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: EricRR who wrote (115711)	6/13/2000 9:13:00 PM
From: Hans de Vries	Respond to of 1571828

Ratbert, I have a question about your PDF. You hint that Mustang might eliminate a couple of stages of Athlon's pipeline. Why would one want to make the pipeline shorter? Is it to reduce power dissipation? It would improve the performance (with equal frequency) because it reduces "the branch-miss-prediction-penalty" The start of the pipeline loads instructions from the instruction cache. It has to guess which way to go in case of a conditional jump (branch-prediction). Only at the end of the pipeline it becomes clear if the guess was OK. By this time the whole pipeline is filled with "speculative" instructions. All these "speculative" instructions are dumped in case of a branch-miss-prediction. A longer pipe- line has a higher branch-miss-prediction-penalty. The Willamette has a 20 stage pipeline and is therefor very susceptible for miss-predictions. Also I've asked this general question to a number of people, and I'd like to hear your opinion. Do you think that heat production is becoming more important of a limiting factor in raising clock speed? It seems to me, that with such high concentrations of power dissipation, the thermal conductivity of the chip is not high enough to ensure a reasonable operating temperature. Do you think that such problems could be dogging Willy? Willy may well dissipate 100+ Watt or so at 1.5 GHz. Even more would not surprise me. Heat dissipation numbers can however be extracted long before there is first silicon. Intel knew this well in advance and it is unlikely to be a real show-stopper. It's a bit strange that the Computex demo ran at only 800 MHz... This might indicate that there are still some serious issues to be resolved. You can see Willamette's ventilator here. insanehardware.com Regards, Hans.