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: Saturn V who wrote (82653)	12/10/1999 10:14:00 PM
From: THE WATSONYOUTH	Respond to of 1572171

Re: "I think that design tweaks on the logic circuitry will also give a significant speed enhancements. So in my mind a 1GHZ Coppermine in Q2 2000 appears probable." You know - based on current and gate capacitance scaling alone, Coppermine should indeed go beyond 1GHz. (I've been told this by people who claim to have looked at it somewhat carefully) However, the design might not scale as ideally as you expect. Based on the IEDM paper,the MINIMUM device is at 1000A physical poly at the bottom of the notch. I think there would have to be 200A per edge physical overlap beyond that to get to the currents specified in the paper. Based on the physical poly of the chip I've seen in cross section, it would have to have been over 925MHz to predict a 1067MHz shippable part (1147MHz with guard band) at a mean of 50A above the minimum physical poly specified in the paper. So, why would Intel ship a 925MHz part as 733MHz. I suspect it was really around 850MHz which would have to ship as 733MHz if the guard band is 7.5% But, then again - I've been known to be somewhat a doubting sort so you need to take that into account. Right or wrong, I will stick to my 1GHz tops prediction. THE WATSONYOUTH

To: Saturn V who wrote (82653)	12/11/1999 4:54:00 PM
From: Ali Chen	Read Replies (1) \| Respond to of 1572171

Saturn5, <I think that you are assuming that the SRAM is running at the same clock speed as the processor clock. Surprisingly (unlike the Celeron) the SRAM on the Coppermine runs at half the processor clock speed. Thus the SRAM will not be the bottleneck for the processor. Thus there may be more headroom than you think.> I think that you have very vague idea what you are talking about. Nobody is saying that SRAM is a bottleneck speed-wise, and the half-speed 32-byte parallel L2 on Floppermines is also well known. From posts of WATSONYOUTH it is quite apparent that he knows what he is talking about, but you are not. It is customary for process "boys" to demonstrate new process capabilities using SRAM. However, the combinatorial logic depth in SRAM decoders under-represent the gate complexity of pipe stages in super-pipelined CPUs, therefore 1.16GHz is not a direct indication of how well the x86 pipeline will fare. <I think that design tweaks on the logic circuitry will also give a significant speed enhancements > The main Pentium-Pro core pipeline has been in design revisions for many years. You seem to think that there are many opportunities left. "Significant".