And it's not even unreasonable. We collectively used transistors in general purpose CPUs in a way that wasn't most efficient but made it easy from a software and operational perspective. And that's changing now.
And it's not even unreasonable. We collectively used transistors in general purpose CPUs in a way that wasn't most efficient but made it easy from a software and operational perspective. And that's changing now.
But multicore on a single (mostly) general purpose architecture (leaving aside vector instructions) is still easier from a software perspective than the very heterogeneous architecture--vector, array, FPGA, etc.--we're increasingly adopting.
are you referring to microsoft?
This was of course somewhat self-serving given all the criticism Intel caught for being slow to adopt multicore. But there were certainly concerns about software support at the time, especially on the desktop.
(In the end, it ended up not being as big a problem as some were fearing for various reasons. But it was a real fear at the time.)