Madshrimps Forum Madness

Madshrimps Forum Madness (
-   General Madness - System Building Advice (
-   -   Myth of CPU temperature (

Sidney 24th December 2004 17:45

Myth of CPU temperature
The Myth of CPU temperature

Billon dollar industry and hundreds of products have been created in dealing with CPU temperature. Most of us know high CPU temperature affects PC performance. Both AMD and Intel list critical operating temperatures for specific processors in their product line-ups. To simplify some of the terminology used, the following lists some of the commonly used terms in the illustrations.

Motherboard temp sensors = varies way in measuring CPU, system and Mosfet temps. Some motherboard makers use “calculation” by coding algorithms in BIOS, others may use surface mounted diode sensors. No matter which method they choose, the location of the temp measurement is not the same as what and how the processor makers have “prescribed” in obtaining TCase temp as stated below. In fact, the only way to measure processor temp accurately and resembling TCase temp listed by processor maker is to drill a hole on the heatsink where a temp probe can be placed as close to the center of the heat spreader as possible. To make the matter worst, diode mounting and “calculated” temp could be so much off (higher and lower than actual), they drive many people crazy in finding the “real” processor temps, hence creating this Billon $ heatsink / water cooling industry.

TCase = temperature at center of Heat spreader for P4 and A64; or center of surface of core for AMD XP series processors. Both Intel and AMD suggest TCase temperature reading for max allowable operating temp for their processors.

System Temp = close proximity of air temp representing air intake to the heatsink. Although system temp has direct effect on CPU temp, it also affects other components such as hard drives, graphic card, memory, Mosfets, and north-bridge temps.

Ambient Temp = room temp where the system is placed.

Processor Thermal output = max thermal loss of a specific processor given by manufacturer.

Thermal Interface Material = material used to better conduct heat from one surface to another, normally expressed in C/W coefficient. No matter what we use some loss is expected.

C/W coefficient factor = characteristic of thermal transfer, the lower the number the more efficient. C/W rating could be found in product specifications from name brand manufacturers.

In a nut shell, some may argue open bench testing will have lower CPU temp than a cased system. This is mostly true to some extend. The reason: the CPU fan intake temp is almost identical to air temp in the room which can easily be assumed lower than the temperature inside an enclosure (chassis or case). However, open bench testing reduces the ability of air movement around the heatsink without the benefit of exhaust fan, unless a fan is placed around the heatsink to create air movement necessary to get rid of hot spot that could impact the intake air as hot air rises. The same principle applies to memory, hard drive and graphic card temperature in open bench testing.

While system temp is important in the effect of CPU temp, it is less important when a fan duct or water cooling is used. By measuring the intake temp to the CPU fan; and water temp inlet will provide a better picture of cause and effect of CPU temp relative to heatsink/fan and “water-cooling-system” efficiency.

While a high performance heatink/fan and water cooling system will provide low CPU temp, controlling or providing a temperature as close to ambient temp is the key.; unless sub-ambient temp such as Vapor Chill is used.

Placing an accurate temp probe as close to the processor heat spreader or core as possible and add 6-8 C to the reading may also provide a temp reading equivalent to TCase temp specified by processor makers.


Using Swiftech MCX478-V and P4 3.0C

We can calculate max CPU temp based on 25C room temp assuming acceptable air movement around the setup.

1) Using data from Swiftech = Delta 80x25 fan @ 0.226C/W
2) Using 81.9 Watts from Intel S-Spec
3) Given 25C room temp
4) Add 0.09 Thermal paste C/W

Total C/W from Swiftech setup = 0.226+0.09 = 0.316

81.9 Watts x 0.316 + 25 = 50.88 C

This is the temp you are expecting and no less; as the system is placed in a case, you will not get another better than this.

Granted the 81.9 watts is norm meaning you could get a processor with higher or lower thermal loss. You now have a theoretical temp; try using an external probe reading and see how close you are. Don’t forget to add 5-8 C to the probe reading.

If your motherboard sensor is giving you 40C temp reading don't go over joy; on the other hand if it reports 65C, don't be sad about it.

jmke 25th December 2004 12:51

interesting read, on a Christmas day :)

Sidney 25th December 2004 16:38

Nothing much to do before I headed off to my brother's house for X'Mas dinner.:)

All times are GMT +1. The time now is 16:42.

Powered by vBulletin® - Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO