[SOLVED] What do I need to pay attention to when comparing CPU air coolers?

[theamazingmettaton]

I'm currently trying to overclock my i5 9600k but need a better cooler. I'd like to stick to air coolers. What specs should I be comparing between air coolers and how to know if its up for the task?

I know all the basics of how an air cooler works, fins, heat pipes, heat sink and fans but I don't know how to properly compare them to others. I've heard about TDP but I've read various forums about whether or not that's reliable so?

 

[Flayed]

The most important spec is cooler height because it must fit inside your case. You need to check your cases CPU clearance and make sure it is greater than the cooler height. Then I guess fan size, number of heat pipes but better to look at reviews and see the rise over ambient temperature for a given load. Or just buy a Noctua D15 or D15S

 

[iPeekYou]

I'm currently trying to overclock my i5 9600k but need a better cooler. I'd like to stick to air coolers. What specs should I be comparing between air coolers and how to know if its up for the task?

I know all the basics of how an air cooler works, fins, heat pipes, heat sink and fans but I don't know how to properly compare them to others. I've heard about TDP but I've read various forums about whether or not that's reliable so?

As @Flayed mentioned, size. No good a big fat heatsink is when you can't physically fit it inside your case. Height is a common concern, but length can be too, with small mATX cases and big coolers such as Assassin III and its ilk would eat up the space to the rear exhaust fan. If your GPU is watercooled, then no problems in that edge case.

The only way to really know is....experimenting. Specs are only as good as marketing material, with some improvements here and there could be either decisive or only worth a few degrees at most. I recall the Snowman M-T6 cooler with 6 heatpipes only achieving 2-3 degrees over the M-T4 model, and that's as apple to apple you can get.

Big size coolers don't always dictate performance. In most cases, yes, but the ancient Deepcool Lucifer isn't as good as the size might suggest. You can get more mileage with more dainty coolers such as Ninja 5 and NH-U12A. Ditto with heatpipes as I elaborated above.

As my own rule of thumb, if it's similar in size and look it's probably gonna perform the same, give or take. The deciding factor would be the included fan(s) or mounting hardware. The latter is less of an issue these days with even budget brands having proper mounting brackets and none of the old plastic ring mounts.

Better fans would drop temps, especially more powerful ones. People might opt for balance between noise and performance, as well. I'd argue that installing better fans to a Hyper 212 would change its standing compared to more expensive coolers of its class. There's always the option to go nuts and get Delta fans or even a Blowie-Matron too.

Also, bigger is better, to a point. If the CPU isn't as demanding in wattage (thus heat production), bigger would result in marginal improvements. The gain in thermal resistance (or should I say, loss) isn't producing enough degrees of difference to be meaningful. For instance, NH-D14/15 won't always score twice as low as NH-U12S, it's marginal if any.

Edit: Additionally, TDP is nowhere as useful as it might suggest. TDP for coolers is rated at best case scenario, and the current smaller nodes in modern Ryzen CPUs are notoriously difficult to cool since there's less surface area to dissipate the heat. On top of that, the rated TDP for both Intel and AMD seems to be funky, I know that even 65W Zens can gobble up to 90+W with PBO. Steve Burke even covered AMD's statement about "marketing TDP" and "engineering TDP". It's a mess to say.

I had a tiny 92mm tower cooler on my i3-550, and that resulted in 50 degrees under OCCT, and that chip was rated for 73W TDP. No chance of doing the same on my 3500X, where a Gammaxx 400 Pro nets me 73 degrees without PBO with the same load.

The 92mm only has 2 heatpipes compared to Gammaxx's 4 heatpipes. Not to mention the increase in heatsink fins (the 400 Pro has more fins than vanilla 400). Small fans, poor airflow, bad mounting hardware, and MX-4 being used vs a pair of SP120 Performance Ed, proper metal/backplate mounting, and KPx thermal paste. If anything, the "65W" chip would result in lower temps vs the 73W one right?

TL;DR: check reviews, ask around, specs are near useless aside from size for clearances. TDP is out of the window especially.

 

[theamazingmettaton]

As my own rule of thumb, if it's similar in size and look it's probably gonna perform the same, give or take. The deciding factor would be the included fan(s) or mounting hardware. The latter is less of an issue these days with even budget brands having proper mounting brackets and none of the old plastic ring mounts.

Better fans would drop temps, especially more powerful ones. People might opt for balance between noise and performance, as well. I'd argue that installing better fans to a Hyper 212 would change its standing compared to more expensive coolers of its class. There's always the option to go nuts and get Delta fans or even a Blowie-Matron too.

Also, bigger is better, to a point. If the CPU isn't as demanding in wattage (thus heat production), bigger would result in marginal improvements. The gain in thermal resistance (or should I say, loss) isn't producing enough degrees of difference to be meaningful. For instance, NH-D14/15 won't always score twice as low as NH-U12S, it's marginal if any.

Edit: Additionally, TDP is nowhere as useful as it might suggest. TDP for coolers is rated at best case scenario, and the current smaller nodes in modern Ryzen CPUs are notoriously difficult to cool since there's less surface area to dissipate the heat. On top of that, the rated TDP for both Intel and AMD seems to be funky, I know that even 65W Zens can gobble up to 90+W with PBO. Steve Burke even covered AMD's statement about "marketing TDP" and "engineering TDP". It's a mess to say.

I had a tiny 92mm tower cooler on my i3-550, and that resulted in 50 degrees under OCCT, and that chip was rated for 73W TDP. No chance of doing the same on my 3500X, where a Gammaxx 400 Pro nets me 73 degrees without PBO with the same load.

The 92mm only has 2 heatpipes compared to Gammaxx's 4 heatpipes. Not to mention the increase in heatsink fins (the 400 Pro has more fins than vanilla 400). Small fans, poor airflow, bad mounting hardware, and MX-4 being used vs a pair of SP120 Performance Ed, proper metal/backplate mounting, and KPx thermal paste. If anything, the "65W" chip would result in lower temps vs the 73W one right?

TL;DR: check reviews, ask around, specs are near useless aside from size for clearances. TDP is out of the window especially.

Wow! This was extremely informative and reassured my hunch that Id need to experiment. Thank you!
I do have a question though, what is OCCT, im findin this, https://www.ocbase.com/ but i'd just like to be sure.
I see your point on why experimenting is important, cuz you wont always be given the most accurate info on the market, like the "65W" chips.
Oh, and what is considered the best case scenario for TDP?

 

[iPeekYou]

Wow! This was extremely informative and reassured my hunch that Id need to experiment. Thank you!
I do have a question though, what is OCCT, im findin this, https://www.ocbase.com/ but i'd just like to be sure.
I see your point on why experimenting is important, cuz you wont always be given the most accurate info on the market, like the "65W" chips.
Oh, and what is considered the best case scenario for TDP?

OCCT is much like Prime95, a software to test overclocking stability and temps. Large data set CPU set is for stability, I use small for most heat generation. Similar to Prime95 with same size data set since they both use AVX. OCbase is right on if you wanted the software for yourself.

AFAIK, cooler manufacturers use a thermal vehicle, basically a plate of copper heated to a BTU point to see how much Watts dissipated from a particular model of cooler. The watt then used as marketing TDP. I might be wrong on the technicalities since I'm not an engineer, but the point about thermal vehicle remains. It's convenient, reliable, and consistent. Not to mention easier than procuring chips of various transistor nodes (14nm, 7nm, so on).

That is what I refer to as best case. In theory, that is. No reviewer is going to use a thermal vehicle for cooler reviews since it's irrelevant to users.

In practice, older processors with larger transistor size is easier to cool, ceteris paribus. Old Intels with their 14nm can be improved by a lot through delidding and either changing the thermal compound under the heatspreader or just plopping a cooler on the bare die, yes, but that's not really comparable to the later soldered heatspreaders with newer Intel and Zens.

In other words, the smaller the chips get, the more difficult it'll cool. Zens can run up to 95°, or 85° per AMD's stock limit, but the old FX series are meant to run at 70ish. Seems worse in comparison, but the FX series were much easier to keep under 60 with light overclocks. People even gotten some OC on 8350 with their stock coolers, yet it's near impossible to do the same with even a 3600. Most people only see 60ish with decent cooling, whereas on stock cooler they can see 80 easily, even max 95° under heavy loads.

Edit: to clarify, you can OC Zens with stock cooler, you just need to set a low voltage (1.25v or so), especially doable with higher tier chips and/or newer batches. I'm talking about how the behaviour in temps between Zen and FX are different, with Zen being a worse case for cooling compared to older microarchitectures. Ditto with Intel, but that's because they essentially just pump more watts into their 14nm. Hotter, yes, but not due to decrease in transistor size.

An addition to your original question: even with reviews, make sure you read numerous sources to double-check. Some coolers here and there are rated higher or lower than they should be. If you're really keen on it, you can also find sources from other countries to expand your cross-check reference. I myself is particular to German websites. Read with a healthy dose of salt, and temper expectations.