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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Patient Engineer who wrote (26162)	11/25/1997 2:49:00 PM
From: Ali Chen	Read Replies (2) \| Respond to of 1584249

Patient, re <LSI is at least 2 years behind Intel and AMD in process technology.> I do not disagree with Leff and i-line steppers at LSI. I disagree with too sharp labelling of processes as "XX generation behind Intel", "Intel is always the best", etc. etc. For example, you are saying that IBM is on top. However, following Yousef's methods of "analysis", the IBM process must be "3 generations behind Intel" because the Cyrix processors they make are running at 180MHz at most. Here is clearly a discrepance, don't you think? In fact, if we would remember that Cyrix-180 about to outperform the Pentium-266, there must be something else besides the transistor gate length and Idsat. This is called the logic design metodology. More complex logic gives better performance per clock but does not result in high frequency. The reason behind this technology was to keep CPU bandwidth and memory bandwidth in balance (since memory speed progressed far slower than the CPU speeds). Intel has adopted another design rule - so-called "short tick design", which they stole from DEC Alpha, I believe. In this design technique the complex logic must be spread into sequential cycles of simplified logic, but you cannot do much with a few gates in this single logical block. This leads to lower performance per CPU clock but makes the overall design to work at higher speed thus offsetting the low per-clock performance. Now, back to AMD. They seems to be somewhere in between. From the fact that initially K6-233 run at 3.2V, Yousef was fast to conclude that the AMD process is "generations" behind Intel, and "made it so" countless times on this thread. You seem to agree: <AMD is rising fast, but still clearly behind Intel.> I think this is not so clear. For much more complex design (e.g. bigger L1 cache with higher associativity levels requires extra layer of logic to handle addressing in one clock latency), AMD is only 30% below Intel in clock rates. If you would renormalize this rate per a gate, you may arrive to a guite different conclusion. For the above reasons I disagree with simplified labeling of these processes as "generations ahead" or "generations behind" in this complex business. Regards, Ali

To: Patient Engineer who wrote (26162)	11/25/1997 9:15:00 PM
From: Bill Jackson	Read Replies (1) \| Respond to of 1584249

Patient; What is the lead time for the latest steppers for .18 and .11 etc?. I have heard that it can be 18-24 months?, if true this makes catchup hard for even well funded corporations, as the queue might not be jumpable(unless you buy from some smart company that placed an order way ahead, with no desire to use it, but to sell it, as they do in the airline industry). It is my understanding that the leading edge machines have very few manufacturers, and in some cases a sole company makes critical machines? Would you be able to comment and ppace a perspective on this area. Bill