CPU Fan Maintenance - Framework 13 Intel Gen 12

Has anyone tried opening up the fan assembly and oiling it? Am using the Intel Gen 12 assembly. Recently removed it and found that there is a lot of dust in there. Fan was almost always on until I disabled S0 sleep. Has always been too hot to handle. Am considering disabling Intel Turbo Boost/Speedstep. Do you have any thermal paste recommendations ( can try replacing the thermal paste)?

Also is there any way to replace the fan with a heatsink?

1 Like

It already has a heatsink, look to the right, and you’ll see the heat pipes coming off the CPU die and going into the fan. A passive heatsink would be far less effective than that fan I assure you. As for thermal pastes, noctua is always a good bet. You might want to consider the honeywell thermal pads as well. Searching the forum should bring up the relevant threads. LTT did a video on it too.


Thanks for your reply. A couple of other questions:

  • How much thermal paste do we need to add? I think it’s possible I added too much on my last attempt.
  • In the heatsink and fan replacement guide, there is a photo that actually shows multiple thermal pads on the replacement heatsink/fan assembly:

    – What is the location of these thermal pads on the mainboard? If I accidentally stacked the extra thermal pads (e.g., because I already replaced the heatsink/fan assembly once), could that adversely affect performance? Also, would I need to remove the thermal pads and apply thermal paste on these additional locations on the mainboard?

Those thermal pads attach to the VRMs and such. Do not replace them with thermal paste. Do not stack multiple thermal pads, it will worsen cooling performance as the thermal pads are designed to transfer heat, not to absorb it themselves. Stacking pads will only impede the transfer. The large square bit on the far left is the main die, that is where the CPU and GPU dies are. A pea sized amount is plenty since the die is bare and fairly small. Again, thin is good, it just needs to fill the microscopic imperfections between the die and the heatsink. I’d compare your temps and scores in Cinebench to one of the benchmarking threads for a sainty check. If temps and scores are within like 2-3 percent of one another. Don’t sweat it, the extra thermal paste won’t hurt anything since it isn’t electrically conductive.

If temps are still too much for you, consider turning off some of the cores in BIOS or limiting power draw using throttlestop or a solution like it. I’m on Linux so I don’t really have much experience with that side of things.

1 Like

Hey @GhostLegion,

Thanks for your advice. I’ve been pretty careful in my thermal paste application this time. I’ve seen some improvements. Unfortunately, am still seeing really bad results when running the benchmark:

Now I know that benchmarks aren’t reflective of everyday usage, but I assume I should be able to get close to average performance (e.g., ~45th to ~55th percentile) on the processor benchmark compared to similar hardware. Instead, as per the above link, am getting 0th percentile results on the CPU benchmark and 8th percentile on the SSD benchmark.

Of course it’s possible that Windows 11 is at fault …

Is there anything else I can try? I can also contact Framework Support.

@Hari_Ravichandran I really would recommend that you stay away from Userbenchmark as a tool. They are notoriously bad. I’d recommend you run Cinebench and compare your score in that to the threads on here.

This looks like a good thread to compare your results to

1 Like

Thanks, I did try the Cinebench as you suggested. Looks like my results are consistent with what’s in the page you linked. I do have my laptop on a vertical stand with the vents exposed so that may contribute to the slightly better score, 8115 vs. 7923.

Thanks for your suggestion, at least I know I’m in line with what’s expected.


In case anyone was interested, I tried repasting with Arctic MX-6. Now I have a multi core Cinebench score of 8723, which is almost a 10% improvement over the benchmark value of 7923. The single score Cinebench score is 1278. Again, the laptop was on the vertical stand prior to running the benchmark.

When I was pasting, I applied the paste, pushed the fan assembly on top, and removed it to check if the entire CPU was covered. Then, I applied additional paste on the uncovered spots on the top of the die and pushed down the fan assembly on the cpu die again. I repeated this process until the entire CPU die was covered with paste. Let me know if this is helpful!

The Userbenchmark performance improved as well: