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.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Mani1 who wrote (50830)	8/12/2001 7:22:12 PM
From: wanna_bmw	Read Replies (1) \| Respond to of 275872

Mani, Re: "That is not possible, just by how a processor works." If you check out my other post on this topic, I captured some of the wording from the datasheets on the difference between TDPmax, TDPtyp, and those numbers being used for the Pentium III-M and Athlon 4. Intel defines the 34W value as the following. TDPMAX is a specification of the total power dissipation of the processor while executing a worst-case instruction mix under normal operating conditions at nominal voltages. It includes the power dissipated by all of the components within the processor. Not 100% tested. Specified by design/characterization. As you said, it is an experimental quality, and it can't be fully tested because there isn't any software that exists to switch all components at the same time. The other value for the Mobile Pentium III, which is 24.8W, is calculated the same way as in the Pentium III-M and Athlon 4. Check out Intel's and AMD's exact words here. Message 16201061 Both Pentium III-M and Athlon 4 say that they are measuring under "worst case power dissipated by the processor while executing publicly available software," and the Pentium III-M datasheet specifically says, "TDP definition is synonymous with the Thermal Design Power (typical) specification referred to in previous Intel datasheets." Therefore, here is how Andreas' post should have looked. Processor, Clock-Speed [MHz], TDP [W] Athlon 4 \| 900 \| 24 Athlon 4 \| 1000 \| 25 Pentium III-M \| 933 \| 20.1 Pentium III-M \| 1000 \| 20.5 Pentium III-M \| 1133 \| 21.8 Pentium III \| 1000 \| 24.8 wanna_bmw

To: Mani1 who wrote (50830)	8/13/2001 2:11:51 PM
From: Ali Chen	Respond to of 275872

Mani, "That number (34.0W) is experimental. They run software like CAD, photoshop or multimedia and measure the output from the CPU". Mani, the kid who wants BMW certainly knows nothing about TDP, but you are not entirely correct either. The max power is measured, but on a specially-designed software. Intel used to distribute it to valuable OEMs under NDA. AMD has a similar software too. True, it is impossible to make every CPU transistor to switch at core clock rate, but it is possible to design a piece of synthetic software that tries to maximize switching rate, by creating artificial patterns like FFFFFFFF-00000000-FFFFFFFF... , or AAAAAAAA-55555555-AAAAAAAA whenever possible, by simultaneously loading all possible execution units with this, both on addresses and data, with every buffer, TLB and FIFO, all running from L1 cache, and avoiding any external references. One way how it is (usually) written is as follows. They take a former top-notch CPU designer who was promoted to technical marketing, and assigned to write the software using inherent knowledge of internal architecture. Another way to approach the max consumption is to measure instant Icc on variety of power-consuming application codes. This current is usually widely fluctuating (depending on where you manages to measure it - before bypass capacitors, or directly at the CPU core). Then you capture an instance of CPU code that causes those high peaks in the current, and wraps this code in an infinite loop, and maximize results by little code/pattern tweaking. In both cases there is no guarantee that you hit the absolute power case, that's why all those marketing excuses and uncertainties. In both cases the result is about 40% above any worst-consuming real apps, as far as I know. You can get good results with some freely-available software like BURNP6. Regards, - Ali