speed = four times the power, roughly). Intel has been working very hard to extract as many instructions per cycle as possible from their chips; that is hard work and uses up immense amount of transistors and therefore power. No doubt ARM could be changed to execute more instructions per cycle, but that would be lots of design work, lots of transistors, lots of power.
On the other hand, ARM processors are tiny and dirt cheap. I think it would be not too difficult to make an eight core design that is quite powerful when needed, with most of the cores powered down most of the time to save battery life.
Last edited:
