hmaly, Re: "The simple truth is that Intel needed to go to .13 in order to stay competitive. AMD didn't need to, and has maintained its market share while still on .18. I would agree that the shift to .13 is tardy, however it hardly has been that detremental, as AMD hasn't lost market share"
You mean that AMD hasn't lost *much* market share? I believe they are a bit lower than their high of 22%, though, which I'm sure a .13u transition would have broken. But, if you want to look at it in another way, then Intel's .13u process certainly has allowed them a competitive edge, while .18u manufacturing would have left them with a larger and hotter, even if they decided to add the extra cache to their older core.
I'll agree that unless Intel has some inventive new ways to reduce power, then AMD seems to have found a way to have a lower power core with smaller die size at the same manufacturing process. And although some people consider it second nature, Intel really doesn't have that many more transistors in their design. It just seems that the custom design has not been very efficient in terms of density. Maybe Intel will change this in the future, since it seems to have given AMD a competitive edge in terms of cost.
And I think that's where the main advantage is. Right now, it seems that AMD's .13u Thoroughbred, and Intel's .13u Northwood may be evenly matched. Hammer will turn the favor back to AMD, but it still remains to be seen how well AMD will do on the marketing side, which by now most people agree is a key ingredient in selling CPUs.
And even if AMD is able to get a far more competitive product that has the marketing to back it up, how long will it take them to increase ASPs, and will the market continue to expand, and allow AMD to grow in volumes? These are the things important to the stock - *not* a few benchmarks on Anandtech.com, even though they may be interesting for the technical speculators to study.
If you are long on AMD stock, you might want to consider how long lived the Hammer processor will really be. It's likely that Hyperthreading in the Prescott CPU will give Intel an enormous marketing advantage, since they will be able to brag two CPUs in one. Will marketing Hyperthreaded performance in single threaded applications be more effective than marketing 64-bit computing without 64-bit applications? Maybe.
There seems to be other things at work, too. Speculation about Prescott New Instructions, larger cache sizes, dual channel DDR memory, and other things that could make a difference to Pentium 4 based systems over the next year or so. Hammer isn't necessarily the end to all of it.
wbmw |
