CPU Cooler Test System Update for 2024

[crazyeyesreaper]

As PC technology continues to advance at a rapid pace, the need for new cooling to keep up with industry changes becomes inevitable. In response, we've completely revamped the AMD and Intel CPU cooler test systems. However, our efforts didn't stop there; we've also introduced new testing methods to provide greater insight into CPU cooler performance and value.

Show full review

 

[oxrufiioxo]

Happy to see this and I think the cpu selection is spot on.

 

[Juventas]

Thanks for the new benchmarks. It would be nice to have the air and AIO coolers different colours on the charts for those of us who don't have the models memorized. Also, I was excited to see the Be Quiet Dark Rock on the introduction page chart, but then it was missing from the reviews.

 

[mechtech]

Nice work!!

Since I was asked to share my thought(s)..........here it is.

When looking for a cooler almost every one comes with a fan with different rpm ranges, etc.

Perhaps to try to standardize coolers cooling capability by using the same 2000rpm noctua or whatever fan and use some specific rpms, maybe 650, 900, 1200, 1500, and 2000.

People could then see if it can perform decently with no noise relative to other coolers, and where the cooling starts to drop off from air flow.

and yeah awesome

 

[crazyeyesreaper]

Nice work!!

Since I was asked to share my thought(s)..........here it is.

When looking for a cooler almost every one comes with a fan with different rpm ranges, etc.

Perhaps to try to standardize coolers cooling capability by using the same 2000rpm noctua or whatever fan and use some specific rpms, maybe 650, 900, 1200, 1500, and 2000.

People could then see if it can perform decently with no noise relative to other coolers, and where the cooling starts to drop off from air flow.

and yeah awesome
View attachment 324615

Adding set fans while interesting means little, You recommend a Noctua fan, but whatever fan I pick will not make most happy, On top of that the majority tend to use the fans included. Also some coolers won't accept different fans. Making it a moot point. Interesting to look at but offers little value for the time invested. Which is why we opt for Noise normalized testing as that evens the playing field in a way we can all relate too. AKA a noise level that is acceptable (keep in mind the 45 dBA threshold is on an open bench. A chassis with a side panel can easily drop noise levels 3-5 dBA depending on the materials and setup.

 

[Nostras]

Hmm, the only thing is that 45dBA sounds a tad bit high to me. Typically I want my fans to spin at at most 1000-1100 rpm or so what according to the 50% PWM noise/rpm relationship seems to be about 40dBA or so. I'll just mentally add a couple degrees to the noise normalized chart but perhaps you could poll for a noise normalized noise level?
Maybe combine it with a video so we can make a better judgment on how many dBA's measured reflect our perception of "acceptable".

I do question how many people unironically need the full max RPM chart, if you're planning on using this data in some kind of way you're probably only concerned how much W it can cool and how noisy it is. Or worded differently, if the cooler can adequately cool the CPU when needed if all the reins are loosened. Think you can simplify the testing somewhat at no cost.
Maybe I'm wrong however.

I'm looking forward to the test results

 

[chrcoluk]

The test is good but would have liked to see a page showing details of each product tested.

I also agree that instead of max rpm and noise normalised results, perhaps results in line with silence bios profile, balanced bios profile, high performance bios profile, the most likely modes used by gamers, as I dont know who in their right mind would run their PC daily with fans pegged at 100%.

 

[Chry]

Wow didn't think my Noctua NH U12S is so weak at cooling. Looks to be near/at the bottom of all charts. Well at least it's more than good enough for low-medium range CPUs like mine. And looks like it's the most silent one (can confirm I basically don't hear the cooler).

 

[damric]

I love it. This is the way.

 

[phanbuey]

Need the thermalright peerless assassin in here....

perf/$ king.

 

[Deesider]

Hmm, the only thing is that 45dBA sounds a tad bit high to me. Typically I want my fans to spin at at most 1000-1100 rpm or so what according to the 50% PWM noise/rpm relationship seems to be about 40dBA or so. I'll just mentally add a couple degrees to the noise normalized chart but perhaps you could poll for a noise normalized noise level?
Maybe combine it with a video so we can make a better judgment on how many dBA's measured reflect our perception of "acceptable".

Bear in mind that 45dBA at 15cm is 29dBA at 1 metre away for an on-the-floor setup - or 35dBA at 50cm away for an on-the-desk scenario.

 

[arbiter]

Would loved this a few weeks back when i bought a board 13700k and mem. I have CM hyper 212 which is 3rd on the list for performance/$ so.

 

[Dirt Chip]

NA-U12A still rocking

 

[Scrizz]

NH-D15

 

[Scircura]

Bear in mind that 45dBA at 15cm is 29dBA at 1 metre away for an on-the-floor setup - or 35dBA at 50cm away for an on-the-desk scenario.

It's nearly 2 decades since I've let any of my PC's ever run that loud. I'm certain my current PC never exceeds 25dBA @ 1m except during power-on.

It's not that I've measured them in a lab myself, but I do have many of the same components that Silent PC Review evaluated so I know what certain fans sound like at RPMs that measure 12-25dBA @ 1m. For example, I use the Define R5 case with stock fans which was thoroughly tested and so I can easily generate a reference 25dBA @ 1m by flipping the fan speed switch.

Anyway, 45dBA @ 6" is not relevant to me. I'd personally like 30dBA but I know that's unrealistic for anywhere but a silence-focused publication. However 40dBA seems to me like a reasonable benchmark.

 

[eq-j]

I am writing this text using DeepL.
Sorry if I speak strange English.

I found this test very confusing.
Which is the test result with default clock?
Which is the test result with overclocking?
What is the operating clock when overclocked?

I am not a computer professional.
I can't imagine the actual operating clock even if only TDP is shown.

It is difficult to intuitively understand which CPU cooler to purchase from this result.

 

[Scircura]

I am not a computer professional.
I can't imagine the actual operating clock even if only TDP is shown.

It is difficult to intuitively understand which CPU cooler to purchase from this result.

This is a good point. I think some example TDP's should be given in the introduction of CPU cooler reviews. I must remember, not everyone reads so many TPU processor reviews that they intuitively know each CPU's operating power!

Here are some supplemental charts to hopefully demonstrate why TDP is a good metric, and more broadly useful than a simple overclocked result. In both charts, the processor is at 95ºC at all points, and runs faster or slower depending on the cooler. Today's processors are essentially "always overclocked" and moreover, they continuously adjust their overclock.

Now eq-j has me thinking that "MHz @ noise-normalized" or "Cinebench score @ noise-normalized" might be an interesting measurement too...

 

[W1zzard]

think some example TDP's should be given

Good idea, let me see if i can add a table later today

 

[HTC]

There are those that are MORE interested in lack of noise, and those that are MORE interested in maximum performance, and thus i suggest the introduction of 2 new charts:

- performance per temperature (@ the same temps)
- performance per noise (@ the same noise level)

While both of these are SIMILAR to what you already provide, they are NOT quite the same, and i do understand that IT WOULD INCREASE the work required for any given review.

Just something to consider.

 

[adilazimdegilx]

It's great to see updated test system and methodology for cpu coolers. Love the wattage charts. Very useful and shows strengths of both block/radiators and fans (noise normalized).
Generally, I like it. Fan speed charts are a little bit excessive imo but more information isn't a bad thing I suppose.

Here's some improvement ideas that I thought of after reading the whole thing

- Fan speed/pwm/dB charts can be merged for shorten the page. I don't think we really need a seperate ascending order rpm chart. Just put the rpms in parantheses after dB readings in the same chart. I understand this can be important for some fans those dont scale well with lower pwm signals but I don't think we need seperate charts just for those presumely 'bad' fans.

- Seperation between air coolers and aios is needed in charts (coloring or something else).

- Current list of cpu coolers are lacking at best. I hope you continue to add older coolers to this list and not rely on new reviews to fill the charts. Deepcool as500, ak620, scythes 120mm single and dual towers and other 'cheaper but capable' coolers needs to present just like in old testing setup.
- This isnt a critique particular for the new cooler reviews but in general tpu cpu cooler reviews desperately needs new thermalright products. Their aircoolers regarded as the best by many reviewers. Not just because they are cheap but they are actually better in thermals too (noise normalized or not). Without them this charts can be misleading. Because if you look at the performance per dollar charts, you can end up with a conclusion that aios are just better in general and air coolers doesn't make sense after 60$. However if we had those coolers around 40$ which can basically match or even surpass those big coolers you could draw an exact opposite conclusion.

- 15cm distance to read dB levels is a weird choice. I understand that is because of new test rig's setup and to be able to show the differences in low pwm reading as those results can be under noise floor. However if we wouldn't be able hear the differences between fans on low pwm from a realistic distance then that means they aren't really that different. This is still a valuable information imo. Even if you put your pc on the desk it should still sit about 40cm~ at the closest. I don't think 15cm represent any realistic scenario at all.

 

[pyron83]

I'm pretty sure che Thermalright BA120 would take the performance per dollar crown, at least just looking at his design (6 heatipes instead of the the 4 of the 224XT). I can find it even at slightly lower prices (around 30 € for the ba120, 35 for the 224xt)

 

[Xuper]

by looking at Deepcool Ak400 , Ak620/Ak500/AG500's Performance should be above it.

 

[nguyen]

My ID Cooling SE226XT with Phanteks T30 is laughing at the pathetic AIOs

 

[Carlyle2020hs]

Calling 30 db silent gave me pause.
An then i remember that all of you have lived with AC all your times, yes?
That makes sense, in a sad way...

 

[mechtech]

Adding set fans while interesting means little, You recommend a Noctua fan, but whatever fan I pick will not make most happy, On top of that the majority tend to use the fans included. Also some coolers won't accept different fans. Making it a moot point. Interesting to look at but offers little value for the time invested. Which is why we opt for Noise normalized testing as that evens the playing field in a way we can all relate too. AKA a noise level that is acceptable (keep in mind the 45 dBA threshold is on an open bench. A chassis with a side panel can easily drop noise levels 3-5 dBA depending on the materials and setup.

Thought about it more. I have a noctua cooler with a 1500 rpm fan and in a case it’s still way too loud.

what about a test that’s noise normalized to 35dba on an open bench to see what cools good with low noise?