Testing grease?

I read the thread on "whats your grease" and started to reply but then figured a new thread was more appropriate.
If I wanted to test different pastes, would I run back to back benchmarks changing paste between each run? How many duplicate runs would be enough to get a baseline? Would I go thru all the different pastes one at a time and then do them over in the same order or run multiple benchmarks with the same paste all in a row?

Step 1: Don't. You'll be wasting your time.

There's a very small difference in temperatures between almost all thermal compounds once you get above toothpaste-level compounds. Even those linked charts make the difference looked larger than normal, but those charts start and end near the spread of data so the differences are exaggerated. 20 of the 26 compounds are within 3C of each other at load, pretty much in margin of error territory.

If you do decide to test thermal compound, the general method from what I've seen is

  1. Apply thermal paste (either keep it consistent for each application, or follow the manufacturer directions)
  2. Let it cure for a day or so (computer on with a slight load)
  3. Get the CPU up to 100% load
  4. Take an average of CPU temperature over a period of time
    a. Also take a reading of room temp since 1C higher in room temp will equate to 1C higher in results
  5. Rinse and repeat for each compound.

In theory it would be best to do each test multiple times per application, with multiple applications per compound. That adds up real quick though, because just doing three runs per application and three applications per compound across five compounds would mean doing the exact same thing 45 times. That is a lot of time to spend on thermal compound.


What @m.meri said.

Life is too short to worry about 2°C. I'd rather know which paste lasts the longest, before turning into clay. But again, who has the time, or the controlled environment?

1 Like

Oh god, the memories.

I had an AMD machine a few years ago that had stock thermal paste on the heatsink. After only a year or so it had practically turned into cement. Ended up ripping the CPU out of the locked socket it was stuck so bad.

1 Like