The main reason is that Microsoft wasn't on track to ship the OS with its grand vision (the "Pillars of Longhorn," which included e.g. WinFS) in any reasonable amount of time.

For example, it wasn't the case that Longhorn was code complete and the Powers that Be said, "You know, the perf isn't where we'd like it, so time to reset." Or that the only work that was happening was perf work.

There were just too many parts to the grand vision of Longhorn, and hence too many moving pieces, and it just never solidified.

IIRC, around the time of the reset, bugs were being opened at a faster rate than they were being closed. So not only was Longhorn not going to be ship-ready in any reasonable amount of time, at that rate it was never going to be ship-ready at all.

At PDC 2003, they also stated they were expecting much faster processors - one slide mentioned 6GHz chips.

Really interesting.

Mid-2003 is when I started at Microsoft. Though: out in MSN, not where the action was. :)

Has Moore's Law really broken down that severely 'lately'? I'm reminded of one of Kurzweil's books that shows 'Moore's Law' holding since before digital computers. But - the book shows - zoomed in, the trend line is actually a succession of periods of fairly rapid growth followed by stagnation. (Obviously though one question is how cherry-picked the underlying data was...)

Edit: The book was The Singularity is Near.

Transistors have kept getting cheaper, but single core perf hasn't been growing as fast for the last decade or so.

I guess one of them was.

But as I said, it was most certainly political as well.

Anyone that went through such big rewrites at corporations similar to Microsoft, knows how much appeasing to different units has more value than technical merits.

Anyway that is just my opinion. For the real issues we would need insider info.