What is the best CPU cooler of all time? There is no easy answer. If only there was some hierarchy chart to… Oh wait, now there is one! After tons of thought, a bunch of thermal paste, hours of testing, and numerous tweaks, our first CPU cooler hierarchy chart is live. It might be the first one ever.
Measuring CPU cooler temperatures isn’t enough in my opinion. If one unit is five degrees cooler than another, what does that mean? Even though a cooler performs better, overclockers can find the improvements disappointing. There had to be a way to compare CPU coolers beyond raw temperature and noise levels. What we need is a hierarchy chart. My years of building computers and reviewing gave me a good idea of how to determine different overclocking levels. After all the tests, the result is several charts for ultimate CPU cooler comparison.
Environment and Tools
The key to making a CPU cooler hierarchy chart work is consistent variables. While a lot of testing focuses on the system, ambient temperature cannot be overlooked. Being that I have to run tests in my house, I could only come within a certain range of ambient temps using my heating and cooling system. The vast majority of testing fell between 73-75 Fahrenheit. Occasionally, my room temp would jump to 76, or fall to 72, but rarely and only for a short period during a test.
Next up, the testing tools used for the job. Not only did I have my digital thermostat for monitoring the room temp, but I used my old S7 phone for decibel measurement. Unfortunately, my newer Pixel 3 could not consistently return the same dB result. The S7 on the other hand, gave me rock solid measurements. With that variable taken care of, I would use an app to measure an average dB level for 30 seconds. The device was placed at a special spot next to the case every time for consistency as well.
In order for a CPU hierarchy chart to exist, I had to decide on a method to heat things up. I used the program OCCT, which has been my favorite for stress testing for years now. First, I make sure to start each new test right after a fresh boot of the computer. Next, I run a small data set test for 10 minutes to warm the cooler up. After that, I run the test for an additional 20 minutes to pull average clock speeds and temperatures. HWInfo gives a continuous average during the monitoring duration. I add up the 8 cores average frequency from the test, then divide by 8 to get my final number. The average core temp is pulled from the CPU (Tdie) result.
I pay close attention to a couple more factors. I make sure to close all running programs in the taskbar. Even though they should run in the background, temps would drop by as much as a degree if not closed. Also, the BIOS fan control wasn’t precise enough to set fan speeds based on dB level. I ended up using the software program from Gigabyte to set manual fan percentages. Once I set the speed, I could close the program and app center without losing the setting.
There are three critical components to this test: the CPU, the motherboard, and the thermal paste. Unfortunately, I had to use a Ryzen 2700X for the first CPU cooler hierarchy charts here. The problem with Ryzen is that the overclocking range is incredibly small. Even though I found 4 reasonable overclocking levels, the range between them is small for my taste. As far as the motherboard is concerned, the Gigabyte Aorus X470 Gaming 7 was my top choice for this. The power delivery system and cooling are great for these tests. Also, both the 8 and 4-pin power connectors were used.
Thermal Paste is also important. I use Gelid GC Extreme for all tests. Not only does it perform well, but it applies very easily using a simple dot method. I decided to run all testing from inside a case. Since most coolers are probably installed in one of two ways, tower cooler airflow was directed out the back and radiator airflow was directed out the top. While the case I used, the be quiet! Dark Base Pro 900, has fans running at a minimal speed, the side panel was left open. This is to help ensure that hot air doesn’t build up around the coolers. Here’s the full system specs I used.
- AMD Ryzen 2700X CPU
- Gigabyte Aorus X470 Gaming 7 Motherboard
- 2x Gigabyte RX 570 Graphics Cards
- Crucial MX500 SSD
- be quiet! Dark Base Pro 900 rev.2
- NZXT E850 PSU
Testing Results: Stock Frequencies and Temps
I started out with a raw temperature comparison at stock settings. Not only did I see how the coolers compare, but I also got to see the effects of cooling on XFR2. What was interesting was how little better coolers improved frequency across all cores. It shows that Ryzen does a great job at heat management on at an architectural level, performing well even with low end cooling. However, the temperatures drop quite a bit when using a better cooler. Remember, lower CPU temps can still lead to better overall system temps in things like gaming.
Two things to point out here. As far as raw performance is concerned, manual overclocking isn’t worth it on the 2700X. The average frequency is representative of a full load running across all cores. However, single-core boost speeds are still in excess of 4.2 GHz during normal use . Actual performance is likely to be better at stock settings because of boost. This bring us to another interesting point. While certain people will always tweak a component no matter what, manual overclocking could be unnecessary for daily use. Regular users can get the best performance by simply opting for a better cooler.
Testing Results: CPU Cooler Hierarchy Chart
Now it’s time to put these CPU coolers into different levels. I can’t tell you how pleased I was with the results. At level 4, I had to stay below 80c average on the CPU temp to pass. Levels 3-1 needed to stay below 75c average. At first, I was trying to pick temperatures based on representing the coolers properly. However, initial testing showed that average temperatures above those limits would always cause OCCT to end due to an error. It goes to show that Ryzen is very temp sensitive.
As you can see, there isn’t a ton of frequency difference between levels. I’m hoping that future CPUs will be better for showing a bit more range between levels. For now, the 2700X was able to bring solid results in my opinion. As you can see, a slight CPU cooler upgrade will not make a huge difference on overclocking. This is something I discovered through numerous CPU cooler upgrades with my FX CPU. Even though I’d get a slightly better unit, I’d be disappointed that I’d get very little or no frequency increase with the better temperatures. What I learned then is what I want to clearly show now. If you want a noticeably better increase in overclocking, I would recommend a cooler that is at least 2 levels higher than your current one.
Testing Results: Noise Limited CPU Cooler Hierarchy Charts
The second chart represents what I believe to be a tolerable dB level. After playing around with numerous fan settings, I decided that around 46 dB was my tolerable noise limit. Also, this is my personal favorite chart comparison. During PC use, I don’t mind hearing the fans ramp up if I’m doing an intensive CPU task like rendering a video. With gaming though, the lower processor load usually means I can’t hear the CPU cooler. Basically, this is for those who don’t want a jet engine of a CPU cooler in their system.
Finally, I represented what I believed to be a fairly silent cooling level. At this point, I only had one cooler bump out of a level due to lowering the fan speed. However, we do see temperatures raising across the board for the louder units. If you don’t mind using custom fan curves, than any unit can work for your noise preferences. On the other hand, it’s easier to plug the fan into the CPU header and not have to think of it. For those type of builders, you can see certain units performing great and keeping it quiet. (I’m looking at you Noctua!)
The Future of the CPU Cooler Hierarchy Chart
In the end, this was exactly what I wanted to see. The results are not surprising for me, but now I can verify what experience has taught me. With CPU coolers, 360mm and 280mm units are the kings of performance. Larger towers and 240mm CLCs perform slightly below that. While some units are very, VERY quiet, they do suffer in cooling performance. In time, more units added to the charts should shed further light to performance comparisons. My current plan is to continually add to these charts.
I already see two huge improvements I want to change in the near-ish future. First, I hope to get a better CPU for showing overclocking levels. I have a feeling an Intel chip would give me more frequency range. Also, the Ryzen 3000 series release might provide better overclocking and XFR2 range if the rumors hold true. Second, I want to change the dB limit on my silent chart to 40 dB. After testing some near silent coolers, I feel like the noise limit is a tad too high. Curbing the noisy coolers a tad more would help highlight the strengths of silent ones a bit better than it is right now.
Overall, I believe this chart does a great job of helping builders decide on the right CPU cooler. For some, silence will be the largest factor in the decision. Others want raw cooling performance. I’m at the point that I want a cooler to look great first, have close to top performance, but also have a medium noise level. That means 280mm CLCs are a perfect fit for me. While I hope to improve these charts over time, I really feel like this is a great start for the first CPU cooler hierarchy charts.
I hope you find a perfect fit for your CPU cooler needs as well. Let us know in the comments if this helps and what CPU cooler you’d like to see added.