Dual core Intel... how hot?

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Tue Aug 8 16:39:25 UTC 2006


On Sat, Aug 05, 2006 at 10:48:20PM -0400, D. Hugh Redelmeier wrote:
> Thanks for all those figures.  I am guessing that they are the
> "thermal design power" ("TDP") figures for each chip.
> 
> One problem with TDP is that AMD and Intel specify them differently.
> In particular, AMD's TDP is the most power that a chip can use when
> the "inputs" are within spec.  Intel's TDP is a nice
> high-but-not-worst-case number; they demand you throttle when the
> temperature gets too high.  So Intel's numbers appear lower.

That is certainly true.  Some people think intel's "typical max" are
nice, while others (like me) prefer AMD's true max.  After all if you
design the cooling system for the typical max, there is a small
possibility that it won't be enough for some loads, in which case the
system will start to throttle itself to keep cool.

> This article is quite  interesting:
>   http://www.silentpcreview.com/article169-page3.html
> I found this from
>   http://en.wikipedia.org/wiki/Thermal_Design_Power
> 
> BTW, speed throttling might not be the most rational approach.  Given
> a fixed amount of computation to perform, full-speed processing
> followed by sleep-when-done may be more energy-efficient than
> throttled-speed processing followed by less sleep.  If sleep can be
> entered and exited quickly enough, this may be generalized to
> interactive computational loads.

It is certainly possible, at least if you have to spin up a disk to
access some of the data, the quicker you can get it all, the sooner you
can spin down the disk again.  In the same way I suspect some laptops
would save power by having more ram, even though ram uses power too,
just because they could reduce disk accesses in some cases.

> | 62W for 90nm Athlon 64 (new revision with virtualization support)
> 
> I didn't realize that this was out already.

It has been for quite a while actually.  Probably the last year or so.

--
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml





More information about the Legacy mailing list