XBit Labs has a new article out which firms up a bit some of the good rumors we’ve previously heard about what AMD plans to do.
It gets current verification from sources (probably Taiwanese mobo makers) that AMD still plans on putting out socket 939 processors down to the 3000+ level, which should mean sub-$200 or maybe even sub-$150 prices.
For those who don’t believe anything until AMD says it, there is a kind-of confirmation from AMD in the story saying there will be a “wider” range of socket 939 in the “not-too-distant” future.
All this (along with earlier information) would seem to point to lower-end socket 939 chips becoming available around the end of the year.
90nm or Not?
Many of you out there would really rather wait until reasonably-priced 90nm chips are out there.
You may get your wish sooner than you think.
If AMD has initial problems making very high-speed (i.e. 2.6GHz or more) processors, there’s at least a decent possibility that 90nm chips will show up in the lower speed bins sooner rather than later.
The problem with that scenario is that the factor which will make that more likely is AMD’s relative inability to make high-speed 90nm chips. In other words, the sooner low-end 90nm chips show up, the lousier they are likely to be.
What’s “lousy?” 2.5-2.6GHz with high-end air is probably a good guessimate for the moment. Over the course of time, they’ll improve a bit, but probably not as much as with previous generations of CPUs, maybe another 400-500MHz by the end of 2005/early 2006.
Of course, well before any 90nm 3000+ shows up, we’ll get a real estimate from real higher-end chips, and if they do better, we’ll certainly tell you.
If they don’t, though, we may be faced with a situation in a couple months where you might have to choose between a 130nm chip now, wait three-four extra months for a 90nm chip that will do little better, or wait closer to a year for a chip that really won’t be a whole lot better than that.
This will probably not be a situation where a clear “Now is the time to buy” signal is going to emerge. It will probably be a dribs and drabs situation based on individual situations as overclocking speeds gradually drift upward from 2.5-2.6GHz to around 3GHz during 2005, and x86-64 software becomes more available.
And many with current socket A systems will probably say, “No thanks, not enough,” and that will be an OK answer, too.
What is clear, though, is that if you’ve waited this long, socket 754 is not something you ought to be buying at this point or thereafter. It’s an end-of-the-line platform. The AMD action is going to shift to socket 939, and stay there.
Intel has fundamental problems with Prescott. I don’t think they can be fundamentally fixed, and especially not for the overclockers market, with tunes and tweaks. For the vast majority of heatsink-and-fan folks, they’re just going to be too hot to handle.
That doesn’t mean Prescott can’t be tuned and tweaked enough to slide by in the OEM markets (which after all is what matters to Intel, not us); it probably can. Nor will it bother those with hefty deep-freeze units.
However, that’s not most of you, and I don’t think there’s even the potential for Intel to come up with an acceptable solution for us until sometime next year.
So, for the meantime, for most people, AMD is going to have to be the default choice, if you decide to choose.