Can you use bread as thermal paste?

Graphics card surrounded by food

The other day I was talking with a friend of mine and he showed me a picture of his friend's sandwich. It looked awful, and he didn't even bother cutting it with a knife, he just ripped the bread with his bare hands. That's not what interested me though, it was the fact that the cheap bread he used looked like a thermal pad, so I bought some of that bread and decided to see if we could actually use it as thermal paste.


Since this is a serious blog, and I'm a serious person, we're going to do some serious science here.

The bread is the typical thin, soft type used for sandwiches and toasts. Out of the box, it is about 4mm thick, so we need to flatten it and make it more malleable like a thermal pad. In order to do that, I drenched a slice of it in water, put it between 2 sheets of paper, and left it under a pile of books for the night. The next day, I let it dry, and this was the result: a slice of bread about 0.8mm thick, dense and malleable. Flattened bread

Now we need a graphics card to sacrifice in the name of science, and I found this Sapphire Radeon HD5450, a low end GPU from 2010 with 1GB of DDR3 (not GDDR3) memory. The TDP is around 20W, and this model is passively cooled with a tiny aluminium heatsink for optimal performance. HD5450

Since this is a 10 year old GPU, I gave it a good cleaning and removed the original thermal paste. It was so dry, I had to scrape it off with a screwdriver.

The experiment

We will divide our tests in 3 groups:

  • Group 1: things that you would normally use as thermal interface
  • Group 2: alternatives to thermal paste
  • Group 3: extremes

To stress the card, we will use Crysis with VSync off to put a constant 100% load on the GPU. The test system is a 6700k with 32GB DDR4 RAM, vastly overkill for this GPU. The latest driver available for this GPU is version 15.9.1 from 2015, luckily it still works on Windows 10 LTSC. No overclocking was applied.
We'll run Crysis until the card starts reaches equilibrium or starts thermal throttling. When it does, we note down the temperature and how long it took to thermal throttle (if it did) since Crysis was started. After that, we'll quit Crysis and wait for it to cool down. When it reaches equilibrium again, we take that as its idle temperature.

The ambient temperature was 26.5 degrees celsius throughout the tests.

Group 1: regular thermal interfaces

We need a baseline for the temperatures, so we'll try:

  • Noctua NT-H1 thermal paste
  • A cheap thermal pad (1mm)

Group 1 results

To no one's surprise, the NT-H1 is the best of the two, but that thermal pad was really really bad, and the card reached thermal throttling in just 4 minutes.

Here's a picture of that cheap thermal pad: Cheap thermal pad

Group 2: alternatives to thermal paste

This is what you're all here to see. We'll try:

  • Bread
  • Bread + olive oil
  • Nutella
  • Solder flux
  • No thermal paste

Note: all of these materials are not electrically conductive.

Group 2 results

This chart doesn't tell us much other than the fact that Nutella is almost as good as proper NT-H1 thermal paste, so let's see a chart of how long it took to reach thermal throttling.

Time to overheat

Now that's more interesting. Even though all of these except Nutella reached thermal throttling, we can see that having no thermal paste is surprisingly good, as it took 15 minutes of Crysis to overheat. Solder flux performs slightly better, with 19 minutes to overheat, but once it's melted it's the same as having no thermal paste. We can also see that bread is pretty much as good as that cheap thermal pad we tried in the first group of tests, both of which are worse than having no thermal paste at all! Incredible. I was expecting bread + olive oil to perform better though.

Here's some pictures from this set of tests. The bread was ripped by hand to honor our friend who inspired this experiment: Bread + olive oil Nutella Solder flux

Group 3: extreme

For our last group of tests, we wanted to see how the GPU performed in its absolute worst and best case scenarios:

  • No heatsink
  • Liquid metal (Thermal Grizzly Conductonaut)

Group 3 results

Without heatsink, the card reached 115 degrees and caused a thermal shutdown before we could even launch Crysis, so I'll call that a fail.

To test the liquid metal, I sanded off the black coating from the heatsink and applied a very tiny amount of liquid metal on the GPU die. Obviously this bested all other thermal interfaces, but it's expensive, and gallium is corrosive for the aluminium heatsink (we'll check back in a few months to see how bad it is).


From our testing, we have determined that:

  • Using bread as thermal paste is a bad idea...
  • ...but so is using cheap thermal pads from China
  • Nutella is a surprisingly good emergency thermal paste if you happen to run out of paste. It would be interesting to see what happens to it as it dries up over time
  • Playing Crysis at 10-20 fps will give you a headache
  • My PC now smells like a bakery
  • The HD5450 survived all tests and its name is now BreadTX 5450

And if that's not science, I don't know what is.

Stay tuned for the next article, where we cook a steak on a Radeon VII.

Update: December 2020

This is what the liquid metal did to the aluminium heatsink.

Heatsink corrosion No liquid metal left on the GPU

Share this article