Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Joe NYC who wrote (66413)	12/28/2001 12:36:23 PM
From: pgerassi	Respond to of 275872

Dear Jozef: Although the proportions may be higher, the number of chips is not. Given any distribution, cutting the tail off does not change the curve, it just scales each y point on the curve upwards. It doesn't look any different if you look from the cutoff up. Take your example and assume that 1.5 GHz was the peak and 100MHz is the sigma (standard deviation). 34% can be sold as 1.5GHz, 30% can be sold as 1.6GHz, 20% as 1.7GHz, 10% as 1.8GHz, 4% as 1.9GHz and 2% as 2.0GHz (numbers are approx). 2% of those chips that can get at least 1.5GHz is 1% of all those that run after packaging. Thus, no more chips are present and it still makes the yield (good saleable packaged chips from good packaged chips) 50%. I did not assume this for my calculations for P4 yields (I assumed 66%). This early cutoff would make the overall yield 33% instead of 45% that I calculated using WAG. Thus using my way, Intel needs 20 million dies on wafers to make 9 million CPUs to sell and with your figures 27 million. This just goes to show that as EP, you and I call yield, this overall number. Yousef and the other process people yelled because only the first term is used by them as yield and comparing processes, they are right to do so, since the other two terms come from the design being manufactured and the market out there in combination. As processes mature and the design is tweaked, the peak moves higher in speed and the sigma either grows (when they push the speed) or shrinks as they get better control (when they don't need to push the speed). On the same process, Athlon is just 10-20% faster than P3 and thus, is easier to produce profitably. P4 is about as fast given the IPC disadvantages on normal software but, costs a lot more in yield and output per wafer line. AMD had to go to the next process half step to gain enough advantage to make inroads on Intel's monopoly without the enormous advertising budget. With Intel on 0.13u copper, the P4 and AXP on 0.18u copper are about even technically and from manufacturing. Marketing allows Intel some advantage. With AMD on 0.13u bulk copper, AXP gains back its 20% technical advantage and evens out against Intel marketing. With AMD on 0.13u SOI copper, AMD gets to 40 to 60% technical advantage over P4 and will continue to gain share against Intel. As things stand now, by Q4 with Hammer on 0.13u SOI copper, AMD gains a 100% advantage over Intel and Intel will have to take drastic actions once again. That is where the push to 0.09u SOI copper is required by Intel to get back to even. Intel better hope that either a new CPU line is designed by then, 0.09u is available quicker than planned or that AMD stumbles badly. Because the situation at end of 2002, will have AMD completely break the CPU monopoly once and for all. In any case, consumers are going to be a beneficiary of this competition. Let the games continue! Pete