https://translate.google.co.uk/translate?sl=sv&tl=en&u=https%3A%2F%2Fm.sweclockers.com%2Ftest%2F27263-noctua-nh-u12a-flaggskeppskylare-med-imponerande-prestanda%2F5

Impressive results in this review. About 2,5 db less at the same thermals as d15.
And same thermal performance at max rpm but still 2,5 db lower noise.
A12x25 also have better subjective noise at the same db as A15.

We have a new reference cpu-cooler for low noise enthusiasts!

So.... i finished, kindoff, i did all the testing, aside from the hearing session, sadly both days had rain in the night making not so quiet or at least the levels that i wanted to test, so maybe some other day ill do it, for now its postpone, as i cant do today nor a weekday because of work, but will see.

First let me say, all the coolers are really good coolers, its not like there is one that its not worth it or that loses severely, all of these coolers will cool most of today's CPUs fine. But do remember, the testing has been done on a delided i7 8086k with custom ihs + liquid metal, this dropped its load max temps around 15C, also take into account that the CPU is only running @ 4.3hz on Aida64, this is intentional or i wouldn't have been able to test lower rpms to compare between fans/heatsinks. So lets go with each cooler into what they excel into,

Noctua NH-D15s
Imo its the best cooler that was tested, but best in what... well its the one that netted the lowest temps and if you use two fans, it can even cool at very low rpms very well. While its an expensive cooler as it is, its very versatile in terms that it can do low and high rpms, downsides its size and weight, in some scenarios you wont be able to mount the frontal fans, specially if you don't go with standard size ram, still this is my top pick of the bunch.

Noctua NH-U12A
I was disappointed at first, as it struggle with low rpm, but its not meant for it, if you see the charts on high rpms it matches or gets close to the NH-D15s on 50/100% PWM, given that the rpms are different, the noise imo its comparable (take it with a grain of salt). The cooler does need the two fans, specially at low rpms, with a single fan was the only cooler that didnt finish the 30min min test out of me fearing it will hit 100c+. The problem with it is that its expensive, more than NH-D15s and HR22, but i do think it has a place/use, for a smaller cooler it cools well on higher rpms, and the Mugen5 cant touch it here, not even with the same fans, so if i was to air cool a 8700k/9900k or alike, and i couldnt fit a NH-d15s or dont like big coolers, then this is as good as it gets.

Thermalright HR22
This is a cooler that many avoid due to its size, but thats whats gives its place in the market, its by far the best low rpm performer of the coolers that were tested, specially with 1 fan win everything, but the NH-D15s performs well here with 2 fans. As with the Mugen5, the HR22 is a widly spread fin design allows this low airflow efficency, but gets better because of its size, it outpeforms the Mugen easily, so you are very likely to be able to plan a fan curve much lower than any other cooler that was tesed, the downside are two, size/weight and the price.... the LeGrandMacho is as expensive as the NH-D15 (none s, that comes with 2 fans) and the chart shows with 2 fans it outperforms on low rpm operation, but as with the Mugen5, it has diminishing returns on high rpm operation, while the NH-D15 perform very well here. So its a niche product, this cooler imo should be priced at $60 to be called the best for us, but at $80 is hard to recommend over the NH-D15, now if you are like me and want the S version because of the PCIe slots, then HR22 picks up its value, as you will need to add another NF-A15 to perform similar to the HR22, taking it to $100, more if you dislike Noctua colors and end up with Chroma NF-A15, where the LeGrandMacho sits at $80 and comes with an attractive and good performing fan. For people wanting more cooling than what the Mugen5 gives and want to remain on low rpms and cant use a twin tower with twin fans (like none standard height memory), this is as good as it gets.

Sythe Mugen5 RevB
I'm going to call this the jewl of the test, it didn't win any but did well in all, and it cost almost half of all the coolers that were tested, and it will remain my top recommendation for SPCR users for its performance/value. The Mugen5 is a cooler that its not meant for high rpm fans, its fin spacing is wider than U12A, more similar to a mini HR22, and thus the test reflect this, on low rpms it wins over the NH-U12A (true on 1 and 2 fan), but once it goes into above 1200rpms it struggles and U12A performs better. But you see, Scythe included a 1300rpm for this reason, there are diminishing return on using a high rpm fan on this cooler, and shines a lot on its range of operation. It was interesting to see how well it perform on the 15% compared to U12A, but the same can be said where it pulls ahead U12A on 2100rpms, but because of what we try to achieve here on SPCR, the Mugen5 imo its a better buy, not only because of its price, but also we are very unlikely to use a cooler above 1200rpms. Now if price is not a concern, then both the HR22 and NH-D15s perform much better, this leads me to the begging, its a great cooler that perform well in most tests, specially at low rpms, but what makes it so good its price/performance.

Is it worth buying the NF-A12x25 for the Mugen5? depends, initially i would say no, the Kamaflex performs well and the range of operation of the fan matches the ideal operation of the Mugen5 design, remember the only thing that the Mugen5 wins is on value, adding a $30 fan will place it in the NH-D15/HR22 price range. I would recommend to start with the kamaFlex, see from there if you want a quieter fan (NF-A12x25 under the same rpms is quieter, to what i remember around 100rpm or so nets same noise subjectively), another scenario would be in case you want lower rpm operation, the kamaflex can't be stopped even at 0% PWM it nets 400rpms (at least on gigabyte PWM bios fan control) so if you wish to stop the CPU fan under certain temps then get the NF-A12x25, or another scenario that i would suggest is if your kamaflex is showing its age and you want to change it, i would make the effort toward the NF-A12x25.

Awesome tests and analysis. Thanks for doing the work

I didn't expected the Mugen 5 doing that well at low speed, and even less the KamaFlex outperforming the A12X25 at low speed. Thanks you for the awesome work and your time

EDIT: You said on the A12X25 thread:

So even if the Mugen 5 achieves lower temps at lower speed, if we ever did a curve with the noise vs temps the A12X25 has much more dynamic in term of min/max cooling at lower noise than the Kama Flex. I think I made my choice, thanks again for all your work.

my kama flex fan that was included with the mugen 5 started (or maybe always had) a little bit of bearing/motor noise. it was pretty unobtrusive and most people probably wouldn't notice it, but it bugged me enough to slap an a12x25 on there.

thanks for testing. looks like I’m better off delidding my 8700k than swapping my heatsink if I want better thermals. I’d probably get the u12a over the mugen if I didn’t already have it, but that’s mostly because I love those fans. I did briefly have a mobo the used the first pci-e slot where there was conflict with the mugen’s fan clip and the gpu backplate, though some zip ties worked well enough.

next time you find yourself bored when it’s quiet, let us know how you think the a15 amd a12 compare!

Abula, your guess...

Which is better on performance per quality of noise, Noctua NH-U12A or Scythe Ninja 5 with a couple of Noctua NF-A12x25 PWM in replacement?

I think the Ninja5 would perform in between the Mugen5 and HR22, since its a bigger cooler than the Mugen5 but not as big as the HR22, spacing on the Ninja4 was similar (not sure on Ninja5), now value wise, i dont think its worth upgrading the fan, it will cost your almost twice as much very near what the U12A, which is even more than the HR22. The issues to chose a cooler with an unlimited budget and dont care what you spend,
1) If you are cooling something extremly hot like the 9900k i would go NH-D15s, on load you wont be able to drop the rpms, the CPU on high load will need the cooling, the U12A would be similar, but you lose the low rpm performance.
2) If you are cooling something easier like a delided 8700k then i would go with the mugen5, you will be able to drop the rpms since it doesnt need as much cooling and you will be spending a lot less. The ninja should be slightly better, but i dont think it will reach HR22 performance on low rpms. But upgrading the fan on either will push them above the NH-D15S which to me its the best cooler, so i dont see the point, specially thinking that Ninja5 overlaps with memory so there is no reason the NH-D15s will not fit there.

@lb_felipe while i was checking other stuff, i stumble on a very nice review on youtube, that has lot of things that i like to see on reviews, specially noise/temperature graphs, take them with a grain of salt as probably him as me, dont have an Anechoic chamber, that said it does look pretty impressive quiet cooler,

ponchato The QUIETEST Cooler I Have EVER Used!

The only thing that i dont understand, is from what i know, they used the same fan as the Mugen5, and two fans will be louder than one, so why is the noise graph reflecting the ninja5 as quieter than mugen5, to the graph should have shown a cooler running ninja5 but the noise should be slightly higher than the mugen5, then it could been error on his part or he didnt ramp the fans, who knows. Either way, seems like a very nice cooler, im probably going to order one in june to compare to the HR22 and NH-D15 (sadly the NH-U12A and Mugen5 are already in use), i still think its going to be tight with the 3 of them, as all have massive areas of dissipation and perform well on low rpms, ill probably will try it with NF-A12x25 just to see how it does, still dont think its worth the extra $60 over using the kazaflex, but will see.

So a friend had their intel cooler break down on Sunday, and ask me if i could get him a cooler that wasn't so expensive and heavy, so suggested the Scythe Kotetsu II $39.99, he liked it specially since it matches his ASUS TUF H370-Pro Gaming board, and the cooler just arrived to the office.

Probably wont be able to test much, as he is picking it up tomorrow, but will see what i manage, maybe just with the NF-A12x25 to compare it to the Mugen5, the fan included should be the same, just with RGB.

The white blades might be a different plastic composition than the std. black..and might lead to different performance. Hopefully, not.

"The only thing that i dont understand, is from what i know, they used the same fan as the Mugen5, and two fans will be louder than one, so why is the noise graph reflecting the ninja5 as quieter than mugen5"

The fans are not exacltly the same model according to Scythe web site.

I didn't have the time that i expected due to family issues, but i tried to do the best that i could with what time i had left with the Kotetsu II, the testing was done on the same motherboard, same cpu (same clocks), same thermal paste, only difference is the ambient temp was a little cooler, so take that into account before judging the differences in performance. Overall the cooler is pretty solid, for $40 the performance is very comparable to the Mugen5, all with in margin of error, for the most part very even.

awesome

Reviewed at TechPowerUp: https://www.techpowerup.com/reviews/Noctua/NH-U12A/

Looks like scythe upgraded the fans for the FUMA 2, opting for one kaze flex and one of their new slim versions. Same (or better?) ram clearance than the u12a, $60 on us newegg. Seems like a direct competitor to the new Noctua

Full details of the Fuma 2 are on the Scythe US web site. One interesting point: the fans rotate in opposite directions. Scythe call it Reverse Jet Flow and claim that "...this generates higher static pressure as well as stable airflow thus pushing (t)he performance to the max...".

The Fuma2 seems a very interesting cooler, specially thinking on how dual fan helps low rpm operation, the only thing that i dont like on it is the frontal fan is a slim 120, so if it goes bad... and not sure you can place a standard 120 on it, the clip seems shorter from the pictures, but its not a big deal either, Noctua also has the NF-A12x15 that should work if you need to replace it down the road in case scythe dont sell their slim 120.

Abula, thank you very much for your copious testing. I think your results help explain some of the difference between SPCR's results of the D15/Mugen/Kotetsu and the ones I saw on ComputerBase: the CB tests appear to only cover the 50-100% fan speed range.

I'm thinking about how the shape of each cooler's base may affect how well they work with the incoming Ryzen 3 series processors. SPCR paid a lot of attention to the shape of each heatsink base, as well as the shape of the processor heat spreader. Ryzen 3 is interesting because its 8-core offerings position the main chip in the corner of the package, rather than in the center. AMD is also said to use a flatter IHS than Intel, which tends to be concave. (SPCR's i7-965 test processor was lapped flat, so may actually be more comparable to ComputerBase's testing on Ryzen 1700x than their previous Sandy Bridge test machine.)

The Kotetsu base looked a bit conical in SPCR's review, and that was credited in part for its good performance. But maybe for Ryzen 3 it would be a liability?

I don't see any passive testing?

Honestly its hard to say, this is the first chiclet design, but you might be into something thinking that the AMD design has the contact points not centered, so the highest pressure of a conical base might not be the best, that said its soldier to the ihs, and their lower architecture and more efficient design might make it not matter as much, it would be interesting to test to see how it behaves.

I don't think you can cool passive a [email protected] on aida64, well you might be able to with a bigger dissipation heatsink like the NoFan95, but for standard usual designs i don't think its viable, that said the 15% runs are almost as passive, the fans barely move any air.

There was a couple of reviews when the noctua NH-D14 was released beating the HR-02 in passive mode but i can't remember what CPU it was , i'm pretty sure it was at stock.it would have been interesting to see how they all held up before thermal throttling kicked in.

Even though the Noctua D14 has the wrong fin pitch there is a gap for air to circulate between the towers if the fans are removed for natural convection , just shows how complex the science of passive cooling is.

I never have tried nor i own D14, but you can see on the testing with NF-A15 on 15% where the HR22 takes 9C (10 if you consider the delta ambient) over the D15 with a single fan, so i would assume without a fan the HR22 will still win over the D15, but niether will be capable of sustaining it passively imo, but could happen underclocked/underevolted, remember im running the [email protected] on all cores, which is the turbo max core that should only be allowed for certain time but im running it full time, the real base clock is 3.7ghz which is more likely to be able to pass with either.

Interestingly there is a new prototype by Noctua shown in Computex2019 that can cool a [email protected] fully passive, it a 1.5kg monster but seems effective. Among other things like a new itereation of the NH-D15 with 7 heatpipes (instead of 6) and a black NF-A12x25, Noctua are Finally Doing it - Chromax Everything

When i first saw that prototype my initial thought was " i want one " followed by " i don't want that weight on my socket " as the HR-02 is bad enough for socket sag when the board is on it's side.

Though Noctua make great mounting hardware so i'm sure they are working on that as well.

I just ordered some heatsinks to test,

The Scythe ninja5, i saw a lot of reviews about it being really good at low rpms, and was checking the specs and the rpm range of the fans included are 300-800rpm, i thought was a typo... but seems its the way it is on scythe webpage, ill confirm if they are like that, nothing wrong with that range, but unsure if it might be able to cool a 9700k/9900k with such low rpm range, but will see, ill also test it included fans and with dual NF-A12x25 allowing them to go above that range of rpms, but its very likely that the NH-D15S finding will also apply here, that the second fan at very low rpms does help a lot.

Scythe Kuma2, i want to see how it performs with dual fan with opposite fan rotation. I have NF-A12x15 and NF-A12x25 to test standard fan rotation. This time ill try to match the rpms on this test not the PWM % to see if the orientation and fan design matters or not.

Thermalright Silver Arrow T8 was recently added by Nansgaminggear, this is an adaptation of the TR4 with 8 heatpipes with a bigger plate to the other none TR styled. I want to test vs NH-D15S, but i have my doubts it will perform as good with the less area of dissipation and less fin spacing (probably will perform very well at high rpms), but maybe the more pipes and bigger plate makes up for it.

Around middle july ill share what i find, as i gotta go on the road for the next two weeks.

Do you guys want me to test it on aida64 on the usual [email protected] (delided with liquid metal, this is how all the previous testing has been ran) or [email protected] (soldier, but very likely wont pass low rpms tests).

I'd be interested in the 8086k results. That way there's more likely to be some valid result of the low RPM tests and a performance comparison can be made.

+1 for 8086k so it'll be comparable to the other data you've already generated. I am interested in how low of fan speed you could cool the 9700k, though. maybe after the 8086k tests if there's an obvious winner or two you could rerun with the 9700k.

I'll reiterate that these tests are awesome. I come back and look every week or so to be reminded that there's no need to swap from my mugen 5

Yeah - stick with the 8086

The coolers have arrived last week,


i had some time this weekend to do some fan testing, wanted to see the difference between all the included fans before doing the heatsinks tests. I have decided to avoid Asus FanXpert because its not reliable 100% of the time, sometimes has issues with some fans like Silverstone PWM, so decided that even though takes more time, as i have to dail the PWM % and reboot, its more accurate and reliable. So here some of the some of the runs on the included fans that came with the heatsinks,






I have also modified some the testing chart to reflect the past testing (red text) and adjusted some of the NF-A12x25 testing PWM to match better what Scythe/Thermalright deliver with their fans,


Ill probably will do the testing during the weeknights if i finish ill share them on Friday, else probably sunday.

Cool.

For the fans that top out with high rpm, you might consider bailing on the 100% PWM and replacing with something lower. I think for this crowd the interesting stuff probably happens below 1000rpm. It'd be nice to know where the inflection point of the rpm vs temp curve is...but not a lot happens beyond that other than a few degrees C and a lot of noise.

While i do understand your point and i agree, its good to know how capable a cooler is, not that we are going to use it like that, but todays cpus like 9900k/9700K run very hot, on high load i can bearly keep my TY147 below 1000rpm without hitting 100C, 8086k is delided and being ran with in intel spec reason that we still see it operational even at low rpms, but most find deliding something that fear and dont do it, so for practical purposes it allows me to do low rpms test and still at 4.3ghz which imo is a pretty decent clock speed for 6core, but in reality you will need more rpms for a none delided cpu (or lower voltages or clocks), for this i think the max rpm is still a good test, and it should also show low gains on high fin spread while bigger gains on tight fin spread. That said, i do agree with your that for SPCR is useless the max speed, so im willing to add a extra line for each test, let me know which would like to see (the ones in red i cant retest atm, coolers are being used in other builds).

On other news, i already contacted a construction contractor, that will come in august to do a purposal of building the extra room on the back garden, so house made chamber will likely happen by years end, im already contacting the AO Pacific for a 1'' mic like Mike had to have good reading below 20db, still a challenge to see if i can build the room that can reach that low. But when i finish it, the goal will be to have similar testing to what SPCR did on the past, to me the end goal is db vs temp, but to do this i need to be able to measures at least between 15-30db (hopefully 10db).