VR-Zone has a graph showing what AMD and Intel will have available the latter half of 2007 which ought to leave AMDers very, very unhappy.
I suspect the forecast is optimistic on the Intel side, but even if you assume that 3Q 2007 ends up meaning 1Q 2008 for plenty of affordable ones to be around, it still should leave AMDers very, very unhappy.
We’ll ignore the quadcores for today, and just focus on the duallies.
On the Blue Team, we see that the 45nm dually will have 6Mb of cache and run at default speeds of 3.5-4GHz, and do this with just 57 watts.
These numbers aren’t spectacular, but they look at least OK for a process shrink.
On the other side, though, we see a second-generation 65nm AMD dually called Antares, but the core numbers are certainly not out of this world. In fact, they aren’t even as good as the 90nm chips due out shortly.
There won’t be a 45nm chip from AMD until late 2008 at best, so what you’ll get is a 65nm CPU with just 1Mb of L2 for two CPUs (plus 2Mb of L3 tacked on), 2.9GHz at best chewing up 89W of power.
More ominously, if you look at the power ratings for Antares, it sure doesn’t look good once you get past 2.7GHz: 24 extra watts just to get an extra 200MHz.
No doubt there will be some architectural tweaks, but rest assured Intel will have some, too.
I’m sorry, but if these numbers are more-or-less right, and assuming just modest overclockability of a Wolfdale will wipe the floor with an Antares, and Intel’s current performance lead ought to widen.
I’m sorry, but if these numbers are more-or-less right, somebody’s having some serious problems with the SOI sauce. Again. Presumably, these numbers assume no miraculous fix (AKA IBM bailing them out, again) down the road, so they could get better, but if they don’t . . . .
I’m sorry, but if these numbers are more-or-less right, how is AMD going to spin this? “Unlike Intel, we listen to our customers, and they demand slow, hot chips!”?
I’m sorry, but if these numbers are more-or-less right, it looks like AMD will go from selling hypercheap CPUs to hobbyists in 2001/2 to selling hypercheap CPUs to Michael Dell in 2007/8.
I’m sorry, but if these numbers are more-or-less right, this looks like a company that shot its load designing Hammer, and thinks it can rest on its laurels (or, more likely, has no other choice given everything else its doing).