We ought to see reviews of Bloomfield Nehalems show up on Monday, since the NDA is supposed to expire then.
However, there are already signs that at least the first generation of Nehalems are likely to be underachievers in the overclocking arena.
The primary reason for this appears to be that QPI seems to be a bottleneck for overclocking . It seems to limit an OC of the 2.66GHz Bloomfield to roughly around 4GHz, or a bit less than that. If you want more than that, you need a Bloomfield with a higher multiplier, which means you’ll pay at least twice as much for that CPU, roughly $600 rather than $300, and someone who ought to know recently said that the best overclocks will come from Extreme Editions.
I had wondered why whatever OCing results we had seen was being done with the 2.93GHz Bloomer. Now we know why.
Combine that with preliminary reports than Bloomfields tend to be on the toasty side, and I don’t see a winner here. Sure, it will win benchmarks, but it’s not going to attract too many benchmarkers at these prices.
And yes, Intel can charge $500, $5,000 or $5,000,000 for a CPU if it feels like it, and it will sell enough of these in workstation and servers and the like without the need for a single OC sale.
But this is what a luxury CPU looks like, and this is what we get when AMD is down on its luck.
If I were anybody looking to buy a highish-end system soon, I’d look really, really, really closely at Monday’s benchmarks comparing these Bloomfields to any Penryns, and unless there’s a huge difference between the two in what you mostly do, you might want to tame the ego a bit before paying an arm and a leg for an extra 10% or so.
We don’t need to feed the monster.