Cooler for overclocked i5 4670k

[doveman]

I've just done a build for my Dad with a i5 4670k and Hyper 212 EVO and that's running at [email protected] at reasonable temps and quite quietly. I replaced the original fan with a quieter one and then put the original one on the back as a pull but it didn't make much difference to the temps or noise, so I'll probably remove it.

Anyway, I've got my own i5 4670k build to do now and I'm trying to decide which HSF to use with it. Whilst the 212 EVO performs quite well and is only £25, I'm hoping I'll be able to OC my CPU a bit further than my Dad's. I stopped at 4.4Ghz with my Dad's as it required quite a jump in voltage and thus temperature to even go to 4.5 or 4.6Ghz and I don't consider it worth the extra stress on the CPU for 200Mhz. I guess the voltage isn't that important (within limits) it's the temperature that causes wear, so with a better heatsink it might have been possible to up the voltage and hit 4.6Ghz or even higher whilst keeping the temperature well within the 'safe' range.

I know there's variability in the CPUs and mine might not even OC to 4.4Ghz so easily, so it's a bit of a gamble spending more on a better HSF and for that reason, I'm not planning to spend silly money. I can get the Phanteks TC12DX for £38, which I've seen good reviews of http://www.kitguru.net/components/cooli ... -review/6/ or I could get the Kotetsu, which has a good review here http://www.silentpcreview.com/article1391-page6.html , for about the same, although I don't know how they compare.

Apparently the Phanteks has a somewhat fiddly clip mechanism for attaching the fan but that's not really important compared to how efficient and quietly it can cool the CPU.

So is there anything better than those two for about the same price (or less) that I should be considering?

[lodestar]

For £38 - £40 in the UK you could consider the Scythe Mugen Max from sources such as Overclockers. This seems to be a larger version of the Mugen 4. The Mugen 4 has been shown to be a fairly decent performer so the Max should do even better. At around the same price as the Mugen 4 it may well be worth a try. The Max is fitted with the 140mm Glide Stream (140mm rotor with 120mm size fixings). The graphics below illustrate the difference in dimensions between the two:



Scythe Mugen Max



Scythe Mugen 4

[Haych]

Also for about £40 (once P+P is factored in) there's the Thermalright HR-02 Macho Rev.A available at Scan. http://www.scan.co.uk/products/thermalr ... d-amd-cpus

[quest_for_silence]

I can't speak for the Mugen Max but, set aside it, none of the aforementioned coolers is preferable to the Scythe Kotetsu (so I guess it's mainly a matter of aesthetics and price).

[doveman]

Thanks for the suggestions guys.

Wow, the Mugen Max is a big ol' beasty! 145 x 86 x 161 mm but surprisingly only 720g (both without fan) compared to the Kotetsu's 130 x 58 x 160 mm and 480g. Weight and height aren't really an issue for me, as I'm using a CoolerMaster HAF XB EVO case but I will have to check whether the Max will clash with the VRM heatsinks on my Asus Z97-A board, or block the top PCI-E slot (in case it's the only slot spare after fitting two graphics cards, which I might want to do one day, it might be useful to have available). I use low profile RAM, so I shouldn't have any worries on that side but the Max is designed to avoid that problem anyway.

The HR-02 Macho Rev.A is 140 x 102 x 162 mm and 710g (Scan has completely bogus specs, so I checked the TM website for the correct ones) but if quest_for_silence is correct that the Kotetsu beats it, then clearly the extra depth/width doesn't equate to better performance and may not for the Max either. There doesn't seem to be much good quality testing data available for it yet though. Gotta love the Press Release "Fan speed can be controlled conveniently by the PWM controller of the mainboard in a range from 500 to 1.300 RPM, which allows airflow between 63,5 and 165,1 m³/h (37,37 ~ 97,18 CFM). Hence allows it to operate virtually inaudible at 13 dBA at lowest speed or really silent at 30,7 dBA when running at maximum." So 30.7 dBA is "really silent", whilst 13 dBA is only "virtually inaudible" I guess they meant "really quiet" but 30.7 dBA certainly isn't that either.

This review http://www.play3r.net/reviews/cooling/s ... -review/7/ shows it performing worse than the Phanteks PH-TC12DX, whilst this one http://www.reviewstudio.net/1937-scythe ... -and-noise shows it comparing quite favourably to more expensive Noctua and Phanteks 140mm fan heatsinks. I've never heard of either site before though, so I'm only taking those results as a rough benchmark comparison.

There's nothing really in it as far as price goes, the HR-02 would cost me £34 (I get free shipping from Scan), the Kotesu is £36.75 and the Max £34.

So considering that the Max is somewhat of an unknown at the moment, is more likely to block my top PCI-E slot and/or clash with my VRM heatsinks and I can replace the stock Kotetsu fan with a Nexus 120mm for better/quieter performance (although I quite like the Arctic F12 PWM, so might use that instead but I think I've got one of each in my box of fans, so I can see which works best), then I reckon the Kotetsu is the winner today and I should go and buy it and complete my jump from Phenom II X4 955 to i5 4670k ASAP.

[doveman]

I just got round to building my i5-4670k system with the Kotetsu yesterday.

I'm running at a fixed 3.8 Ghz at the moment, as I found that it caused latency problems when there was network activity if I let it change clock speed, so I'm using Windows High Performance mode. For some reason, the BIOS has set the uncore/cache to 3.8 Ghz as well, whereas I think with my Dad's build (same Motherboard, BIOS version and CPU) it set it to 3.4 Ghz, which I believe is the correct default (CPU stock is 3.4 Ghz as well, so I'm not sure why that's defaulted to 3.8 Ghz either. Maybe Turbo but I thought that was only meant to increase the clockspeed of one core when the rest were idle and they're all at 3.8 Ghz).

Anyway, I've done some quick stability/heat testing before I start to overclock it and wanted to check whether the temps look reasonable or if I should re-fit the heatsink. I probably put a bit more MX-2 (in a vertical line about 3/4 the length of the heatspreader) than necessary but I think it's impossible to make a 1mm wide line as the guides suggest, as the plunger doesn't allow you to control how much comes out, initially at least, so I find I just have to use the nozzle to drag whatever amount comes out along with, squeezing more out when needed as I go along and do my best to make a reasonably small line. I'm probably making it sound worse than it is and it's what happened when I did my Dad's build, which is cooling fine, so it's probably OK.

Monitored with HwInfo

Idle - High Performance - 3.8Ghz:

CPU Fan: 780 PM
Chassis Fan: 1000 RPM

CPU: 41c
Cores, 42, 38, 40, 36
[Under Nuvoton NCT6791D]
Motherboard: 31c
CPU: 36c
Temp2: 54c

I'm not sure that "Motherboard" is actually showing the Motherboard temperature and I don't know what Temp2 is measuring either.

The fans were a bit noisy like that so I manually edited their profiles in the BIOS. The Chassis fan I tied to the Motherboard temperature but as the sensor associated with that doesn't seem to increase, the fan stays very slow, so I'll change that so it's tied to the CPU temp.

Idle - High Performance - 3.8Ghz:
CPU Fan: 742 RPM
Chassis Fan: 318 RPM
CPU: 46c
Cores: 45, 42, 44, 41
Motherboard: 33c

Aida64 Stess Test for 30mins:
CPU Fan: 940 RPM
Chassis Fan: 325 RPM

immediate:
CPU: 60c
Cores: 59.5, 59, 59, 54
[Under Nuvoton NCT6791D]
Motherboard: 33c
CPU: 43c
Temp2: 52c

after 30mins:
CPU: 60c
Cores: 60.2, 59.3, 59, 54.5 (Avg), 67, 67, 66, 61 (Max)
[Under Nuvoton NCT6791D]
Motherboard: 33c
CPU: 43c
Temp2: 52c

immediately after stopping Aida64:
CPU: 47c
Cores: 47,44,44,41

I wonder if the fact that core 4 is running a few degrees cooler than the others indicates that I haven't applied the TIM very well, or if that's just normal?

For comparison, here's a screenshot from testing my Dad's build (using a Hyper 212 Evo) also running at 3.8 Ghz but at 1.173v, whilst mine is at 1.1v. His CPU fan was running at 1000 RPM and there was an intake fan and exhaust fan running at 792 RPM and 969 RPM (not sure which was which).



Here's another one with it at 4.4Ghz @1.25v (before I stabilised it @1.225v):


I'd prefer to keep my build below 70c if possible, without making it too noisy. I wonder if an Arctic F12 PWM fan I have will work better than the stock Kotetsu fan? That's on my old Phenom II X4 955 which I'll be using as my backup system in another case though, so I'll have to replace that with something if I use it on the Kotetsu.

I've also got

1*Scythe 120mm SY1225SL12M
1*Scythe 120mm SY1225SL12LM-P (PWM)
1*Yate Loon 140mm D14SL-12
1*NB BlackSilentPro 140mm

so I could put one of the 140mm (probably the NB) on the front of my Coolermaster HAF XB case http://www.coolermaster.com/case/lan-box/haf-xb/ to run slow and silent, if that's going to help much. I don't much like the Scythe as I find that they have an annoying tick at the low speeds I like to run them most of the time, to minimise the airflow noise.

[doveman]

Turned out the Noiseblocker box actually had the Yate Loon 140mm in it with a small two-pin connector (too small for a motherboard header) and an adapter to convert it to Molex. I could probably wire up a suitable connector to it sometime but for now I've just put the Scythe 120mm SY1225SL12M in the front as an intake. The other Scythe (PWM) is already being used as the exhaust. God knows what I did with the Noiseblocker, probably put it in the system I built for my brother over a year ago!

I tweaked the CPU fan to run at 780 RPM (PWM - 30% @ 30c), set the intake fan at 440 RPM (DC - 60% @ 40c with Zalman resistor, it was way too noisy without) and increased the exhaust fan to 370 RPM (PWM - 30% @ 30c). Ambient temp was 24c.

Idle:

CPU: 40c
Cores, 41, 38, 39, 33
Motherboard: 30c
CPU: 35c
Temp2: 54c
Temp4: 34c
Temp5: 54c


Aida64 for 30mins:
(Immediate CPU fan increase to 910 RPM and Exhaust to 510 RPM)

CPU: 55.8c (Avg), 62c (Max)
Cores: 55, 54, 54, 50 (Avg), 62, 62, 61, 56 (max)
Motherboard: 31c
CPU: 41c
Temp2: 54c
Temp4: 34c
Temp5: 54c
CPU Fan: 935 RPM
Intake: 435 RPM
Exhaust: 544 RPM
---
Then I overclocked to 4.4 Ghz @1.225v (still running in High Performance to lock the clockspeed). Ambient temp was 25c. I tweaked the fans a bit more.

CPU fan at 820 RPM (PWM 40% @30c)
Intake fan at 670 RPM (DC 90% @40c with Zalman resistor)
Exhaust at 530 RPM (PWM 40% @30c)

Idle:

CPU: 43c
Cores, 43, 40, 41, 36
Motherboard: 30c
CPU: 36c
Temp2: 49c
Temp4: 34c
Temp5: 49c

Aida64 for 30mins:
(Immediate CPU fan increase to 1060 RPM, Intake 710 RPM, Exhaust 810 RPM)

CPU: 66c (Avg), 73c (Max)
Cores: 63, 62, 62, 56 (Avg), 72, 71, 70, 66 (max)
Motherboard: 31c
CPU: 41c
Temp2: 54c
Temp4: 34c
Temp5: 54c
CPU Fan: 1080 RPM
Intake: 715 RPM
Exhaust: 870 RPM

immediately after stopping:
CPU: 44c
Cores: 44,43,42,37

So it doesn't seem to be doing much better than the Hyper 212 Evo in my Dad's build. Here's the results with that at idle and running Aida64 (ambient was 3c higher at 28c):

Even taking into account he's using a different case and fans, which were running at different speeds to mine (mine vs his, at idle: CPU 626 RPM vs 820 RPM, Intake 937 RPM vs 670 RPM, Exhaust 690 RPM vs 530 RPM):


(CPU 990 RPM vs 1080 RPM, Intake 1080 RPM vs 715 RPM, Exhaust 915 RPM vs 870 RPM):


I was hoping for better results with the Kotetsu. Comparing the max. core temps, as those are the only figures from his system that look correct:
His: 77, 76,73, 69
Mine: 72, 71, 70, 66

and taking the +3c in ambient temp into account, the decrease is -2c for cores 0 and 1 and 0c for cores 2 and 4, which could just be down to my more open HAF XB case compared to his NZXT Source 530 (with top vents blocked off).

[doveman]

Damn thing just gave me "WHEA Uncorrectable Error" again, whilst at [email protected]. Surely I don't need to increase the voltage even higher for this basic overclock?

I was running DCS World at the time. Played ArmA3 for several hours the other day at these settings, which was what triggered the same error @1.225v.

Had some weird stuff happening shortly before it BSOD as well. First, I noticed that VAC (Voice Assisted Commands) wasn't working, so I opened the editor to check and the My Documents\VAC folder was completely empty. OK, I thought, at least I'm running Crashplan to backup my files, so I tried to restore with that but each time it would say completed but in the results it showed 0 restored files and indeed, nothing was restored. After rebooting following the BSOD, Crashplan worked fine to restore these files (it restores them to the desktop for you to move to wherever) but then I checked My Documents\VAC and all the files were already back again!

[quest_for_silence]

I was hoping for better results with the Kotetsu. Comparing the max. core temps, as those are the only figures from his system that look correct:
His: 77, 76,73, 69
Mine: 72, 71, 70, 66

Check the article, it's only at the 12dB level that the CM performance collapse: otherwise the difference is around 3°C (in my quick shoot-out I had a difference of 5°C with the fan under 700rpm).

Surely I don't need to increase the voltage even higher for this basic overclock?

With any OC YMMV: did you try to raise the bar? Have you a spare drive to perform a fresh install? How much did you stress the rig while validating? There are lots of questions to work out in order to answer (or try to) to your one, I guess.

[doveman]

Check the article, it's only at the 12dB level that the CM performance collapse: otherwise the difference is around 3°C (in my quick shoot-out I had a difference of 5°C with the fan under 700rpm).

Well with the two different stock fans maybe but with the same reference fan, at 12v (17dBA), 9v (14 dBA) and 7v (12 dBA) the Kotetsu beat the 212 Plus (not Evo, which wasn't available then) by 4c, 6c and 7c respectively.

Now I don't currently have the same fan on the Kotetsu and Evo as the former still has the stock fan, whilst the latter has a Scythe PWM. I think the only Scythe PWM I have left is being used as my exhaust at the moment, so I could take that and put it on the Kotetsu to match them, although I don't know if that will just be quieter but push even less air than the stock fan, at whatever speed.

I wasn't very impressed with the hardware for the Kotetsu. The two crossbars had three holes for the screws into the pillars, depending on which board you have. I had to use the middle one for 1150 but it wasn't actually separated from the inner hole, so the screw would slide into the latter unless you held it tight in place, whilst trying to lower the crossbar onto the pillars and screw it down. It was almost impossible to tell whether the screw had shifted to the inner hole in the process. If you had a board that used the outer hole, then it would be easier as the screw could not move from there to the middle hole. Just a cheap washer might have been all they needed to provide to avoid this issue, or perhaps designing the crossbar with an open edge by the screw holes, so the screw could be partly screwed into the post and then the crossbar slid onto the screws from the side before tightening the screws.

The other issue was tightening the two posts securing the part that went across the heatsink, between the two crossbars. For some reason, someone thought it a good idea to make this like a u-channel, so it has a lip pointing up either side. Of course, trying to tighten the posts with the included mini wrench was made very difficult because of this, as they'd get into a position where the only place the wrench would fit onto them put it so close to these lips that you can't actually move the wrench, as it hits up against the lip as soon as you try. Much time was wasted trying to work around this and get them tightened up.

In contrast, the Evo was much easier to fit and whilst I expect some issues with a cheap cooler, these just seem like very poor design decisions and very much put me off the Kotetsu.

With any OC YMMV: did you try to raise the bar? Have you a spare drive to perform a fresh install? How much did you stress the rig while validating? There are lots of questions to work out in order to answer (or try to) to your one, I guess.

I researched that error and apparently it's sent directly from the CPU to the OS, so I think re-installing a fresh OS would be a waste of time. I stressed the rig with Aida64's stability test, which includes testing the RAM (I think it fills up at least 12GB of my 16GB). I'm trying to think what part of the Mobo/CPU could be being stressed when running games (not necessarily even playing them, as ArmA3 crashed just sitting on the menu. That does load the GPU almost as much as playing it actually but as the error is related to the CPU, the GPU seems irrelevant here) but not when running Aida64.