Grouptest: CPU cooler group review with 3770K

[Hilbert Hagedoorn]

In this review we test over a dozen CPU coolers, mainly heatpipe based. We'll test them on a Core i7 3770K, one of the more difficult to cool processors on the market. In this group test we'll use S...

Grouptest: CPU cooler group review with 3770K

[PhazeDelta1]

I see the NH-D14 is still king of the lab.

[ASLayerAODsk]

I have a CM Hyper 212+ on mine aka /evo interesting to see how it fares at that level... good review tyvm

[SLI-756]

Thanks very much this is an excellent resource.

[King4x4]

Would of have loved seeing some custom WC to see any major differences... lets say a 360 rad with a decent block.

[Denial]

I still don't get what I'm doing wrong with my system. I currently have a CPU370 with 4 120's directly to it as my GTX690 is down and even after delidding I can't seem to get my temps under 70c full load @ 4.5Ghz. I also can't seem to run it under 1.3v stable at that speed either yet I see pretty much everyone else able to do that.

Feel like I got shafted =(

[BLEH!]

H100i with (Noctua ) aftermarket fans on a 980X is 10C cooler than a NH-D14 for me. Good review though boss, insightful.

[benqbiggis]

Originally Posted by Denial

I still don't get what I'm doing wrong with my system. I currently have a CPU370 with 4 120's directly to it as my GTX690 is down and even after delidding I can't seem to get my temps under 70c full load @ 4.5Ghz. I also can't seem to run it under 1.3v stable at that speed either yet I see pretty much everyone else able to do that.

Feel like I got shafted =(

me too

i can't get 4.6ghz fully stable below 1.295v-1.30v

mine is not delidded but even using a custom water cooler with 4 cooler rad and 1500 pump in a simple loop i cant get temps below 90c-95c under prime95 at max cpu torture

even considering my ambient temp over 30°₢ (very hot summer here) and 2 gtx 580 blowing hot air

[SLI-756]

Originally Posted by benqbiggis

me too

i can't get 4.6ghz fully stable below 1.295v-1.30v

mine is not delidded but even using a custom water cooler with 4 cooler rad and 1500 pump in a simple loop i cant get temps below 90c-95c under prime95 at max cpu torture

even considering my ambient temp over 30°₢ (very hot summer here) and 2 gtx 580 blowing hot air

Oh! another candidate for the delid test?
Sounds like it would greatly benefit you but with the 3770k is a huge risk, i was prepared to go weeks without computer before i done my 3570k.
There's a couple of great delid threads here:

3770k Discussion thread - Heat Possible improvement!
http://forums.guru3d.com/showthread.php?t=372566

i5 3570K OC Adventure
http://forums.guru3d.com/showthread.php?t=367095

[Denial]

Originally Posted by SLI-756

Oh! another candidate for the delid test?
Sounds like it would greatly benefit you but with the 3770k is a huge risk, i was prepared to go weeks without computer before i done my 3570k.
There's a couple of great delid threads here:

3770k Discussion thread - Heat Possible improvement!
http://forums.guru3d.com/showthread.php?t=372566

i5 3570K OC Adventure
http://forums.guru3d.com/showthread.php?t=367095

I was at the same temps he was at, like 90c until I delidded it now I'm down to like 77-82c. But then I see people that hit like 60c and I'm like what the hell is going on.

[SLI-756]

Originally Posted by Denial

I was at the same temps he was at, like 90c until I delidded it now I'm down to like 77-82c. But then I see people that hit like 60c and I'm like what the hell is going on.

Try the spread method if you didn't already, on both the die and inbetween cooler @ IHS(light spread on die and over outside of IHS), i recommend the diamond based but it's thick and horrid to spread.

[mmicrosysm]

What no SilverArrow?

[Denial]

Originally Posted by SLI-756

Try the spread method if you didn't already, on both the die and inbetween cooler @ IHS(light spread on die and over outside of IHS), i recommend the diamond based but it's thick and horrid to spread.

I've tried a line, a dot and using a plastic bag and my finger to spread it. The plastic bag finger thing worked the best but I'm still not seeing the sub 70c results other people post. I'm using Noctua NT-H1 for paste, which while not the best, is definitely up there as far as results are concerned. I've considered completely removing the IHS and modifying my board to make it work but I'm not sure if it's worth it.

Like looking at Hilberts results, he's getting 72c at 1.3 and 62c at 1.2 and I'm assuming he didn't replace the TIM/Delid the thing. I'm getting ~79c @ 1.3 after delidding and anything under 1.29v and I can't run prime95 for more than a few seconds without a BSOD.

The only things I can think of is

1. Because I don't have a 690 the CPU is running hotter due to the HD4000 being active? When I get my card back on wednesday I can rule this out. But seeing in the pictures it looks like Hilbert was also using the on die chip during the benchmarks so, yeah.

2. My 370 block just isn't sitting right on my chip. If I pull the block up even a little dab of thermal paste seems to spread out all over so I don't think this is the case.

3. My chip is just bad. I see multiple people running p95 @ ~1.2v without crashing even at 80c+. I simply can't do this. I get a BSOD instantly. So unless the HD4000 is somehow interfering with my results causing it to crash I'm just going to assume I got a bad chip.

I'm not really too concerned about it as it rarely hits 79c in a real game anyway just P95 and even then it's stable at those temps. Plus I usually upgrade every 6 months anyway so the life of the chip being effected is really not a problem. It's just annoying to know that I can't drive it as fast as other people are.

[Agent-A01]

Nice review, but i would have liked to see the Megahalem included. With it lapped and 2 Noctua NF-F12s it did better than a NH-D14 i borrowed from a friend

[Solid_State]

Have a Ninja 3, it definitely has some strengths and also some weaknesses. I ran my i7/920 passive on my old MB, but when overclocking paired with a cheap case it would get to hot for me.

I will definitely put her back in use for a completely silent HTPC in the near future

And about the mounting system on the Ninja, no worse system exists! You literately need the skills of a Ninja to install this cooler

[WhiteLightning]

Originally Posted by BLEH!

H100i with (Noctua ) aftermarket fans on a 980X is 10C cooler than a NH-D14 for me. Good review though boss, insightful.

I wonder if the environment (testbench fully open VS closed case) had anything to do with it.

[Elder III]

Hilbert ---- you are bad for my virtual wallet.

[bonob]

47°c at stock clock with one of the best cooler out there, man i mean with a cpu rated at 77W by intel (http://ark.intel.com/products/65523), someone must be lying somewhere.

[malky]

Originally Posted by mmicrosysm

What no SilverArrow?

my thoughts exactly, are they no longer in production?

[Titan29]

Nice review. Looks like I will be keeping my NH-D14 for a long time

[thatguy91]

Originally Posted by mmicrosysm

What no SilverArrow?

Originally Posted by malky

my thoughts exactly, are they no longer in production?

I believe that is the case! The SilverArrow has been replaced with the SilverArrow SB-E. There is also the 'overkill' one, and designed that way! The Silver Arrow SB-E Extreme.

The review only covers a handful of coolers. I would have been interested to see how the Xigmatek Dark Knight Night Hawk edition would have gone, since that has got good reviews on some sites (and it's the one I have). There's no denying the Noctua is one of the best coolers available, but the results are a little surprising. The limiting factor in the ability for a cooler to cool is the weakest link. The weakest link in this case is the thermal interface between the actual chip and the integrated heat sink (IHS).

Now, is the temperature the core temperature or the temperature of the CPU? At the moment I'm doing encoding in a room that is mid to high 20's (celsius), the CPU temperature is 40C (remembering I am running at 4.4Ghz, 43x102.33). but the individual core temperatures are range between 55C and 64C depending on the core. Actually one core is going down to 50C or less at times, I guess that depends on the thread load on that core.

At the 4600Mhz, 1.3V load test, the CPU temps range between 72C and 96C. I would think if it were the CPU temps, the core temps would actually be significantly higher than that? What if, during the test, the cores hit their thermal maximum and throttling was invoked, does the test take that into account?

I suspect that for the NH-D14 that either throttling may have played a part, or the different CPU mounting pressure meant that the limitations of the interface between the actual core and IHS were slightly overcome. Otherwise, it would be a 'sucked in' to all the idiots who have gone for the inferior Corsair H100 water cooling. Higher temps, more noise, crapper! (just stirring here, I believe the H100 should have done better).

[MadGizmo]

Thanks, Hilbert. Loved to read this.

It's interesting the H100i performed worse than I expected. So I am going for air in my next build for now. I wished the Zalman CNPS 12X was in that list, but you tested that one not too long ago. It seems that the 3770K is another beast altogether. I wonder how it performed with that one, though.

[Uncle Dude]

Originally Posted by thatguy91

all the idiots who have gone for the inferior Corsair H100 water cooling.

4.8GHz/1.4v & 70C with 1350 rpm SP120 Quiet fans and I'm one happy idiot!

[Agent-A01]

Originally Posted by Uncle Dude

4.8GHz/1.4v & 70C with 1350 rpm SP120 Quiet fans and I'm one happy idiot!

5Ghz 50c w/ 1500 rpm nff12s

but still, no point in AIO solutions, its time to go all out on a real setup

[thatguy91]

Originally Posted by Uncle Dude

4.8GHz/1.4v & 70C with 1350 rpm SP120 Quiet fans and I'm one happy idiot!

I was getting at the fact it should have done better than the Noctua Hopefully Haswell is once again soldered (if Intel spend the extra 5 cents a CPU). CPU cooler reviews seem to be trend based, in that a good cooler will trend well, bad cooler will trend badly, but within these trends are some sights saying one cooler is much better than another cooler that a different site says is better than the first. I think if the Noctua and the Corsair H100 were reseated and tested again, the results would be different, such is the nature of CPU reviews.

I could live with a CPU temp of 50C, mine got to that when I was encoding and the room got to 41C (hot day) when I was out, saw the temp when I got back, but the core temps were 70C which was borderline with what I was comfortable with. I can only imagine if the CPU temp, like in the reviews, was 70C the core temps must have been excessive and damaging over time (if thermal protections didn't kick in).