I found this story at Tom's Hardware about the AMD Black Edition 5000+ quite fascinating. I wish AMD would put out more stuff like this, get people excited about their CPUs.

http://www.tomshardware.com/2007/10/22/budget_overclocker/

I agree with Sargent Duck who suggested that Intel is probably not all that worried about AMD's Barcelona et al, so no reason to rush Nehalem.

For all but the embarrassingly parallel workloads, I'm going to wager that SMT won't make the slightest bit of difference. Here's why:

Take a quad-core Nehalem, and add SMT. You now have 8 front ends. However, can you split your tasks 8 ways? Will they find a way in the next year or so? Doubtful.

One thing is for certain, if you can (Servers and graphics processing), that core's gonna need a helluva lotta bandwidth.

I'm only guessing here, but judging from some Initial Barcelona benchmarks, Intel probably isn't feeling all that worried. They probably pushed it back to make sure they got everything right, maybe go through a few steppings and such.

That, and to milk Penyrn a little more.

I won't believe that AMD can ship 45 nm in volume in the first half of next year until I see it. They still haven't figured out how to ship Barcelona in volume on 65nm, and it's almost November.

Agreeing with #3