Do You NEED TO Change Thermal Paste After Removing CPU Cooler?
The thermal paste (also called TIM – thermal interface material) can seem scary. After all, the internet is FILLED with scary stories of computers crashing due to insufficient thermal paste, or CPUs being destroyed due to overly high CPU temperatures.
As a result, when you come to remove and reapply your CPU cooler or heatsink for any reason, you might shy away from changing the thermal paste – hoping that your existing paste will be ‘good enough’ (despite removing the heatsink).
But is this true? Well I perform a number of detailed tests in this video to answer this exact question, along with disproving a few myths about CPU TIM.
Finally I include a ‘bonus’ section where I stress tested my 12 core Ryzen 5900X CPU with NO thermal paste at all!
If you prefer text over video, please read on for the guide/transcript version of this video.
Video Transcript And Guide
Intro
Hey everyone, thermal paste is kind of scary. I mean, it IS just some sticky blue gunk BUT people say that your computer might overheat and explode if you use it wrongly. For what it’s worth, that’s actually not true – as a test I removed all my thermal paste and ran my 12 core CPU through a bunch of stress tests without any MAJOR problems… but I’ll discuss that more towards the end of this video. BUT what I was ORIGINALLY going to say is that if you need to remove your CPU cooler (or heatsink) for any reason, you might be a bit scared about the thermal paste.
After all, many people online say that you should ALWAYS wipe off the old TIM with isopropyl alcohol, leave it all to dry, and then apply new thermal paste. But is this true? OR once you’re done swapping out your fans or plugging in some new RGB lights, can you then put your old cooler back on WITHOUT changing the TIM? Well, let’s find out.
How I Tested
I performed three tests here: the first being a “no change” baseline test where I ran a number of ‘stress tests’ on my computer as-is. I THEN opened up my case, removed the CPU cooler, waited a few minutes and put the cooler back on.
I then ran a number of stress tests again, and recorded the CPU temps. For the final test, I reopened my case and removed the cooler again and this time I wiped off all the two-year old Noctua thermal paste with an isopropyl alcohol wipe. I waited for this to dry, and then I applied freshly purchased Arctic MX-4 paste. I put the heatsink back on, and retested everything again. The results were pretty interesting but before getting to that, I wanted to discuss which individual tests I ran – and why.
The first was a simple idle test, where I booted up the computer and waited 5 to 10 minutes for everything to settle down. I waited because sometimes CPU usage and temps will spike up to 60 or 70 degrees celsius during the initial start-up process, so it’s important to let things settle back to normal first. I then recorded the CPU temperatures within HwInfo after waiting for 5 minutes.
There’s naturally MANY different CPU temperature stats I could look at, and there’s really no such thing as a single “CPU temperature” reading because of how modern CPUs are designed. But I cross referenced the HwINfo stats with Ryzen Master and other programs, and the “CPU Die average” was the closest match to the other programs so THIS is the temperature value that I stuck to during this video. To be honest though, as long as you’re consistent with what specific value you look at (and record), your results will be valid.
The next test that I did was gaming – I played Sniper Elite 5 at 4K with all graphical settings cranked up to max. This always heats up my PC so it felt like a good real world test. After this test, I exited the game and then I waited at least 5 minutes for the temperatures to settle back down before running Prime95 where I ran the Small FFTs test. While there are some good alternatives to Prime95 nowadays, this is still a great stress test because CPU usage is always 100% during it. The final test that I did was to render a 4K video at maximum settings within Premiere Pro. This is actually one of the best real world tests you can run, to be honest, because video rendering maxes out your CPU and also stresses some other components in your PC – so your internal temps will get quite toasty.
So that’s the four tests that I ran. I always kept the room temperatures at just below 20 degrees too, to ensure consistency. But that’s enough background – let’s get onto the tests and test results.
Test 1 – Baseline (Noctua Preapplied Paste)
I built my computer at the very end of 2021 and it uses a Noctua cooler with thermal paste pre-applied and THIS was the baseline that I wanted to test things out on. Y’know, can you remove a CPU heatsink two and a half years on and then put it back on without changing the TIM? I actually did need to swap out my case fans so I was pretty interested to test all this out.
So after turning my computer on and letting it settle down, I reset the HwInfo stats and then I waited 5 minutes. You can see the results on-screen now, with average temperatures at idle being 44 degrees celsius.
When gaming they increased to 57 degrees (with a max of 73 degrees) and then they went decently higher in my other two tests, with Prime95 temps averaging at 69 degrees with them then heading higher – to 72 degrees – during a Premiere Pro render. Temps did also spike at 85 degrees during a 4K video render which surprised me a bit – that’s pretty high (to be honest), so it’ll be interesting to dig into this a little bit later on.
Test 2 – Removed CPU (No TIM Change) Then I Retested
The next thing to do, of course, was to create CHAOS and FIRE – at least if you’re listening to some people online. So I opened up my case, and I professionally removed the case fan from my BEAST of a CPU cooler. After using a long screw driver I was finally able to remove the heatsink, and you can see that the TIM is still in fairly good condition. There appears to be decent enough coverage on both the CPU and the heatsink, and it seemed to be fairly “liquidy”. I’m not sure if that’s a word, but basically it wasn’t hard and cracked.
That’s worth knowing because if you were removing a decade old CPU cooler and the thermal paste had dried out, your results may vary a little compared to the results I’m about to show you. Anywhoo, I put the cooler back on as carefully as I could, and then I put everything back together. I booted the PC back up, and again I waited 5 to 10 minutes for everything to settle down before resetting the HwInfo stats and starting my idle test which saw an average of 47 degrees celsius – 3 degrees more than my baseline test.
I’ll compare both results in a minute, but I wanted to quickly keep showing this graph to point out that now both Prime95 and Premiere Pro were averaging above 70 degrees. I saw higher peak temperatures at idle and also when gaming and in Prime95, too.
So here’s the first two test results compared, and for clarity I’m only showing the average temperatures here:
As you can see, thermal temperatures ARE a LITTLE bit higher in this second test. However it’s not massively significant – it’s not like removing the CPU cooler has led to thermal throttling or degraded system performance. My tests ran through fine from a performance standpoint, it’s just that things were a degree or two higher than they were before.
Now you MIGHT be thinking “okay great, I can remove my cooler and put it back on without worrying about the TIM” but that might not be the correct conclusion here, which brings us onto our third main test.
Test 3 – New TIM applied
So I opened up my case and removed the CPU fan and heatsink, and then I started to remove the old Noctua thermal paste. Obviously you have to do this on both the CPU and the cooler base itself. I started off by using a bigger cloth wipe to remove the majority of the paste, and then I switched on over to isopropyl alcohol wipes to remove any of the residue. I lightly covered my CPU and cooler, and left them dry for an hour. This might have been overkill, but I wanted to ensure that they were completely dry. I then put new Arctic thermal paste on it and I usually prefer to put a bit too much on (compared to not having enough paste).
Then I got my Noctua cooler and attached it back on, and naturally I then put my computer back together and booted up, and waited 10 minutes. I then ran through all the usual tests: running at idle, then gaming in Sniper Elite 5, then maxing out the CPU in Prime95 and finally rendering a 4K video in Premiere Pro.
The detailed results are on-screen now and I’ll naturally show a comparison to the other tests in a second, but what’s immediately clear here is that temperatures are lower across the board. Idle temperatures are below 40 degrees celsius, and none of the averages are above 70 degrees – noice. The max temperature spikes were also lower than before too. Here’s the side-by-side comparisons and they’re interesting:
The results are clearly lower after changing the thermal paste, compared to simply removing my cooler and putting it back on. But what’s also interesting is that the new Arctic thermal paste is leading to better results generally than my baseline tests.
I think I know why this is, and I’lll discuss that in a second, but firstly I should point out that eagle-eyed viewers might have seen that my case fans have changed for this third test. This isn’t actually the cause of the lower temperatures though. My previous case fans were high quality fans and they were working fine. I only swapped them out because I wanted a matching color theme on all of my new fans, and I wanted to move to a quieter Arctic set-up with a fan hub. Due to this new fan hub, my new case fans actually run slower than in my old set-up, so if ANYTHING, temperatures in this new test should have been HIGHER than my baseline tests, not lower.
All Results (Table Form)
The raw results in a table format can be seen below:
Thoughts And General Advice
Okay so at first these results surprised me, but I think I know why my temps dropped so much after changing the thermal paste. You might be thinking it’s because the old paste was two and a half year’s old BUT my first ever video on this channel reviewed my Noctua cooler, and in THAT I tested out idle temperatures and also Prime95 temps and they were inline with my baseline test results here. In other words, my temperatures have been at this baseline level for 2 and half years straight.
What I think ACTUALLY happened is that the Noctua cooler is so big that it can be a bit clunky to install – you have to try and align it onto the support bracket screws and tighten them from a distance, which can be hard to do. It’s easy to “scuff” the thermal paste a bit here, because you sometimes need to move the cooler up and down a bit to get onto the screw properly.
Also plugging the CPU fan onto the motherboard is REALLY hard in this case due to having so little space between the case fans and the CPU cooler. I love this old case, but this is one downside for sure. I can remember back to when I originally built this case and I put the cooler on first, and THEN tried plugging the case fan connector in – and I couldn’t quite reach the connector. So I then took the CPU heatsink off, connected up the cooler fan, and naturally then I mounted the heatsink back again. This might have also scuffed (or disrupted) the pre-applied Noctua thermal paste – of course.
So it’s entirely possible that due to some minor build mistakes that I made two and a half years ago, my system has been running 5 degrees hotter (on average) than it otherwise should have been. Whoops. Argh, lesson learnt though. At least now my system is running as cool as it possibly can be, and CPU temps are less than twenty degrees above ambient temps which is pretty good.
Bonus – Running With ZERO Thermal Paste!
Before wrapping up, I have a quick bonus for you. During my tests here, I had the crazy idea of removing the thermal paste COMPLETELY and seeing what happened. That’s right, I ran my 12-core Ryzen CPU with zero TIM. Yeah. The results weren’t great.
I booted up my PC and immediately the fans went crazy – they were completely maxed out – they were making loads of noise. I then got into Windows and launched HwInfo and THIS was genuinely the face that I pulled. My temperatures were spiking all over the place, regularly hitting over 90 degrees celsius max and averaging in the 80s when running my stress tests. Yes, I ran Prime95 on my 12-core CPU with zero thermal paste. #madlad. If you wanted to see a comparison here, THIS is the results:
The short summary here is… don’t run your CPU without any thermal paste! I’ll actually discuss this in detail in another video which will appear up here somewhere when it’s released in a few week’s time, but for now that wraps up THIS video.