Nitestorm
28th July 2001, 01:33 PM
A good basic guide to how to take accurate CPU temp measurements if an internal temp diode is not an option:
http://www.swiftnets.com/cpu_temp_testing.htm
Goes into why relying on just the in-socket thermistor to measure cpu temps isn't a good idea if you want accurate results, especially when comparing heatsink performance (certainly useful for hardware reviewers). It covers the AMD spec way of measuring cpu temps via a thermocouple drilled into the base of a heatsink, but also using a thermistor placed next to the core, which is something most users can do, even at home. Plus it points out the dangers doing that can cause and how to correctly place a thermistor next to the core. Plus it includes some suggestions as to why you might be getting poor performance from your heatsink. A combination of using one of those methods (the thermocouple method is prefered) with the internal temp diode is AMDs suggestion for their AthlonMP processor.
It's also important to note that the difference between the temp reported by the in-socket thermistor and the real CPU temp will be different for different heatsinks. Thats due to the way that different heatsinks have different airflow patterns and designs which cool the area the in-socket thermistor measures differently. It's a good idea to see how far a heatsink lets you overclock and compare on that as well as reported temps (just like this review test (http://www.overclockers.com/tips519/)). If a heatsink lets you overclock more, it probably is performing better than temps would suggest. Supposedly the Millennium Glaciator suffers from producing higher MBM temps than what the real CPU temp is by it's unsual airflow design, which is explained by the heatsink's designer in an article (http://www.overclockers.com/tips510/). Hopefully review sites will consider those issues in future when reviewing any heatsink. Also I'd hope they'd consider comparing the Glaciator especially with the SK6 and MC462, but also with quieter fans on those two heatsinks, rather than just the delta fans like most places do. Plus some overclocking tests as well as temp tests and some C/W calculations (made easy with latest version of Radiate) would be good.
Although having said that in-socket thermistors probably aren't accurate enough to use for reviewing heatsinks accurately and shouldn't be trusted without futher tests, they are still OK to use day to day for most users to give you a guide to how well your cooling is performing. If it's very high, it's probably failing even if it might be 5oC out of the real temp. Although for home users, there are a number of ways to improve results with the in-socket thermistor. Maybe add some foam used to stop condensation with peltiers in the middle of the socket, with a hole for the in-socket thermistor, to reduce fan airflow issues. Maybe use a silicone sealant around the bottom inside edge of the socket (and/or outside edge) to achieve the same thing. Also bending the in-socket thermistor up to make sure it makes good contact with the back of the processor and maybe adding a drop of thermal compound to the tip. Although not perfect, such measures are better than nothing.
For more indepth articles on in-socket thermistors and alternative CPU temp measuring methods:
Glaciator, Swiftech 462A, Thermalright SK6 Performance Tests (http://www.overclockers.com/tips519/)
MBM and Heatsinks: A Discussion (http://www.overclockers.com/tips510/)
A Cool Fool (http://www.overclockers.com/tips506/)
Socket A Temps: Accuracy and Comparison Issues (http://www.overclockers.com/articles306/)
Measuring Heatsink Temperatures (http://www.overclockers.com/tips437/)
Heatsink Testing Methodology (http://www.overclockers.com/tips263/)
Heatsink Testing (http://www.overclockers.com/tips240/)
Hot Air (http://www.burningissues.net/hardware/hotair/hotair.htm)
AMD Thermal, Mechanical, and Chassis Cooling Design Guide (http://www.amd.com/products/cpg/athlon/techdocs/pdf/23794.pdf)
Why Many Thermal Measurements Are Not Valid (http://www.arcticsilver.com/measurement1.htm)
Calibrating The Internal Thermal Diode In An Intel Or AMD CPU (http://www.arcticsilver.com/diode_calibration.htm)
Thermal Interface Basics (http://www.arcticsilver.com/thermal_interface_basics.htm)
Arctic Silver Application Instructions (http://www.arcticsilver.com/arctic_silver_instructions.htm)
http://www.swiftnets.com/cpu_temp_testing.htm
Goes into why relying on just the in-socket thermistor to measure cpu temps isn't a good idea if you want accurate results, especially when comparing heatsink performance (certainly useful for hardware reviewers). It covers the AMD spec way of measuring cpu temps via a thermocouple drilled into the base of a heatsink, but also using a thermistor placed next to the core, which is something most users can do, even at home. Plus it points out the dangers doing that can cause and how to correctly place a thermistor next to the core. Plus it includes some suggestions as to why you might be getting poor performance from your heatsink. A combination of using one of those methods (the thermocouple method is prefered) with the internal temp diode is AMDs suggestion for their AthlonMP processor.
It's also important to note that the difference between the temp reported by the in-socket thermistor and the real CPU temp will be different for different heatsinks. Thats due to the way that different heatsinks have different airflow patterns and designs which cool the area the in-socket thermistor measures differently. It's a good idea to see how far a heatsink lets you overclock and compare on that as well as reported temps (just like this review test (http://www.overclockers.com/tips519/)). If a heatsink lets you overclock more, it probably is performing better than temps would suggest. Supposedly the Millennium Glaciator suffers from producing higher MBM temps than what the real CPU temp is by it's unsual airflow design, which is explained by the heatsink's designer in an article (http://www.overclockers.com/tips510/). Hopefully review sites will consider those issues in future when reviewing any heatsink. Also I'd hope they'd consider comparing the Glaciator especially with the SK6 and MC462, but also with quieter fans on those two heatsinks, rather than just the delta fans like most places do. Plus some overclocking tests as well as temp tests and some C/W calculations (made easy with latest version of Radiate) would be good.
Although having said that in-socket thermistors probably aren't accurate enough to use for reviewing heatsinks accurately and shouldn't be trusted without futher tests, they are still OK to use day to day for most users to give you a guide to how well your cooling is performing. If it's very high, it's probably failing even if it might be 5oC out of the real temp. Although for home users, there are a number of ways to improve results with the in-socket thermistor. Maybe add some foam used to stop condensation with peltiers in the middle of the socket, with a hole for the in-socket thermistor, to reduce fan airflow issues. Maybe use a silicone sealant around the bottom inside edge of the socket (and/or outside edge) to achieve the same thing. Also bending the in-socket thermistor up to make sure it makes good contact with the back of the processor and maybe adding a drop of thermal compound to the tip. Although not perfect, such measures are better than nothing.
For more indepth articles on in-socket thermistors and alternative CPU temp measuring methods:
Glaciator, Swiftech 462A, Thermalright SK6 Performance Tests (http://www.overclockers.com/tips519/)
MBM and Heatsinks: A Discussion (http://www.overclockers.com/tips510/)
A Cool Fool (http://www.overclockers.com/tips506/)
Socket A Temps: Accuracy and Comparison Issues (http://www.overclockers.com/articles306/)
Measuring Heatsink Temperatures (http://www.overclockers.com/tips437/)
Heatsink Testing Methodology (http://www.overclockers.com/tips263/)
Heatsink Testing (http://www.overclockers.com/tips240/)
Hot Air (http://www.burningissues.net/hardware/hotair/hotair.htm)
AMD Thermal, Mechanical, and Chassis Cooling Design Guide (http://www.amd.com/products/cpg/athlon/techdocs/pdf/23794.pdf)
Why Many Thermal Measurements Are Not Valid (http://www.arcticsilver.com/measurement1.htm)
Calibrating The Internal Thermal Diode In An Intel Or AMD CPU (http://www.arcticsilver.com/diode_calibration.htm)
Thermal Interface Basics (http://www.arcticsilver.com/thermal_interface_basics.htm)
Arctic Silver Application Instructions (http://www.arcticsilver.com/arctic_silver_instructions.htm)