Best Type of CPU Cooler Final Answer Linus Tech Tips
Linus Tech Tips
·Linus Tech Tips
·2013-05-07
·
2,464 words · ~12 min read
0:07
So, I did say that with proper fans, and I would define NFF12s from Noctua to be
0:13
about as proper fans as you can get for heat sink and radiator applications,
0:18
these are directed flow fans that are optimized for static pressure, optimized
0:22
for silence. You can see they have rubber isolating mounts. I mean, honestly, just the build quality of
0:26
these quality of these is just outstanding. I've unboxed this fan
0:30
before, so you probably already know what I think about it if you saw that. But if you haven't seen that, it's
0:35
freaking awesome. Basically, it's like the best
0:39
fan. Um, where was I going with this? Ah,
0:43
yes. So, without the adapter, it runs at 1500 RPM. Personally, I'm going to turn
0:48
these bad boys down. So, I'm going to use the low-noise adapter because I'm a
0:51
bit of a silence freak. This is going to be my standardized test bench. But
0:55
before I get started with my with my heat sink and radiator test bench, I
0:59
wanted to do a quick test cuz I had some people asking me, uh, well, you say
1:03
there's not much difference between just pull and push pull. As long as you're
1:07
using good quality fans, uh, what do you mean? Are you sure? Are you sure,
1:10
Lionus? Are you sure? So, that's what I'm going to be doing. I'm going to be taking these fans with their low-noise
1:15
adapters. I'm going to be running them in just pull on an
1:19
H100i and then I'm going to be running them in pushpull on an H100i to see with
1:23
a pressure optimized fan just how much of a difference there is. So I will be
1:27
back. All right guys, this video is going to serve a couple of purposes. So
1:31
number one, it will answer once and for all that question that people keep
1:35
sending me. What should I get to cool my CPU? And I'm talking like extreme like
1:40
not you know the Hyper22 users out there. I mean, that's a good heat sink
1:44
and all that, but we're talking performance options here. So, we have
1:47
the H100i from Corsair right there,
1:51
representing the dual radiator pre-done liquid cooling class of products. We
1:56
have the Silver Arrow Extreme representing extremely high-end air
2:01
cooling. See, dual towers, like eight
2:04
heat pipes, whatever else is going on in there. Don't worry, I'm not using these
2:08
fans. And we have the H80i representing
2:11
thick single radiators. Oh, I should guess I should go find a thin single
2:16
red. Hold on just a sec. Actually, forget it. I said this was for
2:19
performance options. So, we're doing performance options here. Now, I want to
2:23
talk a little bit about my standardized test bench here. This is a Corsair C70
2:27
Vengeance case. I'm going to be running it with the stock fans in the front. I
2:31
am going to be running it with this fan in the back regardless of whatever else
2:35
is going on. That is a Noctua NFF12. My favorite fan. And whatever cooler I put
2:41
in it is getting more Noctua NF12s.
2:45
These are going to be run with their no their low-noise adapters cuz I
2:49
personally don't think even for the highest performance gaming rig these
2:52
days that there's any excuse for it to be loud. So that's a 3930K in there.
2:56
That's going to be overclocked and overvolted. That's a GTX 580. We're
3:00
going to close up the side panel because that's enough of CPU coolers being
3:04
tested on open test benches. I know it's not realistic, so I'm not going to be
3:08
lazy about it anymore, guys. We're testing all CPU coolers in a closed case
3:12
without perfect cable management, just like they're going to be run in the real
3:16
world. So, we're going to take idle and load temperatures of each of these
3:21
setups. We're only using Noctua NF12 fans. We're replacing whatever comes on
3:25
whatever coolers we use. In the event that a cooler must use a larger than 120
3:31
mm fan, we have A15s, so we can use
3:34
those instead. Those are 140 mil fans.
3:39
Um, so let's get started. Oh, right. One
3:42
more thing. The reason I'm using the same fans on every cooler because the
3:45
stock coolers that or the stock fans that are included with most coolers are
3:48
rubbish anyway. So you and again, we're talking performance options. You should
3:53
be buying your own fan that's better. Um, also I'm using aftermarket thermal
3:58
compound on all of them. I'm using IC Diamond thermal compound. So there you
4:02
go. I've had a lot of people talk about IC7 versus IC 24. Guys, it's the same
4:07
stuff. The 7 and 24 for carrots. Carrots are just weight. They're not purity.
4:14
Um, right. So, yeah, consistency. So
4:17
that way we are going to have a consistent sort of overall noise level
4:21
to the system because we're going to be running good quality fans and we're
4:24
going to get a consistent experience with fans that are correctly optimized
4:27
for whatever kind of cooler we're using because these are correctly optimized
4:31
for any heat sink or radiator cuz they're awesome. High pressure, focus
4:35
flow, just good, good, good, good fans. Step one, clean the existing thermal
4:40
goop and any residue off of both surfaces. So clean CPU is needed. Clean
4:44
bottom of the heat sink is
4:47
gain up. Also needed guys, when applying thermal
4:52
compound, less really is more. Remember, this is a 2011 chip, so that's how much
4:57
goop it takes for that. For an 1155, it will be much less. For this next shot,
5:02
prepare for your mind to be blown. I'm going to turn around my shotgun
5:06
mic. There we go. Get that pointed at the computer so you guys can hear what a
5:11
gaming machine sounds like when it's correctly optimized.
5:22
Oh. Oh, it's silent. L minus. Why do you
5:25
care so much about the silence of your gaming rig? Well, because it doesn't
5:30
have to be loud. Hold on. Turn my mic back around. It's
5:33
unnecessary. It's just ridiculous. If you want a loud gaming rig, then why
5:37
invest in these elegant cooling solutions that do such a good job? Why
5:41
not just use a stock cooler? And why don't you get an Xbox while you're at
5:45
it? You want a loud gaming machine? I've explained this before in a
5:50
previous video, but I just want to go through my methodology again really quick, guys. So, basically, I'm using
5:54
Realtemp GT to take temperatures. I am
5:58
recording idle temperatures after about 5 minutes just idling at the desktop.
6:03
Um, I am recording ambient temperatures
6:06
using my multilogger thermometer right here with a I forget something. Type K
6:11
type KT type sensor. There it is with a Ktype sensor um right at the intake for
6:17
the case. So remember guys, the front is the only intake that we're using here.
6:21
So the air is uh basically coming from there. So that gives us a pretty good
6:25
idea of what temperature the air cooling the CPU heat sink is. And then I am
6:30
taking load temperatures after about 5 minutes. For air coolers, I don't need
6:33
to wait quite as long, but for liquid coolers, you have to wait for the liquid
6:36
temperature itself to equalize. And then I'm recording everything here. So, I no
6:41
longer need these noise notes unless uh a water cooler has a pump noise because
6:46
remember guys, with water coolers, you have whatever fans are running on it
6:49
plus a pump. Every mechanical thing you add to a cooler is a another point of
6:54
failure and b another potential noise source. And I want to explain again how
6:59
I'm correcting for idle temperature for ambient temperatures. So, I'm going to
7:03
correct to 20°. I I had people debating
7:06
me about this last time. Please go find a physics professor or something and ask
7:10
them if this is how it works. If the temperature were to go up to 20°, it
7:15
would scale by 11.8°. That that's just that's how it
7:20
works. It's not it's not something that I want to debate about. So, I'll be
7:24
correcting all temperatures and reporting them as if the room was at
7:27
20°, not as if it's at 9.2 degrees cuz
7:31
I'm in Canada in my garage. And another thing for low temperature testing, I
7:35
will be running the GPU at full load as well as the CPU at full load just to
7:40
simulate again a closedcase environment where the system is being taxed and a
7:45
worst case scenario for an actual real
7:48
system that's actually been built. So let's go ahead and take our load
7:52
temperature. So I'm taking the second hottest core, which looks like it's at
7:55
about 61°. Again, guys, this isn't like the super scientificest way to do this.
8:00
So 61 and a 10 degree ambient. There we go. All right. It looks like the
8:05
temperatures haven't gone up since I did my H100i load test. So here's H80i. And
8:11
I've been saying for a long time, and I was never quite sure how right I was
8:15
that single rad, like a single 120 mm rad solutions are not even close to a
8:20
dual 120 mm rad. So you can look at
8:24
there's our there's our idle temperature. It's around 29 on the
8:27
second hottest core, which gives us a difference between an H100i and H80i of
8:32
like 15 almost 15° under uh at idle. So, let's see how
8:37
that gets changed under load. All right. So, here's the H80 with
8:42
one fan, but I'm going to give the H80 the benefit of two fans since it comes
8:46
with two fans. So, we'll test it that way as well. But it looks like our load
8:50
temperature is
8:53
85° and our intake is
8:56
10.1. Well, I was a little surprised at how much of a difference it made to
9:00
switch to a push Whoops. 10 Oops. 10.8.
9:07
There we go. To switch to pushpull on the H80i. I guess it shouldn't surprise
9:11
me that much because it is a thicker radiator. I know on the H100 it doesn't
9:15
make that much of a difference, but there you go. Our corrected temperatures
9:18
at idle show about a 3°ree improvement over just a single pole fan. And then at
9:23
load, it's about a 10°ree difference. However, it's still nowhere near the
9:27
performance of the H100i. Or rather, sorry, hold on. Here's
9:32
our corrected one. So, yeah, about 10 degrees delta and then our difference
9:35
between them and uh still 13 degrees worse than an
9:41
H100i. Well, it's time to measure the contender. It should be noted, guys,
9:45
that there is more to CPU cooling solutions than just performance. I mean,
9:50
for example, these Corsair coolers do come with the link software, which
9:53
allows you to set up things like fan curve profiles and cool stuff like that.
9:58
Whereas, I mean, in my experiment, I'm limiting the factors as much as
10:01
possible. So, I'm keeping all the fans at a constant RPM, but that's something
10:04
you have to bear in mind. Also, installation for the water coolers is
10:08
much easier than these huge bulky air coolers, however blingastic they might
10:12
look with their multiple fans on them. You can see the configuration is still
10:16
standard. Also, with big air coolers, you can see there's very little clearance between that lowrofile dim and
10:21
the uh fan itself. So, you can't you're really limited in terms of what memory
10:25
you can install. Um, more wires. Actually, not necessarily more wires,
10:29
especially with that USB cable that runs to the HI series. So, let's find out how
10:33
this does. Let's boot her
10:39
up. All right, so idle temperatures are 26° for our Silver Arero Extreme and
10:46
ambience have gone up a little bit since we started here to
10:51
11.1°. Well, there you have it, guys. The H100i, even with all other factors
10:57
being the same, is the ruler of the roost. So, our ambient are still
11:02
11.1°. And with the Whoops, sorry. With
11:05
the Silver Arrow Extreme, we are at 67°.
11:10
Whoops. Sorry about that. We are at 67°
11:13
on the second hottest core. So, let's go ahead and fill out those temperatures.
11:19
1.1 67. So, you can see that it per
11:22
Whoops, Freudian slip there. So, it's
11:26
67° under load. So that corrects to
11:29
about 75 degrees under load. So there you go. So there's the
11:34
H100i. There's the H80i with dual fans.
11:37
There is the Silver Arrow Extreme with the same fan. So you can see the silver
11:42
a high-end air cooler. So remember guys, something like an NHD14 uh will be quite
11:47
similar to this in terms of performance. Destroys something like a thick single
11:51
radiator even in pushpull with highly optimized pressure optimized fans.
11:56
Whereas the H100i, especially under heavy loads, really pulls away from
12:01
those other solutions. See, at idle, these I mean, even these two look very
12:05
similar, but you can see how under load the uh Silverado Extreme really pulls
12:10
away. Now, that really demonstrates that as the as the heat goes up, so to speak,
12:15
the uh the men separate themselves from the boys in a test like this. I mean,
12:19
even this is only a 3° difference. That's why I don't like the way that Corsair does the packaging on these
12:24
products, at least the way they used to with the H100, where they would compare
12:28
it. Yeah. See, they still do this. It's just goofy. They compare it against an
12:31
Intel box cooler showing Well, okay, now
12:34
they don't even try to show a proper temperature, but they used to and uh it
12:39
didn't show that much of a difference because it's not until you overclock that it really makes a big difference.
12:43
So, thank you for checking out this video on Linus Tech Tips, demonstrating
12:48
that the H100i is the ruler of the roost. Don't forget to subscribe for
12:52
more unboxings, reviews, and other computer videos.