Vince Ref < ..point I think we will have seen all the trump that Intel has to play until late next year. .. > and AMD as an investment.
If you base your Intel-AMD decisions based upon the trumps or cards you can see, you will be in for extreme disappointments. The two companies have diametrically opposing styles when it comes to showing their cards.
AMD is forced to show its trumps early, particularly for the K-7. It needed to have third parties develop, and manufacture mother boards and chipsets, BIOS etc. The only way you will get their attention is to loudly blow your horn, and hype your product. And the AMD management is outstanding at over hyping. Worse AMD misses its public commitments, because it is under resourced and,and has poor implementation strategies.
Intel is not in the same position and prefers to play its cards very close to its chest.It is forced to show its cards to get third party software support, as needed for MMX or Katmai SSE software.Again Merced has been hyped to get other vendors to write Operating system and applications software.Otherwise Intel will stay mum, except to publish a roadmap for PR reasons, and to keep its customers comfortable.
Andy Grove of Intel is known for his paronia, and galvanized Intel into a frenzy to protect the low end, because he is afraid of a day when the low end becomes "good enough for everyone", and wipes out the high end.[Historically Intel ignored the low end, because of the slim margins. ] So Intel's Celeron counter attack has decimated the low end competitors. The K-7 targets Intel's Fort Knox, the bulk of its profits. Andy Grove has had plenty of warning to be ready for the K-7. Knowing what is at stake, he will surely have several defenses in depth.
I can only speculate to what the defenses might be:
A. Pulling in Willamete.
B. Having a 0.13micron Coppermine. The 0.13 micron may not be that far away for Intel.
C. A hybrid 0.13/0.18 micron process. A 0.13 micron stepper can be used for the poly layer, and the other layers can still be 0.18. A 0.13 micron poly with thinner gates and new implants will definitely take Coppermine well past a GHz. The 0.13 micron poly can also be achieved with 0.18 lithography. Simple tricks with the photoresist exposure and developer, or poly etching can also pattern 0.13 poly.
Andy will definitely do whatever it takes to protect his Fort Knox.
The deck is stacked against AMD with very high risk issues:
A. The 0.25 micron is being ramped down while 0.18micron is ramped up. This is a very high risk. New processes typically have hiccups. A hicuup on 0,18, and AMD loses all revenue.[ Intel ramps up a new process in a new fab, while the old process still continues to run. So a possible hiccup on the new process is not catastrophic.]
B. Dresden starts on a new Copper process developed by Motorola. The Copper technology is very immature.[Historically Copper has been the enemy of silicon processing.] Motorola is having enormous trouble shipping Apple processors made on the Copper process. It is not clear what the problem is, but this is a definite red flag.
AMD cannot afford a hiccup in either Fab 25 or Dresden. Its financial situation is so precarious, that one misstep might cause insolvency. The only bright star is Flash,which will have to be sold if the current cash drain continues.
So an AMD investment is very risky today. |
