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: Petz who wrote (25191)	10/21/1997 2:00:00 PM
From: Yousef	Read Replies (2) \| Respond to of 1574848

John, Re: "Give me a quote from a book ..." I have never said that operating voltage is the only determinant of speed ... in fact, please review this post to you where I discuss this matter: siliconinvestor.com My point has been that for a given process generation, the objective is to get as much drive current as possible for a particular operating voltage. The "trick" involves reducing gate thickness and printed gate length, but both of these will force one to use lower voltages. It turns out that a "good rough" estimate of maximum drive current for the .35um and .25um generations is 600ua/um for NFET and 300ua/um for PFET. This can always be lowered by lowering the voltage, but can't be easily exceeded due to gate breakdown, device reliability ... Performance then is well estimated by: Perf ~ Idsat/(CV) ... Idsat=drive current .. V= voltage .. C=capacitance So let's run an example, let's assume that Intel and AMD have optimized their processes to reach the 600ua/um drive current limit for their .35um processes and that capcitances are equivalent (to first order). Now let's compare the PII@300mhz to the K6@233mhz. These are the highest frequencies possible for each device. Let's try to estimate the AMD operating voltage needed to get 233mhz K6's: Perf = 300mhz = 600/(2.5C) ... Voltage = 2.5V for Intel .35um Perf = 233mhz = 600/(VC) ... V = Voltage AMD needs for 233mhz Thus... you can go through the math ... V =(3002.5)/233 ... and lo and behold V = 3.22 Volts This shows that the theory above is backed with "real world" data. So now let's look at AMD's new* .25um process. Based on the report on Tom's WebPage, AMD's .25um process will have K6's running at 2.5V. Are you beginning to see the problem. If AMD optimizes their 2.5V .25um process, then they will have equivalent drive currents AND performance as compared to Intel's .35um process. I would agree that AMD's .25um process will give much denser circuits and thus smaller die size compared to Intel's .35um process. However, perfromance will be equivalent. This is borne out by AMD's statement that K6's in the .25um process will run at 266mhz and 300mhz. I firmly believe that for AMD to reach 350 - 400mhz that they will have to re-design their devices to operate at lower voltages. This will be off in the future !! Hope this helps. Make It So, Yousef