This will depend greatly on your case airflow. And if you only run short load bursts (browsing etc.) which can easily be absorbed by the heatsink heat capacity or continous loads (games, work), where the exchange of heat from the heatsink to the outside world limits cooling.
With good fan control you could set the fan to switch off below a certain temp. So it could be silent 99% of the time but with a low rpm fan there if needed.
I did similar with my old SLACR Q6600 95w CPU - Noctua D14 (I think) with a fan on a controller. At stock speeds (with a pair of Noctua case fans on it) it had just enough airflow to run without the CPU fan running at all. When I wanted performance, I could overclock from 2.4ghz to 3.something ghz (I can't remember but I think they went to 3.6ghz?) and just turn the CPU fan up to 'normal' speeds and it'd never get above 70deg and it was still a very quiet machine - HDD noise was far more noticeable than fan noise.
I really must get some decent fans for my current rig - a slightly long-in-the-tooth A8-3870 mit 16gb RAM that is still running the OEM cooler. Yes, I've got bored of overclocking. I still have that noctua kit kicking around somewhere, really must dig it out and see if I can get an adapter for it. I'm sure that'll tide me over till we see if Zen is worth a light...?
My current machine is an i7-4790T (45W, 2.7ghz, quad-core, hyper-threaded, 3.9ghz turbo with HD4600 graphics) in an Akasa Euler case, which means the case acts as a heat-sink. As I type this I'm transcoding video on all eight hardware threads with a total load of about 760% (where 800% is max), at a CPU temperature of 85ºC and a clock speed of 2.97ghz.
Of course that's for a passive case rather than a heat-sink on its own, but as long as you have somewhere for that heat to go it definitely seems doable. For example if you used an open-air case then ought to just rise out between the heatsink fins so airflow may not be required at all.
Basically keeping the case from becoming a big box of hot air is crucial; the Euler case with my processor (which is a slightly higher TDP than the 35W that the case recommends) gets pretty hot internally, which isn't great for internal drives. I ended up having to swap an mSATA SSD for a 2.5" one, as the mSATA drive just got too hot, while the 2.5" one has a bigger surface area and a metal body. Even so, I squeezed a tiny 40mm fan inside just to help pull hot air out on warmer days.
So ehm… yeah, possible, but you have to be sure you've considered where that heat is going to go before you attempt it. But as others have said; if your case has room then you should just put fans in there anyway and set them to switch off at lower temperatures; you can also use very slow, quiet fans so even if they do run they're silent.
I fully agree with the Oxford Guy. I've got a NoFan CR-95C cooling my non-OC i7 4790K. This in conjuction with a couple of M.2 SSD-units, a passively cooled PSU and a passively cooled graphics card makes for a 100% quiet and rather powerful computer. To be on the safe side I've added a Noctua D14 which is configured to force air across all components when the motherboard temperature gets over 50 degC. I is almost never active, though. I've run the Prime95 "Torture Test" for prolonged periods but the CPU-temp consistently stays below 70 degC. In my opinion the NoFan unit is doing a splendid job, although at a price.
Listen to Oxford Guy. I've used three NoFan models and they all work amazingly well...as long as your CPU's power consumption stays under 100W. If you use a 6- or 8- core i7, or if you overclock enough to hit the 100W envelope, fanless is not for you. Note that NoFan coolers benefit only slightly when a fan is used. They are truly built as fanless coolers from the ground up.
My HTPC has a G2120 with NH-U12P, HD5670 with Accelero S1r2, 64GB Samsung 830 + 1TB WD Black along with 80+ Plat 400W fanless PSU. Inside Lian Li A05N. Only fans being filtered intake Gentle Typhoon @~600rpm and exhaust Slip Stream ~400rpm.
Pretty overkill cooling wise. Could drop the fan speeds even further.. To answer your question, yes it can easily handle it provided there's a teeny weeny bit of airflow in the case.
Work rig has a HR-02 Macho with 800rpm Slip Stream cooling a [email protected]/1.336V. Could run it fanless if I'd drop the clocks to say 4.3/1.1 or so..
It would simply be the ratio of surface area of the fins to the surface area of the top of the CPU making contact with the heat collector. The fans merely dissipate the heat more quickly over the same area.
The advantage of the fans are to transfer the heat by convection to the outer environment more quickly than allowing the heat to build up closer to other components in the system.
If designed for heat transfer, the other components are likely to have been designed assuming an ambient temperature at a particular max level, say 100-130degF. As the delta Temp between the environment and the part generating the heat will increase, so will the heat flow by conduction.
Intent is to draw the heat as far away from the components as possible.
One problem in these designs is to get the heat away from the CPU, as well as the Motherboard components, as well as other components in the case, so the interior case temperatures don't approach the environmental max design temps of those components.
A disadvantage in building by components, is that the component manufacturers are likely to only design for their particular component or one they support.
A common problem in Data Centers is how to remove all the heat from the racks and equipment within them. ANSI/TIA 942 stds go a long way to coordinate between disciplines and trades to effect proper HVAC in the server areas, but even within the racks and cabinets, too many designs limit themselves to providing a temperature set point at different areas in the room, but fail to flow adequate air over the equipment to transfer the heat away from the local electronics environments.
Computer Room Air Conditioners (CRAC) units are notorious for being installed to remove heat, but fail to provide adequate ventilation (air movement) within the computer rooms.
Since most of the CRAC units use split systems (condensate lines in 1/2" copper tubing running through the wall to a condenser outside the building), The natural trend would be to incorporate a small heat exchanger using a CPU water cooling fluid as the secondary, and the chilled water from the condensate of a HVAC system as the primary chilled water to remove the heat.
I haven't shopped the Enterprise level systems. I wonder if such systems are commodities.
True. It should be the default heatsink to compare with. Now that majority of Intel's CPUs become low power and efficient, these dual tower designs seem overkill except for the unlocked multiplier overclocker or fanless PCs.
Obsession with CM Hyper 212 EVO "Hypertwohundredtwelvetitis" is a disease also prevalent in Overclock.net. People go berserk over the 212, almost as if they have been mass brainwashed or mass hypnotized. To my best understanding, this mass hysteria is due to the fact that cheap "enthusiasts" may save up to the hugely important sum of $9.99 if they go with the 212 compared to other coolers for the wondrous performance gain of 0.8 Celsius. In other words, the mass hysteria with the 212 is because if you go with the 212, you will save enough money in the end to buy a pack of cigarettes and a can of beer.
Well every little bit counts, and to be honest I can understand why people would not want to spend $70-$80 on a heatsink. Getting a decent heatsink for $30-$40 makes sense for a lot of people. However, if you consider wasting money buying cigarettes to be reasonable, I can understand why you wouldn't put much stock in saving a few bucks.
Given how many overclockers and enthusiasts actually use the CM Hyper Evo 212 in their rigs (as eveidenced at Overclock.net) I think that Zodiacfml's suggestion of the CM hyper Evo 212 being used as a baseline cooler is a good one and I recommend the OP to take it.
I agree. It would be helpful to know how the 212 compares both with regards to cooling and quietness. I typically prefer "quieter" so I'd be curious to know how much better the "Dark Rock Pro 3" is than the 212... is it $40+ better?
I actually bought the 212 and added a second fan to it, not because it was cheap, but because it would fit between the 2x2 banks of memory on my Rampage IV Formula perfectly (with 1 mm of space on either side) allowing the tall memory heatsinks to rise up, and inconsequentially get a breeze from the fans. The CPU runs nice and cool (and quiet) with a modest overclock.
I'd get the offset Noctua NH-D15S if I ever upgrade from a 4 core 3820 to a 6 core 4930K.
Exactly my thoughts. The Cooler Master Hyper 212 EVO needs to be included in this test, to show exactly what 90% who buys CPU coolers are missing out on, or eventually not missing out on.
I'm confused about your complaint. What's wrong w/ the 212? What's wrong w/ saving $10 for 1 deg celcius difference?
I genuinely don't know, I have a 212 from almost 4 yrs ago? It works. It's quiet (for the time). The only complaint is if I try to go super small form factor, it won't work.
If it were only $10 I might agree with you, but when it's half the price, and sometimes a whole lot less, it makes a lot of sense. I looked at the Noctua when I built my 5820 last fall, and couldn't justify the 2.5x price. For $35, the 212 EVO is a great cooler. As good as the Noctua is, it's not two and a half times as good. That's why the 212 is so popular. It's in the proverbial sweet spot.
I agree. I've bought several 212 EVO's and I've been very happy with them. I was mostly looking for something that would run quiet under load (without overclocking) and I think they've been great. I've used some less expensive coolers and they were much worse - so in my opinion it's the cheapest cooler that met my needs.
Obsession is pretty harsh given the facts... I went and looked up a comparison over on frostytech, and it looks like the Hyper 212 evo is only 2C hotter than the Noctua chosen as the realistic 'best choice'.. for 1/3 the price. Given that my ambient temps change by more than that 2C over the course of a year, 2C is never the stability margin I use on my overclocks.
Except that the 212 is not a "premium" cooler. When you start getting into extreme overclocks like I have (i5 4690k @4.8GHz, or a 23% overclock) and into water cooling needs territory, the 212 falls well more than a 2C behind which is where it is on lower level overclocks (5-15%) on my chip.
Who cares whether if the 212 isn't a "premium" cooler when I can simply buy the 4790K at stock 4.4GHz instead of premium cooling to barely OC a 4690K past a 4790K. You overclockers STILL think there is tremendous value to be had with OCing when the 2500K ship have long sailed.
According to my benchmark tests in games and applications like Sony Vegas Studio, my overclock to 4.8GHz yields quite a bit of performance increase over the stock turbo of 3.9GHz. Oh and since I was on a budget and game about 80% of the time on that 4690K rig, I justified saving the extra $100 over a 4790K and put it towards a better GPU solution.
And yes, I still have a 2500K build as well (not sure what that has to do with the price of ketchup), which used to be overclocked to 4.6GHz on that NH-D14 cooler (it is now relegated to backup duties and running at stock speeds on a Zalman 9700LED cooler). Which, incidentally, roughly equals the performance of my Devil's Canyon chip running at 4.2GHz.
Which with simulated thermal loads ranging from 60-340W should have been made apparent in the course of testing.
I would hope and expect that most if not all of these coolers would out perform it, especially at higher loads. But as a de-facto standard budget cooler for people who want something better than Intel's I think it should've been added to the matrix to show how much better these bigger ones performed. A stock Intel cooler should've been included as well for the same purposes (at least at the lower loads; no sense risking burning the test setup by trying to broil 340W through it). Including a mainstream reference point is especially important in this case because E. Fylladitakis's synthetic test load means that we can't cross reference his results with those found elsewhere.
Totally agree - to get the full picture and to aid comparability with other tests the reference points of the Intel stock cooler (free) and CM 212 EVO (£25) should be included.
If you buy one of those things on sale I've seen them as low as $30, which if you don't need better cooling, is a good deal. The reason the Hyper 212 EVO is popular is that it's cheaper than most of it's competition and easily available. They're good enough for moderate overclocking on a 1150-series chip so they do fit the bill for a lot of people. Something being popular doesn't make it bad.
I have 2 of the original Hyper 212 (Non-evo) on two different I5-2500Ks that have been running super quiet at 4.4ghz for the last 4+ years. No coil whine or bearing degradation on the fans. I paid $20 for each Hyper 212 Evo. The value for the money is amazing. It's an excellent quality reliable product and it's easy to install.
I paid $220 for my I5-2500k, 20$ for the Hyper 212 Evo, and $70 for 8gbs of PC 1333 in February of 2011. At 4.4ghz, it's still within 5% as fast as any CPU on the market. Sandy Bridge FOREVER! I'll keep buying video cards. You can waste your money on HSF upgrades for CPUs that become less important every day. DX12 is just going to make the CPU even less useful.
Remember, the average exchange in 2011 was 1.011 (CAD to USD). It is presently 0.79 (CAD to USD). Assuming US/CAN price parity in 2011, that $220 cooler would cost ~$278, not that far behind the $284 i5 4690K.
Used to have 212+ once. Later I swapped out for a cheap closed-loop. Though the CPU temp is a few C lower, the closed-loop was much louder than the 212+ (due the the 'pump whine').
Years later I upgrade the rig to a Core i5, which is not really that hot, and I'm not interested in overclocking anymore (being more mature I guess).
I find the 212 is pretty good for its price. It's a great entry-level cooler for those who want to upgrade. I also think that it could serve well as a baseline for the comparision.
I didn't know that overclocking enthusiast would prefer a lower performing heatsink instead of the best available. The reason is simple; CPUs consume less power throughout the years even with continuous but non synthetic workloads including gaming.
Many years ago, I was a fan of watercooling then big-air heatsinks then not anymore. It is just not logical anymore as they are more expensive, larger, and cumbersome.
To me, the whole point of overclocking is to get a better cpu than what you paid for. So overpaying for a heat sink doesn't make sense - the whole point is to get the best possible performance, while spending the least amount of money. At least that is what overclocking means to me, and I'm sure a lot of other people as well.
That's usually how most people start with overclocking. For others, it's getting the best performance regardless. That's why people still that the i7-K and push it, rather than a Pentium-K and tweak it.
I bought my Hyper 212 for $19.99 - a much bigger savings than $10. It does the job, and in the end my overclock was not limited by temperature, but by the CPU itself. A more expensive heatsink wouldn't gain me anything.
What are you even talking about. The 212 is $35 and half the cost of many of these heatsinks. Its been the gold standard for years, if you only get another 1 C out of a HSF that cost double then its not worth it.
(In case someone is wondering: No, I didn't over-torque it, the bolt got caught in the back-plate and sheared under the power of a screwdriver lightly applied; And yes, that bolt is hollow; And yes, that is the mounting hardware from a CM 212 EVO; And yes, I'm done buying CM products.)
Are you kidding me? Intel's CPUs might be efficient compared to AMD's, but there is hardly a valid reason to dismiss the dual-tower crowd. Intel's CPUs are hotter than they have been in years, thermal performance having declined steadily since Sandy Bridge due to sh*ttier and sh*ttier TIM and other reasons.
They did one of these round ups with the 212 Evo or + a while back, also involving high ends from Noctua (U14 and U12S I believe). They found that it doesn't quite match up, but I know it got a mention for exceptional performance for cheap. I think it falls behind more in noise than performance.
It definitely would have been interesting to see it in here, but nevertheless, thanks for the review!
I would have also liked to have a "decent" CPU cooler like that included, as well as the stock Intel/AMD HSF. It's great seeing how these coolers stack up to one another, but it doesn't truly quantify how much of an improvement you're getting over a cheap alternative, or the stock fan. For the record, I haven't run with a stock fan on any main PC I've owned in the past 15 years, but I would be curious to see how much I'm actually gaining.
Especially given the raw value of the Hyper 212 Evo at $30, it may get within a degree or two of some of these for half the price or less... which is why my last build had the Hyper 212+ (It's been a few years) - I could have gained maybe 5C by spending 4 times as much.. which didn't seem worth it to me.
AGREED. I have one, and at $29.95 from newegg just a few months ago on sale it was an AWESOME deal. i4790k can do massive oc's with it and even at full load is not terribly annoying with my 5850 causing most noise when gaming. This is still a top seller and for good reason.
I'm actually really glad to see this article, it's been ages since I've seen a good CPU air cooler roundup and sockets have changed several times over the years so it's nice to know what works well these days.
air cooling has plenty much run into a wall; heatpipes to copper base, aluminum fins on the heatpipes, put 140mm or 120mm fan... there is not a lot of wiggle room, so performance of those that follow this recipe is very close.
differentiators now for most part are: socket compatibility, price, installation method. Raw performance/noise is no longer the focus imho if you want a successful product
It's not so much socket compatibility, so much as how compatible you can make your heatsink against mobos that have poor design choices.
Although not as common on mATX and larger boards, mITX suffers a lot from this, because manufacturers attach fragile bits onto the back of the mobo, near the CPU socket, that interfere with the mounting bracket. Either that, or the CPU socket is placed too close to the PCIe, etc.
That Reeven Okeanos is something I haven't seen since Athlon 64 days, which are heatsinks paired with a stupidly loud fan. Look, if I wanted a heatsink with stupidly loud fan, I would buy an amazing heatsink or watercooler first, then attach the stupidly loud fan to that, instead of some mediocre heatsink with a mediocre fan.
I would of course like to see the testbed updated to have fan/s controlled by thermals, ie something like ASUS Fan Expert. Set for a target temperature and loads in the more realistic range of 15-150 watts. And of course when reducing voltage it should be via PWM and not simply reducing static voltage. The results should then be presented in temperature variance and noise levels/profile.
Good suggestions. I hate those charts with "full fan voltage" and "fan voltage reduced to ..". What I really care about is "how silent can the cooler be for a given temperature / cooling performance?" And "which one cools better at similar noise level".
It doesn't help much to see a strong fan with inbearable noise in those charts. Even if someone is interested in such solutions - wouldn't his question rather be "which heatsink performs best with this high-speed fan"? Which would again be something he couldn't answer from this data.
I know making noise based comparisons is difficult. But the raw sound pressure could be accompanied by some subjective remarks regarding the noise spectrum.
I don't know if the stock HSFs have changed since Sandy Bridge, but my 2500k would hit 98+ degrees and throttle under AVX loads with the pathetic little stock thing whilst sounding like a small tornado had developed inside my case (how is it that a 95w CPU comes with an HSF half the size of the ones they used to ship with 65w C2Ds?!)
Whichever cheap tower cooler I replaced it with does the job just fine, though. It's been running like a champ at 4.5GHz for near enough 4 years now. (I think it's a Xigmatek SD1283 - I haven't even taken the side off my machine for over a year, those heady days of tinkering and yearly upgrades long since passed.)
No it isn't. The mounting mechanisms on these higher priced sinks are, as, had you read the article, solid. If you are missing an arm and wanted to mount a 100$ heat sink in your 25$ case, you might care about the weight, in which case, the manufacturer's website is a one armed lazy click away.
Cheaper still, at under $50, is the ZALMAN CNPS9500A. I use it on almost every build due to it's all-copper design and razor thin fins. I can keep the fan on the lowest speed without my Core i7 going over 40 Deg. C under load. It's silent no matter what I'm doing.
My only complaint is that the fan is not PWM. Quibbles.
The only thing I haven't liked about Zalman, is that the fans they use in their heatsinks aren't exactly the quietest and are usually not easily swappable. At least, not when compared against manufacturers like Scythe and Noctua.
I have the 9700LED variant of that design for my old Core 2 Duo E8400 build. It does okay for medium overclocks (running my 3.0GHz E8400 at 3.4GHz), but that was about it. And considering I paid $55 for it back in 2009 when I built that rig (about $60 in today's money), it wasn't exactly a cheap option. I would have spent a little more for a better cooler so I could get a higher overclock, but the new Sandy Bridge chipset was coming out soon and I decided to just keep it as a backup rig.
I just want to mention that the Cryorig R1 Ultimate CPU Cooler that you listed and said was currently available only through a foreign store registered in Amazon.com that ships from Korea is currently being sold at PC Case Gear here in Australia for AU$89 (US$66.82).
I would have liked to see the Noctua's real arch nemesis included in this shootout, the Thermalright Silver Arrow. Crushes the Macho Zero. I have a Silver Arrow on my i7 4790k running seti@home, and it owns.
The Phanteks PH-TC14PЕ is a better cooler using noctunas own fans on it and even better if u add AF140 fans to it . Is why i took it over the Noctua for myself.
It's not so much as updated, as it is an off-center version of the same NH-D15. I would need the NH-D15S that you linked to on my MB as the former would hit my GPU in the first slot. I was considering modifying the D15 to move it up and away until I saw the D15S. A good link none the less.
I'm a bit disappointed that you didn't comment on the install process on any of the heatsinks. I recently bought a Dark Rock Pro 3 and while I love how quiet it is and the temps are actually lower than the Corsair Hydro H80 it replaced, the install process requires you to screw the heatsink in from the back of the motherboard. That and the size of the supplied backplate made the heatsink install more difficult that is really necessary.
If you buy a Dark Rock 3 Pro I recommend removing the motherboard from the case entirely and installing it by flipping the heatsink upside down and balancing the motherboard on top of it in the correct position. This makes it fairly easy to screw in. But if you are using a normal thermal paste you might need to put it on the heatsink instead of the CPU heatspreader. I use IC Diamond and that stuff is so thick that it just stuck there upside down for long enough to finish mounting the thing.
I never gave it much thought. Installation is such a small concern to me, maybe I do this more than most, price and performance are preferable. That said, I still think Noctua's mounting can be improved, it seems unnecessarily complicated to me. First off, you really do need to replace the plastic bracket, there's no way around that. But secondly, why include a 140mm screw driver? Why not make the screws 140mm taller? Then you can just use a common screw driver, even a stubby or a pocket knife. And make them captive so they do not fall out and you do not need to line them up. These will certainly add to the cost due to extra engineering time and unique screws.
Because the 140mm screwdriver: A: Is cheaper than re-engineering the entire product B: long screws are *very* easy to cross-thread due to the extra sideways torque you can apply when inserting them.
I *do* like the Noctua setup system. It's strong, comprehensive, and lets be honest, you only do it once. I'm pretty sure that any gotchas with installation were caught in the descriptions of each cooler, too.
For the Macho Zero especially, I'd want to see the tests conducted (additionally) in a vertical motherboard orientation (as you'd have in most tower cases), since convection may have an effect on performance. It may also reveal differences in fan noise due to bearings rubbing more or less on different surfaces.
Not likely. Convection is slow. Any fans will blow away convection currents. Besides, orientation is strictly a "case by case" basis and beyond the scope of an empirical HSF comparison.
Fan noise due to orientation may be good to check for though. I doubt it will be any different, if it is, the aberration should be noted.
Something really important that never is reviewed is dust control and maintenance.
As radiator fins get closer, they have more dissipation area, but air flow gets worse. Yet the main problem of fins closeness is that they accumulate dust much faster, and get occluded.
No review makes a dust test. I have a Thermaltake cooler, which works excellent when is clean, but it rapidly loses his capacity due to being occluded by dust.
The easiest way to remove his dust is to use canned air, but it is a short lived fix, because an air cleaned radiator occludes itself very fast, sometimes in matter of weeks.
The only way to clean it for good is to take out the dissipator, and put in in the dish washer. But that is a cumbersome task. I need to do a lot of work just to clean it, and I need thermal paste to place it again on the motherboard.
So, maintenance is as important as cooling and noise.
A good cooler should pass a dusting test (being exposed to a day of dirty and dusty air current), and should be possible to clean it without much hassle, without applying thermal paste, without using screws, without need to access both sides of the motherboard, and without his pegs breaking or being degraded by manipulation.
Interesting concern. It makes sense, yet is never considered. I think the best answer is the same as electricity, you need to provide the best environment you can for your computer. Higher quality components are more sensitive to dirty power, and dusty air. Get that PC off the floor, and get a case with good dust filters. And if your room is bad enough, get an room air filter. I have one on my desk right in front of my PC. A Honeywell HEPA air filter.
I understand why cases are tested with stock fans, replacing 3+ quality fans will significantly alter the price equation... It's a trivial difference for $50-100 heatsinks tho, many are often even sold sans fans anyway.
If nothing else, a single apples to apples test with all running the same fan would've been welcome. In addition to something like the Hyper 212 as a baseline, the different Thermalright TRUE Spirit variants have remained a solid value over the years (they perform better than the 212 and even closer to some of these as per HardOCP's last roundup).
Is there gonna be a midrange roundup? Spending upwards of $65 on an air cooler never made much sense to me when more affordable options were so close in both noise and performance. Out of 9 coolers tested only 3 or so come in at a sensible price, at $75+ wouldn't it make more sense to go AIO WC?
If you want low noise, good thermals and dead quiet idle well into the land of diminishing returns, high end air is still a very compelling alternative to water, especially if you're willing to put the effort into some ducting.
However midrange products can be excellent and are well worth a good long look before hanging a hundred bucks off your socket.
I was a HSF fanboy until I actually needed a water cooling system. Specs and testing does not show response time. Fast, transient loads can overwhelm a HSF, as heat pipes cannot transfer heat as fast as water. For low power CPU's a HSF is fine, but when you go over 100 watts the noise is less of an issue, as a HSF will need much more airflow to compare to the higher heat transfer ability of W/C. And instantaneous loads are more likely to cause system instability due to the less efficient heat transfer rates of heatpipes.
I am going to nitpick on what you're saying there, because you have it confused.
Heatpipes use some form of vaporized liquid inside and they actually transmit heat quite good, especially when compared against sold copper rods. Where they fail, is that they do not have a lot of capacity for transferring large amounts of heat.
Water is actually quite poor at transmitting heat, as it's a non-metal and an insulator. It does, however, have great capacity to hold heat.
This is why large bodies of water are warm during winter and cold during summer and results in mild weather near the coast and extreme weather in the interior.
The advantage of water cooling systems, over air cooling, is the ability to place large radiators, with a lot of surface area, in spots that get fresh air from outside the case, instead of warmed up air that is already inside the case. Heatsinks are also limited to size and weight constraints around the socket.
Were it possible to attach a triple 120mm fan heatpipe heatsink, as in parallel layout, instead of serial dual towers, directly to the CPU, I'm quite sure the results would be similar to a 360mm radiator water cooling setup.
Yeah, he's describing the thermal mass of the coolant, not any transfer capability. It seems that running coolant through radiator fins does have an advantage of relying on the fins to radiate the heat out from a heatpipe, but for a similar surface area, it's questionable how large the advantage is, especially at CPU TDPs.
GPUs on the other hand have a more constrained form factor for large air cooling and a higher TDP, so they are a more promising place for CLCs.
I'd be very interested to see a size-per-performance chart.
I usually buy coolers that perform the best without being too large. Although I guess "Top Tier" coolers isn't the right article to look for smaller size coolers :p
Leaving it to the manufacturers to select their entry is always messy or questionable IMO, TR choosing the Macho over the Silver Arrow for one... Still curious whether there's a midrange round up planned or if this is it for air coolers at AT for a couple more years...
Despite the criticism I do appreciate E Fyll's thorough process, otherwise I imagine I and many others wouldn't even bother commenting. Low end (212+) and high end ($120 AIO?) base lines would make the article much more useful tho.
The into even suggests some people prefer high end air over WC at the start but then leaves that up in the air... No pun intended.
why didn't you test these all with the same fan? Then we could see how the cooler performed independent of the included fans? Like the Thermalright, it comes with no fan, apparently you got a super quiet slow fan and put on there, but that isn't fair to Thermalright saying they are hotter, when it could be because it doesn't have the amount of air moving over the fins as others.
When you talk about ultimate cooling, you should've at least tested the True Spirit 140 Power Edition with a TY-143 fan, or the Silver Arrow SB-E Extreme. Those are the most interesting products from Thermalright, not the Macho Zero.
Maybe as an addendum at some point? I'm very interested to see how the Silver Arrow and the 140 PE fare against the D15.
Apparently the Macho is what TR themselves chose to send AT, if the intro is accurate... Big /facepalm on their part. They probably have some of the best value coolers in the TRUE Spirits (I have the original Cogage version myself), and the Silver Arrow might've ranked up there with the Noctua and Phantek. The article did make me pretty curious about the latter tho, call me vain but the color choices are cool.
Thanks for doing a heatsink round-up. They are refreshing to see these days.
It is a shame that Thermalright did not send in its top-tier performer. Then again, the Macho Zero is nothing to sneeze at. ~40C over ambient for a 340 watt load is still a good result. (Perhaps, instead, they could have thrown the TY143 performance fan instead? Har!)
I'd like to see the measurements all with the same fan(s) - whatever is considered "the best" fan. That would give an indication of how much you could get out of the cooler with aftermarket fans.
I wish there was more review cross-over with water coolers, these two camps seem to be at odds with each other. They never seem to be compared effectively to each other, so it is difficult for consumers to determine the "best" cooler for themselves. With Noctua getting up to $93, there are water coolers out there for less. I bought my Noctua NH-D14 for $75 and thought that was high for a HSF.
I strictly used air coolers until I got an AMD APU, among them are several Noctua models. It was apparent to me that this CPU, after a bit of easy O/C, got hot much too fast for an air cooler to absorb. It would crash after just 4 seconds of starting Tomb Raider and the cooling fins were still ambient temperature. I tried three coolers including the Noctua NH-D14. Fan speed did not matter as the rapid increase in temperature exceeded the heatsinks' ability to draw the heat off the CPU itself. I would guess that it take anywhere from 30 seconds to a minute for the heat to actually get to the fins, so if your heat sink cannot "sink" the heat all by itself, no fan, for a full minute, then it has inadequate heat transfer and no fan will fix that.
I installed a Corsair H100i and that works very well. I had previously thought that any cooler with less surface area would have less cooling performance, but I have found that if you cannot transfer the heat to the fins, they make no difference. I think a Corsair H60 would have been fine now. I heard that water coolers were "better" and more efficient, but nobody ever explains WHY.
From this experience at least, it appears that water coolers have better heat transfer performance. Fan speeds and fins are secondary to that, as they do not matter until the heat gets to them. If they get hot, then low speed fans can easily remove that heat as higher temperature differentials generally allow for greater heat transfer. If you run high-power and high-heat for a long time, then higher fan speeds help.
How quickly can your test bench ramp up in power? Was that tested? Was that considered? CPU's can hit maximum power in nanoseconds, and crash in milliseconds. Only the base of the HSF would see anything from that event. I think this test is more academic, and not very relevant in the real-world with actual CPU's. It only tests for maximum heat generation over time, like when running benchmarks, not the dynamic nature in which CPU's operate for most useful loads. But then, that's just my perspective.
Someone should test for this. It is quite plausible as it will take time to increase the temperature of water/coolant in the system. Yet, for longer and continuous loads, surface area matters to dump heat in the environment which AIO coolers has less to a similarly priced Big Air cooler which shows in benchmarks.
It'd be nice if there was a base line (say intel's current stock cooler) and an AIO water cooler to show the difference between a top tier air cooler and what many consider to be the next step.
One point, SPCR always test each cooler with a 'reference' fan, making it easier to compare the heatsink's efficiency. Can't help but feel you didn't complete the task you set out to by levelling the playing field with your simulated CPU heat source.
Great roundup! i for one still believe high end air a good alternative to water if you want dead silence.
What i still can't understand is how can Nocuta get away with shipping the most expensive air cooler, likely to be used by enthusiasts who care about aesthetics, with the most hideous fan colours?
Noctua colors are odd, but I think it forces you to consider performance more than aesthetics. They do look nice, but not "cool" - no pun in tended.
Dead silence basically depends on how much heat you need to dissipate, temperature of ambient air, surface area of fins/radiator, air flow, and heat transfer capability of the cooler. Or more basically, how easily you can transfer heat from CPU to the air. Water increases the efficiency of heat transfer, so once you reach a certain power level they will overtake HSF's. They will never be as quite as a good HSF, at least not on low-power CPU's that arguably do not need anything but the stock cooler to begin with.
Noctua gets away with those colors entirely because they're Noctua. Yes they're not pretty colors, but unlike nice bright/garish colors they are an unmistakable sign that they're the best or at least tied for best.
Well, serious people are more likely to buy Noctua anyway. And no serious user cares about what the fans look like, and even if they did, they would still take the Noctua, since you can CLEARLY see the extremely high quality they are just by looking at them. Windows in cases and other stupid bling bling is for the PC-ricer gang. They only choose after looks anyway, and dont care if their fans make noises from day one.
Excellent review of what seem to be very good HSFs throughout (but then again, you did ask them to send their best).
A comparison with the stock Intel cooler as supplied with something like the 4790K would be nice, as that is what any expense on a third-party cooler must be compared against. How would the stock-cooler compare at 340W? I was very impressed with the degC/W figures for all those coolers; it almost makes 340W seem an amount of heat that can be dealt with, though I'm thinking it's crazy.
As well as the stock-cooler from a highish end CPU, one or two commonly used AIO liquid-coolers needs to be added to provided a second comparison point, as these high-end air-coolers are presumably expected to be near, possibly even better than a liquid AIO.
I love the fact you are using a properly designed heating rig, and your comprehensive review, but I came away thinking: they're good! But I've no idea how much better they are than the stock cooler, or how any of them would compare with the liquid AIOs also available, both in terms of cooling or noise.
This is probably the best CPU cooler article I've read (and I read such articles when Anandtech when it was on Geocities). Excellent writing, useful measurements, and a cogent conclusion. Thank you, Emmanouil!
I've had great performance with the Spire Thermax Eclipse II for five years now, and for the last three years, even better performance when paired with two Corsair SP120's. It's still the highest rated heatsink on Frostytech; not sure why it isn't included here.
Whatever was supplied - it's a test of the product as supplied (for most of them) or as recommended by the manufacturer (as in the TR one, where they provided a suitable fan when requested, as the rest came with one).
Thermal pastes are best tested on *one* cooled as a known quantity, really, which would be a separate test of it's own, I'd say.
Thanks for the review! This is certainly your best review since joining AT, kudos to getting a good spread of products from around the world. There were a few in there that I hadn't heard of so it was great that the final results would be a surprise.
The results were somewhat what I expected and nice to see where each of the companies's products sit. I have to say that not having the Cyrorig on the final recommendation is a bit sad as their products are simply bespoke. I don't have the R1 Ultimate but I do use their C1 in my SFF case on an overlocked Core i5 4690k and the temperature to noise performance is unbelievable, especially compared to some SFF coolers I've used in the past from other vendors.
Looking forward to more of these, perhaps a SFF cooler showdown could be next...?
I went with the Phanteks PH-TC14PE back when I built my Haswell system back in 2013. It and my (de-lidded) 4770K are still holding up fine.
All those coolers are really close though, especially at full fan speeds. We've probably hit the limit of air cooling with heatpipe technology at this point.
Why review the Macho Zero and not the True Spirit 140 Power or the IB-E Extreme? Those 2 are far a better match for the rest of the lineup that the one you reviewed from Thermalright.
LOL - "Liquid-based cooling solutions are becoming easier to install and AIO kits generally are hassle-free, yet they are still not favored by the majority of the users. Their space requirements, increased complexity and price hold most people to simple air-based cooling solutions."
so they become easier to install, hassle-free but somehow managed to keep their complexity? what are talking about??? :)
anyway, I am sure you know this and you just overlooked this fact to support your own story, liquid cooling sets are NOT more expensive (Corsair H50 costs $60 shipped), are NOT bigger (H50 compared to any of the air coolers here for example), and are NOT complex (whatever you mean by that), or even the noise factor where air coolers need to run on higher RPMs to achieve same cooling effectiveness as liquid coolers - simply there is no comparison between liquid solution and air solution. Cheap Corsair H50 is far superior in every way than those colossal monstrosities you are reviewing here now. Smart user would never chose an air cooler simply because it doesn't make sense to chose an old and less effective idea.
Therefore please do not spread bullshit just so you can cash a check for an article.
Liquid cooling doesn't always work well. I got a Corsair H60 to put in a Corsair Obsidian case with an ASUS Sabertooth R2.0 mainboard and an AMD FX8350 processor. It worked brilliantly with CPU temperature at ~35C under load. Unfortuately, the layout of the case and lack of airflow around the CPU heatsink lead to the VRMs/MOSFETS around the CPU reaching temperatures of over 85C, no matter how I arranged the case fans. I replaced it with a Noctua cooler similar to this one which also kept the CPU nice and cool, but with the side effect of the large 14mm fans producing sufficient airflow onto the mainboard to eliminate the dead space which caused overheating.
This isn't to say that closed loop coolers are bad; their performance can be very good. But they aren't compatible with every case/mainboard.
First: it's been almost a month - E. Fylladitakis - must be on vacations since he has yet to reply to my comments.
Second: rleigh - very sorry to hear that you had to install additional cooling for your motherboard. It is best to keep high air flow inside the case, if you didn't provide that then no wonder you had an overheating problem.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
135 Comments
Back to Article
Pissedoffyouth - Monday, July 6, 2015 - link
I wonder how a D15 or similar, with the fans removed, would work with a 45w or 65w APU to make a passively cooled PC.ImSpartacus - Monday, July 6, 2015 - link
That's a really interesting consideration.MrSpadge - Monday, July 6, 2015 - link
This will depend greatly on your case airflow. And if you only run short load bursts (browsing etc.) which can easily be absorbed by the heatsink heat capacity or continous loads (games, work), where the exchange of heat from the heatsink to the outside world limits cooling.iamezza - Tuesday, July 7, 2015 - link
With good fan control you could set the fan to switch off below a certain temp. So it could be silent 99% of the time but with a low rpm fan there if needed.Mumrik - Tuesday, July 7, 2015 - link
You'd probably still need some level of airflow.I've never had a fan on my Scythe Ninja that cools an i5-2500k. I think that's a 95W TDP.
It's close to the single 12cm rear exhaust though.
Beany2013 - Wednesday, July 8, 2015 - link
I did similar with my old SLACR Q6600 95w CPU - Noctua D14 (I think) with a fan on a controller. At stock speeds (with a pair of Noctua case fans on it) it had just enough airflow to run without the CPU fan running at all. When I wanted performance, I could overclock from 2.4ghz to 3.something ghz (I can't remember but I think they went to 3.6ghz?) and just turn the CPU fan up to 'normal' speeds and it'd never get above 70deg and it was still a very quiet machine - HDD noise was far more noticeable than fan noise.I really must get some decent fans for my current rig - a slightly long-in-the-tooth A8-3870 mit 16gb RAM that is still running the OEM cooler. Yes, I've got bored of overclocking. I still have that noctua kit kicking around somewhere, really must dig it out and see if I can get an adapter for it. I'm sure that'll tide me over till we see if Zen is worth a light...?
Essayjedii - Friday, July 10, 2015 - link
I have made a post about D15 in here <a href="http://www.dumblittleman.com/2015/04/14-problems-f... Hope ou find it interesting and useful.Essayjedii - Friday, July 10, 2015 - link
I have made a post about D15 in here http://www.dumblittleman.com/2015/04/14-problems-f...Hope ou find it interesting and useful.
Haravikk - Sunday, July 12, 2015 - link
My current machine is an i7-4790T (45W, 2.7ghz, quad-core, hyper-threaded, 3.9ghz turbo with HD4600 graphics) in an Akasa Euler case, which means the case acts as a heat-sink. As I type this I'm transcoding video on all eight hardware threads with a total load of about 760% (where 800% is max), at a CPU temperature of 85ºC and a clock speed of 2.97ghz.Of course that's for a passive case rather than a heat-sink on its own, but as long as you have somewhere for that heat to go it definitely seems doable. For example if you used an open-air case then ought to just rise out between the heatsink fins so airflow may not be required at all.
Basically keeping the case from becoming a big box of hot air is crucial; the Euler case with my processor (which is a slightly higher TDP than the 35W that the case recommends) gets pretty hot internally, which isn't great for internal drives. I ended up having to swap an mSATA SSD for a 2.5" one, as the mSATA drive just got too hot, while the 2.5" one has a bigger surface area and a metal body. Even so, I squeezed a tiny 40mm fan inside just to help pull hot air out on warmer days.
So ehm… yeah, possible, but you have to be sure you've considered where that heat is going to go before you attempt it. But as others have said; if your case has room then you should just put fans in there anyway and set them to switch off at lower temperatures; you can also use very slow, quiet fans so even if they do run they're silent.
Oxford Guy - Tuesday, July 14, 2015 - link
Stick with a cooler like the NoFan models which are made specifically passive cooling. They will be much more effective.Narcissist - Tuesday, July 14, 2015 - link
I fully agree with the Oxford Guy. I've got a NoFan CR-95C cooling my non-OC i7 4790K. This in conjuction with a couple of M.2 SSD-units, a passively cooled PSU and a passively cooled graphics card makes for a 100% quiet and rather powerful computer. To be on the safe side I've added a Noctua D14 which is configured to force air across all components when the motherboard temperature gets over 50 degC. I is almost never active, though. I've run the Prime95 "Torture Test" for prolonged periods but the CPU-temp consistently stays below 70 degC. In my opinion the NoFan unit is doing a splendid job, although at a price.Sivar - Friday, February 5, 2016 - link
Listen to Oxford Guy. I've used three NoFan models and they all work amazingly well...as long as your CPU's power consumption stays under 100W. If you use a 6- or 8- core i7, or if you overclock enough to hit the 100W envelope, fanless is not for you.Note that NoFan coolers benefit only slightly when a fan is used. They are truly built as fanless coolers from the ground up.
lagittaja - Monday, July 20, 2015 - link
My HTPC has a G2120 with NH-U12P, HD5670 with Accelero S1r2, 64GB Samsung 830 + 1TB WD Black along with 80+ Plat 400W fanless PSU. Inside Lian Li A05N.Only fans being filtered intake Gentle Typhoon @~600rpm and exhaust Slip Stream ~400rpm.
Pretty overkill cooling wise. Could drop the fan speeds even further..
To answer your question, yes it can easily handle it provided there's a teeny weeny bit of airflow in the case.
Work rig has a HR-02 Macho with 800rpm Slip Stream cooling a [email protected]/1.336V. Could run it fanless if I'd drop the clocks to say 4.3/1.1 or so..
Cvengr - Friday, December 25, 2015 - link
It would simply be the ratio of surface area of the fins to the surface area of the top of the CPU making contact with the heat collector. The fans merely dissipate the heat more quickly over the same area.The advantage of the fans are to transfer the heat by convection to the outer environment more quickly than allowing the heat to build up closer to other components in the system.
If designed for heat transfer, the other components are likely to have been designed assuming an ambient temperature at a particular max level, say 100-130degF. As the delta Temp between the environment and the part generating the heat will increase, so will the heat flow by conduction.
Intent is to draw the heat as far away from the components as possible.
One problem in these designs is to get the heat away from the CPU, as well as the Motherboard components, as well as other components in the case, so the interior case temperatures don't approach the environmental max design temps of those components.
A disadvantage in building by components, is that the component manufacturers are likely to only design for their particular component or one they support.
A common problem in Data Centers is how to remove all the heat from the racks and equipment within them. ANSI/TIA 942 stds go a long way to coordinate between disciplines and trades to effect proper HVAC in the server areas, but even within the racks and cabinets, too many designs limit themselves to providing a temperature set point at different areas in the room, but fail to flow adequate air over the equipment to transfer the heat away from the local electronics environments.
Computer Room Air Conditioners (CRAC) units are notorious for being installed to remove heat, but fail to provide adequate ventilation (air movement) within the computer rooms.
Since most of the CRAC units use split systems (condensate lines in 1/2" copper tubing running through the wall to a condenser outside the building), The natural trend would be to incorporate a small heat exchanger using a CPU water cooling fluid as the secondary, and the chilled water from the condensate of a HVAC system as the primary chilled water to remove the heat.
I haven't shopped the Enterprise level systems. I wonder if such systems are commodities.
sjakti - Monday, July 6, 2015 - link
Interesting article, thank you! I especially appreciate the "Quick Conclusions", that's a great table.Shadow7037932 - Monday, July 6, 2015 - link
I wish you guys had included the Hyper 212+/EVO in the review as the base comparison.zodiacfml - Monday, July 6, 2015 - link
True. It should be the default heatsink to compare with. Now that majority of Intel's CPUs become low power and efficient, these dual tower designs seem overkill except for the unlocked multiplier overclocker or fanless PCs.Achaios - Monday, July 6, 2015 - link
Obsession with CM Hyper 212 EVO "Hypertwohundredtwelvetitis" is a disease also prevalent in Overclock.net. People go berserk over the 212, almost as if they have been mass brainwashed or mass hypnotized. To my best understanding, this mass hysteria is due to the fact that cheap "enthusiasts" may save up to the hugely important sum of $9.99 if they go with the 212 compared to other coolers for the wondrous performance gain of 0.8 Celsius. In other words, the mass hysteria with the 212 is because if you go with the 212, you will save enough money in the end to buy a pack of cigarettes and a can of beer.Nagorak - Monday, July 6, 2015 - link
Well every little bit counts, and to be honest I can understand why people would not want to spend $70-$80 on a heatsink. Getting a decent heatsink for $30-$40 makes sense for a lot of people. However, if you consider wasting money buying cigarettes to be reasonable, I can understand why you wouldn't put much stock in saving a few bucks.Achaios - Monday, July 6, 2015 - link
Given how many overclockers and enthusiasts actually use the CM Hyper Evo 212 in their rigs (as eveidenced at Overclock.net) I think that Zodiacfml's suggestion of the CM hyper Evo 212 being used as a baseline cooler is a good one and I recommend the OP to take it.mrvco - Monday, July 6, 2015 - link
I agree. It would be helpful to know how the 212 compares both with regards to cooling and quietness. I typically prefer "quieter" so I'd be curious to know how much better the "Dark Rock Pro 3" is than the 212... is it $40+ better?Eidigean - Monday, July 6, 2015 - link
I actually bought the 212 and added a second fan to it, not because it was cheap, but because it would fit between the 2x2 banks of memory on my Rampage IV Formula perfectly (with 1 mm of space on either side) allowing the tall memory heatsinks to rise up, and inconsequentially get a breeze from the fans. The CPU runs nice and cool (and quiet) with a modest overclock.I'd get the offset Noctua NH-D15S if I ever upgrade from a 4 core 3820 to a 6 core 4930K.
effortless - Monday, July 6, 2015 - link
Exactly my thoughts. The Cooler Master Hyper 212 EVO needs to be included in this test, to show exactly what 90% who buys CPU coolers are missing out on, or eventually not missing out on.randomlinh - Monday, July 6, 2015 - link
I'm confused about your complaint. What's wrong w/ the 212? What's wrong w/ saving $10 for 1 deg celcius difference?I genuinely don't know, I have a 212 from almost 4 yrs ago? It works. It's quiet (for the time). The only complaint is if I try to go super small form factor, it won't work.
icrf - Monday, July 6, 2015 - link
If it were only $10 I might agree with you, but when it's half the price, and sometimes a whole lot less, it makes a lot of sense. I looked at the Noctua when I built my 5820 last fall, and couldn't justify the 2.5x price. For $35, the 212 EVO is a great cooler. As good as the Noctua is, it's not two and a half times as good. That's why the 212 is so popular. It's in the proverbial sweet spot.andrewaggb - Monday, July 6, 2015 - link
I agree. I've bought several 212 EVO's and I've been very happy with them. I was mostly looking for something that would run quiet under load (without overclocking) and I think they've been great. I've used some less expensive coolers and they were much worse - so in my opinion it's the cheapest cooler that met my needs.Araemo - Monday, July 6, 2015 - link
Obsession is pretty harsh given the facts... I went and looked up a comparison over on frostytech, and it looks like the Hyper 212 evo is only 2C hotter than the Noctua chosen as the realistic 'best choice'.. for 1/3 the price. Given that my ambient temps change by more than that 2C over the course of a year, 2C is never the stability margin I use on my overclocks.Nfarce - Monday, July 6, 2015 - link
Except that the 212 is not a "premium" cooler. When you start getting into extreme overclocks like I have (i5 4690k @4.8GHz, or a 23% overclock) and into water cooling needs territory, the 212 falls well more than a 2C behind which is where it is on lower level overclocks (5-15%) on my chip.StrangerGuy - Monday, July 6, 2015 - link
Who cares whether if the 212 isn't a "premium" cooler when I can simply buy the 4790K at stock 4.4GHz instead of premium cooling to barely OC a 4690K past a 4790K. You overclockers STILL think there is tremendous value to be had with OCing when the 2500K ship have long sailed.Nfarce - Wednesday, July 8, 2015 - link
According to my benchmark tests in games and applications like Sony Vegas Studio, my overclock to 4.8GHz yields quite a bit of performance increase over the stock turbo of 3.9GHz. Oh and since I was on a budget and game about 80% of the time on that 4690K rig, I justified saving the extra $100 over a 4790K and put it towards a better GPU solution.And yes, I still have a 2500K build as well (not sure what that has to do with the price of ketchup), which used to be overclocked to 4.6GHz on that NH-D14 cooler (it is now relegated to backup duties and running at stock speeds on a Zalman 9700LED cooler). Which, incidentally, roughly equals the performance of my Devil's Canyon chip running at 4.2GHz.
TheJian - Tuesday, July 7, 2015 - link
http://www.anandtech.com/show/6830/cpu-air-cooler-...It is far more impressive than you give it credit for. I easily hit >5ghz on this fan with i4790K.
Shadow7037932 - Tuesday, July 7, 2015 - link
If you're going to extreme OCing, why the hell aren't you on custom water cooling or Dice/LN2?Nfarce - Wednesday, July 8, 2015 - link
Because where I am in overclock, the best air keeps up with the best closed loop kit cooling for far less money.tabascosauz - Monday, July 6, 2015 - link
What farce said is true. At high voltages and overclocks, the 212 EVO breaks down and the dual towers begin to shine.DanNeely - Monday, July 6, 2015 - link
Which with simulated thermal loads ranging from 60-340W should have been made apparent in the course of testing.I would hope and expect that most if not all of these coolers would out perform it, especially at higher loads. But as a de-facto standard budget cooler for people who want something better than Intel's I think it should've been added to the matrix to show how much better these bigger ones performed. A stock Intel cooler should've been included as well for the same purposes (at least at the lower loads; no sense risking burning the test setup by trying to broil 340W through it). Including a mainstream reference point is especially important in this case because E. Fylladitakis's synthetic test load means that we can't cross reference his results with those found elsewhere.
SUpstone - Thursday, October 29, 2015 - link
Totally agree - to get the full picture and to aid comparability with other tests the reference points of the Intel stock cooler (free) and CM 212 EVO (£25) should be included.Flunk - Monday, July 6, 2015 - link
If you buy one of those things on sale I've seen them as low as $30, which if you don't need better cooling, is a good deal. The reason the Hyper 212 EVO is popular is that it's cheaper than most of it's competition and easily available. They're good enough for moderate overclocking on a 1150-series chip so they do fit the bill for a lot of people. Something being popular doesn't make it bad.Pastuch - Monday, July 6, 2015 - link
I have 2 of the original Hyper 212 (Non-evo) on two different I5-2500Ks that have been running super quiet at 4.4ghz for the last 4+ years. No coil whine or bearing degradation on the fans. I paid $20 for each Hyper 212 Evo. The value for the money is amazing. It's an excellent quality reliable product and it's easy to install.I paid $220 for my I5-2500k, 20$ for the Hyper 212 Evo, and $70 for 8gbs of PC 1333 in February of 2011. At 4.4ghz, it's still within 5% as fast as any CPU on the market. Sandy Bridge FOREVER! I'll keep buying video cards. You can waste your money on HSF upgrades for CPUs that become less important every day. DX12 is just going to make the CPU even less useful.
Pastuch - Monday, July 6, 2015 - link
Just to clarify, those are Canadian prices which makes them even more amazing. The new I5-K in Canada is almost $300! No thanks.northward - Monday, July 6, 2015 - link
Remember, the average exchange in 2011 was 1.011 (CAD to USD). It is presently 0.79 (CAD to USD). Assuming US/CAN price parity in 2011, that $220 cooler would cost ~$278, not that far behind the $284 i5 4690K.mr_tawan - Monday, July 6, 2015 - link
Used to have 212+ once. Later I swapped out for a cheap closed-loop. Though the CPU temp is a few C lower, the closed-loop was much louder than the 212+ (due the the 'pump whine').Years later I upgrade the rig to a Core i5, which is not really that hot, and I'm not interested in overclocking anymore (being more mature I guess).
I find the 212 is pretty good for its price. It's a great entry-level cooler for those who want to upgrade. I also think that it could serve well as a baseline for the comparision.
zodiacfml - Monday, July 6, 2015 - link
I didn't know that overclocking enthusiast would prefer a lower performing heatsink instead of the best available. The reason is simple; CPUs consume less power throughout the years even with continuous but non synthetic workloads including gaming.Many years ago, I was a fan of watercooling then big-air heatsinks then not anymore. It is just not logical anymore as they are more expensive, larger, and cumbersome.
kmmatney - Monday, July 6, 2015 - link
To me, the whole point of overclocking is to get a better cpu than what you paid for. So overpaying for a heat sink doesn't make sense - the whole point is to get the best possible performance, while spending the least amount of money. At least that is what overclocking means to me, and I'm sure a lot of other people as well.Ian Cutress - Monday, July 6, 2015 - link
That's usually how most people start with overclocking. For others, it's getting the best performance regardless. That's why people still that the i7-K and push it, rather than a Pentium-K and tweak it.kmmatney - Monday, July 6, 2015 - link
I bought my Hyper 212 for $19.99 - a much bigger savings than $10. It does the job, and in the end my overclock was not limited by temperature, but by the CPU itself. A more expensive heatsink wouldn't gain me anything.aj654987 - Wednesday, July 8, 2015 - link
What are you even talking about. The 212 is $35 and half the cost of many of these heatsinks. Its been the gold standard for years, if you only get another 1 C out of a HSF that cost double then its not worth it.CummingsSM - Thursday, July 9, 2015 - link
Yep. You save $9.99 and then this happens: https://i.imgur.com/COC5qW9.jpg(In case someone is wondering: No, I didn't over-torque it, the bolt got caught in the back-plate and sheared under the power of a screwdriver lightly applied; And yes, that bolt is hollow; And yes, that is the mounting hardware from a CM 212 EVO; And yes, I'm done buying CM products.)
LittleLeo - Thursday, July 9, 2015 - link
Or a Beer and a bag of chipstabascosauz - Monday, July 6, 2015 - link
Are you kidding me? Intel's CPUs might be efficient compared to AMD's, but there is hardly a valid reason to dismiss the dual-tower crowd. Intel's CPUs are hotter than they have been in years, thermal performance having declined steadily since Sandy Bridge due to sh*ttier and sh*ttier TIM and other reasons.rickon66 - Monday, July 6, 2015 - link
Regarding the CM 212+/EVO -They did not want to show a $25 cooler that beat the expensive guys.Drumsticks - Monday, July 6, 2015 - link
They did one of these round ups with the 212 Evo or + a while back, also involving high ends from Noctua (U14 and U12S I believe). They found that it doesn't quite match up, but I know it got a mention for exceptional performance for cheap. I think it falls behind more in noise than performance.It definitely would have been interesting to see it in here, but nevertheless, thanks for the review!
kmmatney - Monday, July 6, 2015 - link
I usually undervolt the fan a little - takes care iof any noise issues.Arnulf - Monday, July 6, 2015 - link
+1Nagorak - Monday, July 6, 2015 - link
I would have also liked to have a "decent" CPU cooler like that included, as well as the stock Intel/AMD HSF. It's great seeing how these coolers stack up to one another, but it doesn't truly quantify how much of an improvement you're getting over a cheap alternative, or the stock fan. For the record, I haven't run with a stock fan on any main PC I've owned in the past 15 years, but I would be curious to see how much I'm actually gaining.Araemo - Monday, July 6, 2015 - link
Especially given the raw value of the Hyper 212 Evo at $30, it may get within a degree or two of some of these for half the price or less... which is why my last build had the Hyper 212+ (It's been a few years) - I could have gained maybe 5C by spending 4 times as much.. which didn't seem worth it to me.TheJian - Tuesday, July 7, 2015 - link
AGREED. I have one, and at $29.95 from newegg just a few months ago on sale it was an AWESOME deal. i4790k can do massive oc's with it and even at full load is not terribly annoying with my 5850 causing most noise when gaming. This is still a top seller and for good reason.LittleLeo - Thursday, July 9, 2015 - link
Since its about the most popular cooler for gamers that would have been nice.jay401 - Monday, July 6, 2015 - link
I'm actually really glad to see this article, it's been ages since I've seen a good CPU air cooler roundup and sockets have changed several times over the years so it's nice to know what works well these days.jmke - Monday, July 6, 2015 - link
air cooling has plenty much run into a wall; heatpipes to copper base, aluminum fins on the heatpipes, put 140mm or 120mm fan... there is not a lot of wiggle room, so performance of those that follow this recipe is very close.differentiators now for most part are: socket compatibility, price, installation method. Raw performance/noise is no longer the focus imho if you want a successful product
meacupla - Monday, July 6, 2015 - link
It's not so much socket compatibility, so much as how compatible you can make your heatsink against mobos that have poor design choices.Although not as common on mATX and larger boards, mITX suffers a lot from this, because manufacturers attach fragile bits onto the back of the mobo, near the CPU socket, that interfere with the mounting bracket. Either that, or the CPU socket is placed too close to the PCIe, etc.
That Reeven Okeanos is something I haven't seen since Athlon 64 days, which are heatsinks paired with a stupidly loud fan. Look, if I wanted a heatsink with stupidly loud fan, I would buy an amazing heatsink or watercooler first, then attach the stupidly loud fan to that, instead of some mediocre heatsink with a mediocre fan.
MartenKL - Monday, July 6, 2015 - link
I would of course like to see the testbed updated to have fan/s controlled by thermals, ie something like ASUS Fan Expert. Set for a target temperature and loads in the more realistic range of 15-150 watts. And of course when reducing voltage it should be via PWM and not simply reducing static voltage. The results should then be presented in temperature variance and noise levels/profile.MrSpadge - Monday, July 6, 2015 - link
Good suggestions. I hate those charts with "full fan voltage" and "fan voltage reduced to ..". What I really care about is "how silent can the cooler be for a given temperature / cooling performance?" And "which one cools better at similar noise level".It doesn't help much to see a strong fan with inbearable noise in those charts. Even if someone is interested in such solutions - wouldn't his question rather be "which heatsink performs best with this high-speed fan"? Which would again be something he couldn't answer from this data.
I know making noise based comparisons is difficult. But the raw sound pressure could be accompanied by some subjective remarks regarding the noise spectrum.
Cookiespy - Monday, July 6, 2015 - link
It would be interesting to see how the stock coolers compare to this high performance cooler. I wouldn't pay $80-100 just to see 5degrees improvement.Eidigean - Monday, July 6, 2015 - link
Chips big enough to need these coolers, such as Socket 2011, do not come with stock coolers.meacupla - Monday, July 6, 2015 - link
The stock heatsink cools great and is pretty silent with stock settings in a case with decent airflow, end of story.These kinds of $80 heatsinks are what you want when you overclock, but with the same or lower noise levels.
If you don't overclock, then a $20~25 heatsink can do a 5~20C improvement and keep the computer quieter at the same time too.
Eidigean - Monday, July 6, 2015 - link
What are you, a shill for Intel? The Intel stock heatsinks are the absolute worst. Check out the graphs from this Anandtech article:http://www.anandtech.com/show/6830/cpu-air-cooler-...
Dead last in performance AND noise. Stock heatsink was greater than 30 degrees C hotter, and 20 decibels louder.
meacupla - Monday, July 6, 2015 - link
It says right there in the system specs used to test those coolers..."Intel Core i7-2700K overclocked to 4.4GHz @ 1.4V"
I'm surprised the stock intel heatsink was able to complete the tests.
meacupla - Monday, July 6, 2015 - link
no wait, look, it says the stock heatsink and a low profile heatsink failed on an overclocked i7.Like I said, stock intel heatsink, especially the one with a copper core, works great at stock speeds.
Azurael - Friday, July 17, 2015 - link
I don't know if the stock HSFs have changed since Sandy Bridge, but my 2500k would hit 98+ degrees and throttle under AVX loads with the pathetic little stock thing whilst sounding like a small tornado had developed inside my case (how is it that a 95w CPU comes with an HSF half the size of the ones they used to ship with 65w C2Ds?!)Whichever cheap tower cooler I replaced it with does the job just fine, though. It's been running like a champ at 4.5GHz for near enough 4 years now. (I think it's a Xigmatek SD1283 - I haven't even taken the side off my machine for over a year, those heady days of tinkering and yearly upgrades long since passed.)
bug77 - Monday, July 6, 2015 - link
What, no mention of the weight of each cooler? I think that's a rather important aspect.AssBall - Tuesday, July 7, 2015 - link
No it isn't. The mounting mechanisms on these higher priced sinks are, as, had you read the article, solid. If you are missing an arm and wanted to mount a 100$ heat sink in your 25$ case, you might care about the weight, in which case, the manufacturer's website is a one armed lazy click away.Schickenipple - Monday, July 6, 2015 - link
Cheaper still, at under $50, is the ZALMAN CNPS9500A. I use it on almost every build due to it's all-copper design and razor thin fins. I can keep the fan on the lowest speed without my Core i7 going over 40 Deg. C under load. It's silent no matter what I'm doing.My only complaint is that the fan is not PWM. Quibbles.
meacupla - Monday, July 6, 2015 - link
The only thing I haven't liked about Zalman, is that the fans they use in their heatsinks aren't exactly the quietest and are usually not easily swappable. At least, not when compared against manufacturers like Scythe and Noctua.Nfarce - Monday, July 6, 2015 - link
I have the 9700LED variant of that design for my old Core 2 Duo E8400 build. It does okay for medium overclocks (running my 3.0GHz E8400 at 3.4GHz), but that was about it. And considering I paid $55 for it back in 2009 when I built that rig (about $60 in today's money), it wasn't exactly a cheap option. I would have spent a little more for a better cooler so I could get a higher overclock, but the new Sandy Bridge chipset was coming out soon and I decided to just keep it as a backup rig.Eidigean - Monday, July 6, 2015 - link
The conversion from Euro to Dollar for the Reeven Okeanos is incorrect. 60€ != $54. 60€ == $66. I think you divided instead of multiplied.GeekTech - Monday, July 6, 2015 - link
I just want to mention that the Cryorig R1 Ultimate CPU Cooler that you listed and said was currently available only through a foreign store registered in Amazon.com that ships from Korea is currently being sold at PC Case Gear here in Australia for AU$89 (US$66.82).Link: https://www.pccasegear.com/index.php?main_page=pro...
letmepicyou - Monday, July 6, 2015 - link
I would have liked to see the Noctua's real arch nemesis included in this shootout, the Thermalright Silver Arrow. Crushes the Macho Zero. I have a Silver Arrow on my i7 4790k running seti@home, and it owns.Innokentij - Tuesday, July 7, 2015 - link
The Phanteks PH-TC14PЕ is a better cooler using noctunas own fans on it and even better if u add AF140 fans to it . Is why i took it over the Noctua for myself.http://www.xbitlabs.com/articles/coolers/display/n...
UltraWide - Monday, July 6, 2015 - link
There is an updated version of the noctua: http://www.noctua.at/main.php?show=productview&...Eidigean - Monday, July 6, 2015 - link
It's not so much as updated, as it is an off-center version of the same NH-D15. I would need the NH-D15S that you linked to on my MB as the former would hit my GPU in the first slot. I was considering modifying the D15 to move it up and away until I saw the D15S. A good link none the less.Peichen - Monday, July 6, 2015 - link
I expect many R9 Fury X users are shopping for large air-cooler now as they have to give up their AIO CPU-cooler to Fury.BTW, I really wish you added a Hyper 212X as reference.
meacupla - Monday, July 6, 2015 - link
You could just buy a case like the Fractal Design Define S, which could fit two separate 120mm CLLC in the front, allowing both to get fresh air.Flunk - Monday, July 6, 2015 - link
I'm a bit disappointed that you didn't comment on the install process on any of the heatsinks. I recently bought a Dark Rock Pro 3 and while I love how quiet it is and the temps are actually lower than the Corsair Hydro H80 it replaced, the install process requires you to screw the heatsink in from the back of the motherboard. That and the size of the supplied backplate made the heatsink install more difficult that is really necessary.If you buy a Dark Rock 3 Pro I recommend removing the motherboard from the case entirely and installing it by flipping the heatsink upside down and balancing the motherboard on top of it in the correct position. This makes it fairly easy to screw in. But if you are using a normal thermal paste you might need to put it on the heatsink instead of the CPU heatspreader. I use IC Diamond and that stuff is so thick that it just stuck there upside down for long enough to finish mounting the thing.
meacupla - Monday, July 6, 2015 - link
Yeah, this is actually quite important. Noctua's mounting brackets are, by far, one of the easiest to work with.'nar - Tuesday, July 7, 2015 - link
I never gave it much thought. Installation is such a small concern to me, maybe I do this more than most, price and performance are preferable. That said, I still think Noctua's mounting can be improved, it seems unnecessarily complicated to me. First off, you really do need to replace the plastic bracket, there's no way around that. But secondly, why include a 140mm screw driver? Why not make the screws 140mm taller? Then you can just use a common screw driver, even a stubby or a pocket knife. And make them captive so they do not fall out and you do not need to line them up. These will certainly add to the cost due to extra engineering time and unique screws.Beany2013 - Wednesday, July 8, 2015 - link
Because the 140mm screwdriver:A: Is cheaper than re-engineering the entire product
B: long screws are *very* easy to cross-thread due to the extra sideways torque you can apply when inserting them.
I *do* like the Noctua setup system. It's strong, comprehensive, and lets be honest, you only do it once. I'm pretty sure that any gotchas with installation were caught in the descriptions of each cooler, too.
der - Monday, July 6, 2015 - link
WOOOO!golemB - Monday, July 6, 2015 - link
For the Macho Zero especially, I'd want to see the tests conducted (additionally) in a vertical motherboard orientation (as you'd have in most tower cases), since convection may have an effect on performance. It may also reveal differences in fan noise due to bearings rubbing more or less on different surfaces.'nar - Tuesday, July 7, 2015 - link
Not likely. Convection is slow. Any fans will blow away convection currents. Besides, orientation is strictly a "case by case" basis and beyond the scope of an empirical HSF comparison.Fan noise due to orientation may be good to check for though. I doubt it will be any different, if it is, the aberration should be noted.
flashbacck - Monday, July 6, 2015 - link
wow, cpu cooler roundup! It seems so rare to see these nowadays.marraco - Monday, July 6, 2015 - link
Something really important that never is reviewed is dust control and maintenance.As radiator fins get closer, they have more dissipation area, but air flow gets worse. Yet the main problem of fins closeness is that they accumulate dust much faster, and get occluded.
No review makes a dust test. I have a Thermaltake cooler, which works excellent when is clean, but it rapidly loses his capacity due to being occluded by dust.
The easiest way to remove his dust is to use canned air, but it is a short lived fix, because an air cleaned radiator occludes itself very fast, sometimes in matter of weeks.
The only way to clean it for good is to take out the dissipator, and put in in the dish washer. But that is a cumbersome task. I need to do a lot of work just to clean it, and I need thermal paste to place it again on the motherboard.
So, maintenance is as important as cooling and noise.
A good cooler should pass a dusting test (being exposed to a day of dirty and dusty air current), and should be possible to clean it without much hassle, without applying thermal paste, without using screws, without need to access both sides of the motherboard, and without his pegs breaking or being degraded by manipulation.
'nar - Tuesday, July 7, 2015 - link
Interesting concern. It makes sense, yet is never considered. I think the best answer is the same as electricity, you need to provide the best environment you can for your computer. Higher quality components are more sensitive to dirty power, and dusty air. Get that PC off the floor, and get a case with good dust filters. And if your room is bad enough, get an room air filter. I have one on my desk right in front of my PC. A Honeywell HEPA air filter.Impulses - Monday, July 6, 2015 - link
I understand why cases are tested with stock fans, replacing 3+ quality fans will significantly alter the price equation... It's a trivial difference for $50-100 heatsinks tho, many are often even sold sans fans anyway.If nothing else, a single apples to apples test with all running the same fan would've been welcome. In addition to something like the Hyper 212 as a baseline, the different Thermalright TRUE Spirit variants have remained a solid value over the years (they perform better than the 212 and even closer to some of these as per HardOCP's last roundup).
Is there gonna be a midrange roundup? Spending upwards of $65 on an air cooler never made much sense to me when more affordable options were so close in both noise and performance. Out of 9 coolers tested only 3 or so come in at a sensible price, at $75+ wouldn't it make more sense to go AIO WC?
xthetenth - Monday, July 6, 2015 - link
If you want low noise, good thermals and dead quiet idle well into the land of diminishing returns, high end air is still a very compelling alternative to water, especially if you're willing to put the effort into some ducting.However midrange products can be excellent and are well worth a good long look before hanging a hundred bucks off your socket.
'nar - Tuesday, July 7, 2015 - link
I was a HSF fanboy until I actually needed a water cooling system. Specs and testing does not show response time. Fast, transient loads can overwhelm a HSF, as heat pipes cannot transfer heat as fast as water. For low power CPU's a HSF is fine, but when you go over 100 watts the noise is less of an issue, as a HSF will need much more airflow to compare to the higher heat transfer ability of W/C. And instantaneous loads are more likely to cause system instability due to the less efficient heat transfer rates of heatpipes.meacupla - Tuesday, July 7, 2015 - link
I am going to nitpick on what you're saying there, because you have it confused.Heatpipes use some form of vaporized liquid inside and they actually transmit heat quite good, especially when compared against sold copper rods. Where they fail, is that they do not have a lot of capacity for transferring large amounts of heat.
Water is actually quite poor at transmitting heat, as it's a non-metal and an insulator. It does, however, have great capacity to hold heat.
This is why large bodies of water are warm during winter and cold during summer and results in mild weather near the coast and extreme weather in the interior.
The advantage of water cooling systems, over air cooling, is the ability to place large radiators, with a lot of surface area, in spots that get fresh air from outside the case, instead of warmed up air that is already inside the case. Heatsinks are also limited to size and weight constraints around the socket.
Were it possible to attach a triple 120mm fan heatpipe heatsink, as in parallel layout, instead of serial dual towers, directly to the CPU, I'm quite sure the results would be similar to a 360mm radiator water cooling setup.
xthetenth - Tuesday, July 7, 2015 - link
Yeah, he's describing the thermal mass of the coolant, not any transfer capability. It seems that running coolant through radiator fins does have an advantage of relying on the fins to radiate the heat out from a heatpipe, but for a similar surface area, it's questionable how large the advantage is, especially at CPU TDPs.GPUs on the other hand have a more constrained form factor for large air cooling and a higher TDP, so they are a more promising place for CLCs.
PitneFor - Monday, July 6, 2015 - link
wheres Prolimatech, my precious...mejobloggs - Monday, July 6, 2015 - link
I'd be very interested to see a size-per-performance chart.I usually buy coolers that perform the best without being too large. Although I guess "Top Tier" coolers isn't the right article to look for smaller size coolers :p
Impulses - Monday, July 6, 2015 - link
Leaving it to the manufacturers to select their entry is always messy or questionable IMO, TR choosing the Macho over the Silver Arrow for one... Still curious whether there's a midrange round up planned or if this is it for air coolers at AT for a couple more years...Despite the criticism I do appreciate E Fyll's thorough process, otherwise I imagine I and many others wouldn't even bother commenting. Low end (212+) and high end ($120 AIO?) base lines would make the article much more useful tho.
The into even suggests some people prefer high end air over WC at the start but then leaves that up in the air... No pun intended.
'nar - Tuesday, July 7, 2015 - link
I know right? I was so looking forward to a W/C vs A/C comparison after an intro like that.Margalus - Monday, July 6, 2015 - link
why didn't you test these all with the same fan? Then we could see how the cooler performed independent of the included fans? Like the Thermalright, it comes with no fan, apparently you got a super quiet slow fan and put on there, but that isn't fair to Thermalright saying they are hotter, when it could be because it doesn't have the amount of air moving over the fins as others.Beany2013 - Wednesday, July 8, 2015 - link
Thermalright provided the fan, so they can't grumble at the results.trandoanhung1991 - Monday, July 6, 2015 - link
When you talk about ultimate cooling, you should've at least tested the True Spirit 140 Power Edition with a TY-143 fan, or the Silver Arrow SB-E Extreme. Those are the most interesting products from Thermalright, not the Macho Zero.Maybe as an addendum at some point? I'm very interested to see how the Silver Arrow and the 140 PE fare against the D15.
Impulses - Tuesday, July 7, 2015 - link
Apparently the Macho is what TR themselves chose to send AT, if the intro is accurate... Big /facepalm on their part. They probably have some of the best value coolers in the TRUE Spirits (I have the original Cogage version myself), and the Silver Arrow might've ranked up there with the Noctua and Phantek. The article did make me pretty curious about the latter tho, call me vain but the color choices are cool.Calculatron - Monday, July 6, 2015 - link
Thanks for doing a heatsink round-up. They are refreshing to see these days.It is a shame that Thermalright did not send in its top-tier performer. Then again, the Macho Zero is nothing to sneeze at. ~40C over ambient for a 340 watt load is still a good result. (Perhaps, instead, they could have thrown the TY143 performance fan instead? Har!)
siberus - Tuesday, July 7, 2015 - link
I actually wish they would have sent 2 of the current fans so we could see if push/pull could push it up a performance bracket.rrohbeck - Tuesday, July 7, 2015 - link
I'd like to see the measurements all with the same fan(s) - whatever is considered "the best" fan. That would give an indication of how much you could get out of the cooler with aftermarket fans.'nar - Tuesday, July 7, 2015 - link
I wish there was more review cross-over with water coolers, these two camps seem to be at odds with each other. They never seem to be compared effectively to each other, so it is difficult for consumers to determine the "best" cooler for themselves. With Noctua getting up to $93, there are water coolers out there for less. I bought my Noctua NH-D14 for $75 and thought that was high for a HSF.I strictly used air coolers until I got an AMD APU, among them are several Noctua models. It was apparent to me that this CPU, after a bit of easy O/C, got hot much too fast for an air cooler to absorb. It would crash after just 4 seconds of starting Tomb Raider and the cooling fins were still ambient temperature. I tried three coolers including the Noctua NH-D14. Fan speed did not matter as the rapid increase in temperature exceeded the heatsinks' ability to draw the heat off the CPU itself. I would guess that it take anywhere from 30 seconds to a minute for the heat to actually get to the fins, so if your heat sink cannot "sink" the heat all by itself, no fan, for a full minute, then it has inadequate heat transfer and no fan will fix that.
I installed a Corsair H100i and that works very well. I had previously thought that any cooler with less surface area would have less cooling performance, but I have found that if you cannot transfer the heat to the fins, they make no difference. I think a Corsair H60 would have been fine now. I heard that water coolers were "better" and more efficient, but nobody ever explains WHY.
From this experience at least, it appears that water coolers have better heat transfer performance. Fan speeds and fins are secondary to that, as they do not matter until the heat gets to them. If they get hot, then low speed fans can easily remove that heat as higher temperature differentials generally allow for greater heat transfer. If you run high-power and high-heat for a long time, then higher fan speeds help.
How quickly can your test bench ramp up in power? Was that tested? Was that considered? CPU's can hit maximum power in nanoseconds, and crash in milliseconds. Only the base of the HSF would see anything from that event. I think this test is more academic, and not very relevant in the real-world with actual CPU's. It only tests for maximum heat generation over time, like when running benchmarks, not the dynamic nature in which CPU's operate for most useful loads. But then, that's just my perspective.
Pissedoffyouth - Tuesday, July 7, 2015 - link
I agree with you. I think the heat transfer through heatpipes takes quite some time to get to the finszodiacfml - Tuesday, July 7, 2015 - link
Someone should test for this. It is quite plausible as it will take time to increase the temperature of water/coolant in the system. Yet, for longer and continuous loads, surface area matters to dump heat in the environment which AIO coolers has less to a similarly priced Big Air cooler which shows in benchmarks.Navvie - Tuesday, July 7, 2015 - link
Pretty good.It'd be nice if there was a base line (say intel's current stock cooler) and an AIO water cooler to show the difference between a top tier air cooler and what many consider to be the next step.
One point, SPCR always test each cooler with a 'reference' fan, making it easier to compare the heatsink's efficiency. Can't help but feel you didn't complete the task you set out to by levelling the playing field with your simulated CPU heat source.
maximumGPU - Tuesday, July 7, 2015 - link
Great roundup! i for one still believe high end air a good alternative to water if you want dead silence.What i still can't understand is how can Nocuta get away with shipping the most expensive air cooler, likely to be used by enthusiasts who care about aesthetics, with the most hideous fan colours?
'nar - Tuesday, July 7, 2015 - link
Noctua colors are odd, but I think it forces you to consider performance more than aesthetics. They do look nice, but not "cool" - no pun in tended.Dead silence basically depends on how much heat you need to dissipate, temperature of ambient air, surface area of fins/radiator, air flow, and heat transfer capability of the cooler. Or more basically, how easily you can transfer heat from CPU to the air. Water increases the efficiency of heat transfer, so once you reach a certain power level they will overtake HSF's. They will never be as quite as a good HSF, at least not on low-power CPU's that arguably do not need anything but the stock cooler to begin with.
xthetenth - Tuesday, July 7, 2015 - link
Noctua gets away with those colors entirely because they're Noctua. Yes they're not pretty colors, but unlike nice bright/garish colors they are an unmistakable sign that they're the best or at least tied for best.piroroadkill - Wednesday, July 8, 2015 - link
Your keyboard seems to have a problem, you've typed "hideous" when clearly "fantastic" would be better. Noctua fans are neat.Beaver M. - Wednesday, July 15, 2015 - link
Well, serious people are more likely to buy Noctua anyway. And no serious user cares about what the fans look like, and even if they did, they would still take the Noctua, since you can CLEARLY see the extremely high quality they are just by looking at them. Windows in cases and other stupid bling bling is for the PC-ricer gang. They only choose after looks anyway, and dont care if their fans make noises from day one.PrinceGaz - Tuesday, July 7, 2015 - link
Excellent review of what seem to be very good HSFs throughout (but then again, you did ask them to send their best).A comparison with the stock Intel cooler as supplied with something like the 4790K would be nice, as that is what any expense on a third-party cooler must be compared against. How would the stock-cooler compare at 340W? I was very impressed with the degC/W figures for all those coolers; it almost makes 340W seem an amount of heat that can be dealt with, though I'm thinking it's crazy.
As well as the stock-cooler from a highish end CPU, one or two commonly used AIO liquid-coolers needs to be added to provided a second comparison point, as these high-end air-coolers are presumably expected to be near, possibly even better than a liquid AIO.
I love the fact you are using a properly designed heating rig, and your comprehensive review, but I came away thinking: they're good! But I've no idea how much better they are than the stock cooler, or how any of them would compare with the liquid AIOs also available, both in terms of cooling or noise.
Sivar - Tuesday, July 7, 2015 - link
This is probably the best CPU cooler article I've read (and I read such articles when Anandtech when it was on Geocities).Excellent writing, useful measurements, and a cogent conclusion. Thank you, Emmanouil!
orangesky - Tuesday, July 7, 2015 - link
SilentPcReview just posted a review of the new Scythe Ninja 4: http://www.silentpcreview.com/article1462-page1.ht... The review includes some good comparison tables with many of the popular air & water coolers.Sounds like a pretty decent option, and probably the best Ninja since the original.
jenesuispasbavard - Tuesday, July 7, 2015 - link
I've had great performance with the Spire Thermax Eclipse II for five years now, and for the last three years, even better performance when paired with two Corsair SP120's. It's still the highest rated heatsink on Frostytech; not sure why it isn't included here.jenesuispasbavard - Tuesday, July 7, 2015 - link
Edit: Maybe too old? Released in mid-2010 I think.Iketh - Tuesday, July 7, 2015 - link
did I miss the thermal medium? what paste is used? I've looked over the article twiceBeany2013 - Wednesday, July 8, 2015 - link
Whatever was supplied - it's a test of the product as supplied (for most of them) or as recommended by the manufacturer (as in the TR one, where they provided a suitable fan when requested, as the rest came with one).Thermal pastes are best tested on *one* cooled as a known quantity, really, which would be a separate test of it's own, I'd say.
Iketh - Tuesday, July 7, 2015 - link
also, EVGA's cpu ACX cooler should be in a cheaper roundupcreed3020 - Wednesday, July 8, 2015 - link
Thanks for the review! This is certainly your best review since joining AT, kudos to getting a good spread of products from around the world. There were a few in there that I hadn't heard of so it was great that the final results would be a surprise.The results were somewhat what I expected and nice to see where each of the companies's products sit. I have to say that not having the Cyrorig on the final recommendation is a bit sad as their products are simply bespoke. I don't have the R1 Ultimate but I do use their C1 in my SFF case on an overlocked Core i5 4690k and the temperature to noise performance is unbelievable, especially compared to some SFF coolers I've used in the past from other vendors.
Looking forward to more of these, perhaps a SFF cooler showdown could be next...?
JimmiG - Wednesday, July 8, 2015 - link
I went with the Phanteks PH-TC14PE back when I built my Haswell system back in 2013. It and my (de-lidded) 4770K are still holding up fine.All those coolers are really close though, especially at full fan speeds. We've probably hit the limit of air cooling with heatpipe technology at this point.
PPB - Thursday, July 9, 2015 - link
Why review the Macho Zero and not the True Spirit 140 Power or the IB-E Extreme? Those 2 are far a better match for the rest of the lineup that the one you reviewed from Thermalright.katinacooker - Saturday, July 11, 2015 - link
Cryorig stock is availble in the UK from Alternatehttps://www.alternate.co.uk/html/search.html?query...
I got my R1 Universal from there
DPOverLord - Saturday, July 11, 2015 - link
You can't do a real air cooling review without the Thermalright Silver Arrow, this review is lacking.kraznal - Thursday, July 16, 2015 - link
LOL - "Liquid-based cooling solutions are becoming easier to install and AIO kits generally are hassle-free, yet they are still not favored by the majority of the users. Their space requirements, increased complexity and price hold most people to simple air-based cooling solutions."so they become easier to install, hassle-free but somehow managed to keep their complexity? what are talking about??? :)
anyway, I am sure you know this and you just overlooked this fact to support your own story, liquid cooling sets are NOT more expensive (Corsair H50 costs $60 shipped), are NOT bigger (H50 compared to any of the air coolers here for example), and are NOT complex (whatever you mean by that), or even the noise factor where air coolers need to run on higher RPMs to achieve same cooling effectiveness as liquid coolers - simply there is no comparison between liquid solution and air solution. Cheap Corsair H50 is far superior in every way than those colossal monstrosities you are reviewing here now. Smart user would never chose an air cooler simply because it doesn't make sense to chose an old and less effective idea.
Therefore please do not spread bullshit just so you can cash a check for an article.
rleigh - Saturday, July 25, 2015 - link
Liquid cooling doesn't always work well. I got a Corsair H60 to put in a Corsair Obsidian case with an ASUS Sabertooth R2.0 mainboard and an AMD FX8350 processor. It worked brilliantly with CPU temperature at ~35C under load. Unfortuately, the layout of the case and lack of airflow around the CPU heatsink lead to the VRMs/MOSFETS around the CPU reaching temperatures of over 85C, no matter how I arranged the case fans. I replaced it with a Noctua cooler similar to this one which also kept the CPU nice and cool, but with the side effect of the large 14mm fans producing sufficient airflow onto the mainboard to eliminate the dead space which caused overheating.This isn't to say that closed loop coolers are bad; their performance can be very good. But they aren't compatible with every case/mainboard.
kraznal - Wednesday, August 5, 2015 - link
First: it's been almost a month - E. Fylladitakis - must be on vacations since he has yet to reply to my comments.Second: rleigh - very sorry to hear that you had to install additional cooling for your motherboard. It is best to keep high air flow inside the case, if you didn't provide that then no wonder you had an overheating problem.
Cvengr - Friday, December 25, 2015 - link
http://serverfault.com/questions/263931/why-datace...Here's an interesting link regarding the use of water cooling in data centers (limited to air cooling). Primary issue is safety.
alexbagi - Monday, March 7, 2016 - link
Good picks. I'd also suggest looking at http://www.144hzmonitors.com/cpu-cooler-buyers-gui... for guidance.I am going with the 110i myself, as I need water cooling.