On the subject of the 2.0-GHz clock-speed ceiling, in its press release, issued Friday, AMD did say that it "expects its native quad-core processors to scale to higher frequencies in Q407 in both standard and Special Edition [i.e., 125-W server] versions."
Such muted language seems like a tacit admission that AMD realizes people will notice that it hasn't been to hit the high clock speeds which had been expected (See, for example, Charlie Demerjian in The Inquirer.)
On the face of it, the 450-million-transistor Barcelona is an impressive achievement. Each of its four cores has its own, 512-kB L2 cache. The cores also will share a nice, big 2-MB L3 cache. The chip also will support a fast, DDR2/DDR3 memory interface
Even that stuff is just so much specmanship when taken against the real achievements of Barcelona: 128-bit floating-point execution units, an instruction-fetching window that's been widened to 32 bits, as bigger memory pages for improved multimedia calculations.
On the downside, AMD is initially manufacturing Barcelona in 64-nm technology. Intel is already churning out chips at 45-nm. The latter provides a big advantage in terms of holding down power dissipation, although in terms of the specs released thus far, this doesn't appear to be an issue for AMD.
The big eyebrow-raiser in the AMD announcement is that admission that the initial Barcelona parts won't run faster than 2.0-GHz. There's been a bit of buzz, and unhappiness, about this on the AMD Forums. There, users are wondering why the new Barcelonas will have a slower clock speed than Intel's Clovertown. That's the code name for the Xeon 5300 series, which already is shipping in devices running up to 2.66 GHz. (That would be the quad-core Xeon X5355.)
Stepping back and looking at the new Barcelona from a broader perspective, one can assume that any issues AMD is having cranking up the clock speed is part and parcel of the normal growing pains of fielding a new architecture. Unless you believe 10h is fundamentally flawed (i.e, there's a situation such as that which caused Intel to implement its painful right-hand turn in 20004), then what we're talking about here is simply that AMD is going to have to work out the nits as it revs through early steppings of Barcelona.
Which means that the real question is, how will AMD price these things? Can they price them low enough to steal market share from Intel and still make money? Given that, since 2003, AMD has offered great bang for the buck, and also that AMD is under the gun to succeed big-time with Barcelona, this is the big area to watch as the quad-core duel unfolds over the next few months.
P.S. More on AMD's Phenom And Opteron Quad-Core Architectures.