InvestorsHub Logo

wbmw

03/02/06 7:29 PM

#70964 RE: j3pflynn #70963

Re: you did say the Turions exceeded their TDP, which was demonstrably false from the very article you used as your source.

The measured power was *above* the rated TDP. The footnote mentioned that VRM loss could account for some of the power, and I acknowledge that, but it still doesn't change my point. AMD's mobile TDP is demonstratably less conservative than their server and desktop TDPs, whether the actual power comes in slightly above, or slightly below the rated TDP. People like mas are nit-picking the details to escape the conclusion.

Re: On the other hand, I did find it interesting that the P-M was doing a possibly better job of staying further within its TDP under load than the Turions.

Exactly. You are the first one to acknowledge this, even though I've mentioned it at least 3 times.

Re: I wouldn't have expected that with Intel's normal TDP criteria. Did they change their TDP system for the P-M line?

No. The common misperception is that Intel's TDPs are designed to be exceeded and protected by throttling, but I think that's more a specific case with Pentium 4, than something universally sanctioned by the company. The mobile chip line has always been more conservative in the TDP measurements. Intel has stated in their datasheets that they pick a TDP arbitrarily by testing real world applications, and they make it high enough so that most if not all real world applications will struggle to reach the TDP on a steady-state basis. Of course, those synthetic apps used to reach loaded conditions are not comprehended in the TDP number, but that's what throttling is for.

AMD's "TDP", at least on their server and desktop parts, is based on the maximum allowable instantaneous current times the nominal voltage of the chip. This accounts for momentary current spikes that are not sustainable over long periods of time. This might be a worthy number for power supply designers, but it doesn't accurately tell you how much the processor will dissipate under real world sustained conditions. Intel's TDP does, and I suspect the AMD mobile TDP is designed this way as well.