A possible way for estimating CPU temperatures

maxSaleen · 02-26-2005, 10:05 PM

I'm just throwing something out that I thought of...

It is understood that different processors and different core architectures within a product line can effect heat output. An example of this would be the P4 2.8C and the P4 2.8E.

Here is my point:

If one were to know the C/W of a cooling loop, and know the heat output of a processor at stock speed, wouldn't it be possible to estimate the heat output at an overclocked speed?

An example of what I mean:

You have a stable room holding at 22 degrees C, a cooling loop with a C/W of .22, and a 2.8 pescott rated for (I'm guessing) 92w. This means that at stock speed, full load, the processor runs at 42.2 degrees C. Now let's take that processor and OC it to 3.6, which we'll say outputs 120w (roughly a 30 percent increase in speed, I'd guess the thermal output increases 30 percent). The processor now runs at 48.4 degrees C. Couldn't you hypothetically scale this in a linear fashion.

Perhaps I'm missing something here.... maybe there's another way to look at it.... I'm curious to hear what you guys think.

Butcher · 02-27-2005, 12:13 AM

Estimating heat output at overclocked speeds has been done for years. Heat output generally scales linearly (or near enough) with frequency and in a square fashion with voltage. That's why overvoltage makes things so hot.

See also: http://www.overclockers.com/articles430/

AngryAlpaca · 02-27-2005, 12:59 AM

I believe that IS the way to estimate CPU temperatures.

pHaestus · 02-27-2005, 01:06 AM

You have one equation: C/W = (Tcpu-Tcoolant)/watts

No biggie to measure Tcoolant whether it's water or air to reasonably accuracy. But you still have two unknowns that are very difficult to accurately measure: Tcpu and Watts. Without some elaborate way to measure one of them then you're screwed.

BillA · 02-27-2005, 01:16 PM

all temps are crap
but Watts are worse
oohhhhmmmmmmmmmm

maxSaleen · 02-27-2005, 02:45 PM

Thanks for the link, Butcher, a very useful little program. pHaestus: I understand the C/W equation, I was just suggesting that one could have an equation for estimating the heat output of a given processor (I mean architecture, not necessarily an individual chip) at a given voltage and at a given speed.

02-26-2005, 10:05 PM	#1
maxSaleen Cooling Savant Join Date: Dec 2004 Location: Orlando, Florida Posts: 383	A possible way for estimating CPU temperatures I'm just throwing something out that I thought of... It is understood that different processors and different core architectures within a product line can effect heat output. An example of this would be the P4 2.8C and the P4 2.8E. Here is my point: If one were to know the C/W of a cooling loop, and know the heat output of a processor at stock speed, wouldn't it be possible to estimate the heat output at an overclocked speed? An example of what I mean: You have a stable room holding at 22 degrees C, a cooling loop with a C/W of .22, and a 2.8 pescott rated for (I'm guessing) 92w. This means that at stock speed, full load, the processor runs at 42.2 degrees C. Now let's take that processor and OC it to 3.6, which we'll say outputs 120w (roughly a 30 percent increase in speed, I'd guess the thermal output increases 30 percent). The processor now runs at 48.4 degrees C. Couldn't you hypothetically scale this in a linear fashion. Perhaps I'm missing something here.... maybe there's another way to look at it.... I'm curious to hear what you guys think.

02-27-2005, 12:13 AM	#2
Butcher Thermophile Join Date: May 2001 Location: UK Posts: 1,064	Estimating heat output at overclocked speeds has been done for years. Heat output generally scales linearly (or near enough) with frequency and in a square fashion with voltage. That's why overvoltage makes things so hot. See also: http://www.overclockers.com/articles430/ __________________ Once upon a time, in a land far far away...

02-27-2005, 12:59 AM	#3
AngryAlpaca Cooling Savant Join Date: Jan 2004 Location: Alberta Posts: 631	I believe that IS the way to estimate CPU temperatures.

02-27-2005, 01:06 AM	#4
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	You have one equation: C/W = (Tcpu-Tcoolant)/watts No biggie to measure Tcoolant whether it's water or air to reasonably accuracy. But you still have two unknowns that are very difficult to accurately measure: Tcpu and Watts. Without some elaborate way to measure one of them then you're screwed.

02-27-2005, 01:16 PM	#5
BillA CoolingWorks Tech Guy Formerly "Unregistered" Join Date: Dec 2000 Location: Posts: 2,371.493,106 Posts: 4,440	all temps are crap but Watts are worse oohhhhmmmmmmmmmm

02-27-2005, 02:45 PM	#6
maxSaleen Cooling Savant Join Date: Dec 2004 Location: Orlando, Florida Posts: 383	Thanks for the link, Butcher, a very useful little program. pHaestus: I understand the C/W equation, I was just suggesting that one could have an equation for estimating the heat output of a given processor (I mean architecture, not necessarily an individual chip) at a given voltage and at a given speed.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)