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: Elmer who wrote (82406)	12/8/1999 7:53:00 PM
From: Ali Chen	Read Replies (1) \| Respond to of 1579162

re < MaxIcc and thermal power> This topic re-emerges every time with every new processor. The Ohm law is valid indeed, even on Intel premises, and always will be. Once again, the power dissipation of a CMOS processor is due to switching gates; there are no resistors. When a gate is either ON or OFF, it consumes almost no power. The dissipation occurs only during switch transients, when a capacitance associated with the gate load (trace etc.) gets charged or discharged via the corresponding gate channel. Therefore the dissipated power is directly proportional to the amount of gates switching per unit of time. The total obviously depends on device clocking frequency, and also on what kind of data and code pattern it is running. The dissipated power is an integral of all these spikes. When the power rails are shunted by capacitors, one can observe only an AVERAGE current, so the dissipated power is pretty much equal to the AVERAGE Icc times AVERAGE Vcc. I found it interesting that Intel specifies MaxIcc at "...maximum signal load" and says very little about their test data/code patterns. There is some indication about their HIPWR30 utility developer.intel.com but they say that it does not produce the maxIcc. They also posted the same current for 600, 650, and 667 Floppermines, which is an apparent nonsense. So I suspect they are lying. AMD is also not clear how they determine MaxIcc. For example, in one document they say: amd.com "To measure output voltage response, run a program that creates a large current draw on the processor. AMD suggests using the popular benchmark program SPECfp under Microsoft Windows NT 4.0." .... "AMD has developed a maximum power utility to assist in designing systems that comply with the processor power and thermal requirements. This utility can verify that the supply voltage remains stable during a transition to a higher power and current consumption level." .... "The maximum power utility is available under a nondisclosure agreement. Contact the AMD sales office for information." Therefore, given incompatible and uncertain measurement conditions, it is impossible to conclude that Floppermine consumes half power as compared to Athlon. I think the real difference could be around 30-40% only, just roughly proportional to more execution units and more buffers in Athlon.