Personal computing discussed

Moderators: Flying Fox, morphine

 
WhatMeWorry
Gerbil First Class
Topic Author
Posts: 100
Joined: Mon Apr 13, 2009 6:01 pm
Location: Illinois

Power/Heat/Performance frequency question?

Thu Jan 27, 2011 12:54 pm

Sorry if this is a obvious question but Is power and heat and performance of a processor a linear or near-linear function of frequency? In other words if a given processor at 3.0 GHz is overclocked to 6.0 Ghz, would the power required, heat generated and performance obtained be approximately doubled as well?

And would underclocking (if possible) behave in a similar fashion?

Thanks.
I dont think, therefore I am not.
 
Althernai
Gerbil
Posts: 88
Joined: Thu Nov 19, 2009 8:02 pm

Re: Power/Heat/Performance frequency question?

Thu Jan 27, 2011 1:06 pm

As a function of frequency, it's linear -- but it also depends quadratically on the voltage. The simplified equation is:

Power = C * Frequency * Voltage^2

The value of C depends on a bunch of different things (e.g. the capacitance of the chip so necessarily the temperature, size, etc.). If you could change the frequency from 3GHz to 6GHz without changing the voltage or anything else, the power usage would indeed go up by a factor of two... but since there are no chips that are capable of this, it usually needs a lot more than that.
 
morphine
Gold subscriber
Gerbilus Supremus
Posts: 11302
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: Power/Heat/Performance frequency question?

Thu Jan 27, 2011 4:16 pm

I have no idea about the theory, but in practice, once you go a bit past the rated clockspeed, the power draw and heat increase a LOT, that is to say, either quadratically or exponentially. It's definitely not linear, as there's a very defined point of diminishing returns.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
Flying Fox
Gerbil God
Posts: 25323
Joined: Mon May 24, 2004 2:19 am
Contact:

Re: Power/Heat/Performance frequency question?

Thu Jan 27, 2011 10:41 pm

Performance is definitely not linear to frequency. There are other parts of the system that does not scale in the same magnitude as the CPU clock (such as memory, disk, etc.) that a program is not going to run twice as fast with twice the frequency.
The Model M is not for the faint of heart. You either like them or hate them.

Gerbils unite! Fold for UnitedGerbilNation, team 2630.
 
WhatMeWorry
Gerbil First Class
Topic Author
Posts: 100
Joined: Mon Apr 13, 2009 6:01 pm
Location: Illinois

Re: Power/Heat/Performance frequency question?

Fri Jan 28, 2011 11:46 am

I came across this from today's short bread link to AnandTech Brazos review. Seems germane to the discussion.

Generally a single microprocessor architecture can cover an order of magnitude of power envelopes. You can take an architecture from 10W - 100W using clock speed, voltage scaling and disabling features (e.g. cutting cache sizes). You can’t efficiently take a 100W architecture and scale it down to 1W. Intel realized this with Atom, and what resulted was a new architecture designed to span the 0.5W - 5W range
I dont think, therefore I am not.

Who is online

Users browsing this forum: No registered users and 1 guest