Testing TIM

[MadAd]

How would you go about testing whether a compound that claims its 9W/mK really is 9W/mK and not 8 or 7?

I assume its not something that can be done at home?

 

[Matthias99]

I'm unclear how you can claim a Watts/Kelvin or Watts/degree C value for thermal compound, since the temperature rise per Watt is going to vary wildly based on how good the heatsink you are interfacing with is, how good the airflow over the heatsink is, what the ambient temperature is, etc.

The only way I could think of to test reliably would be by comparison testing with something else, changing just the TIM. If you know product A on CPU X with heatsink Y gets up to 55 degrees under load, and product B on CPU X with heatink Y only gets up to 50 degrees C, then product B is doing a better job transferring heat.

If you picked a particular heatsink, and varied the thermal compound, and knew exactly what power dissipation the heat source had, you could get Watt/degree C values for different thermal compounds with that heatsink, in that configuration.

 

[Howard]

I'm guessing you could test the thermal conductivity the same way as any other material.

 

[CycloWizard]

Originally posted by: Matthias99
I'm unclear how you can claim a Watts/Kelvin or Watts/degree C value for thermal compound, since the temperature rise per Watt is going to vary wildly based on how good the heatsink you are interfacing with is, how good the airflow over the heatsink is, what the ambient temperature is, etc.

The only way I could think of to test reliably would be by comparison testing with something else, changing just the TIM. If you know product A on CPU X with heatsink Y gets up to 55 degrees under load, and product B on CPU X with heatink Y only gets up to 50 degrees C, then product B is doing a better job transferring heat.

If you picked a particular heatsink, and varied the thermal compound, and knew exactly what power dissipation the heat source had, you could get Watt/degree C values for different thermal compounds with that heatsink, in that configuration.

You can do it, but it's not so easy. 😛 The way it's typically done is to have two pieces with a well-defined geometry (e.g. two cylinders). You apply the thermal paste to one end of each cylinder and butt them together. Then, by measuring the temperature profile within the cylinders, you can mathematically back out the thermal conductivity. Of course, this parameter is relatively unimportant (as long as it's above some minimum value - fairly high in this case), since the limiting factor will be how quickly you can remove heat from the heat sink.

 

[Mark R]

Originally posted by: Matthias99
I'm unclear how you can claim a Watts/Kelvin or Watts/degree C value for thermal compound, since the temperature rise per Watt is going to vary wildly based on how good the heatsink you are interfacing with is, how good the airflow over the heatsink is, what the ambient temperature is, etc.

That's why heatsink compund's fundamental specification is conductivity, measured in Watts/meter-Kelvin. (W/m.K) 🙂

You'd measure it like you'd measure conductivity for any material. Prepare a sample of known geometry, and measure the thermal resistance across it.