WEBVTT

00:00:00.000 --> 00:00:05.120
CPUs run hot. So hot that they would destroy themselves if it weren't for built-in protections

00:00:05.120 --> 00:00:10.480
that slow them down. But we don't want to slow down. We want to go fast and thus nerds are willing

00:00:10.480 --> 00:00:15.280
to spend hundreds of dollars to keep their CPUs cool. But what if there's a whole other side

00:00:15.280 --> 00:00:19.920
of cooling that no one's talking about? The back side, why not put a cooler there as well?

00:00:19.920 --> 00:00:26.800
So we did. Before you furiously comment saying, cooling a CPU from the back doesn't work, there's too many layers of stuff between the CPU and the

00:00:26.800 --> 00:00:31.040
cooler for it to be effective. Well take a look at this footage. That is heat on the back of the

00:00:31.040 --> 00:00:35.600
motherboard and there's enough that removing it might actually make a meaningful difference.

00:00:35.600 --> 00:00:42.720
So does it? And if it does, is it easy or cheap to do? Well that depends. Not until after this

00:00:42.720 --> 00:00:55.760
segue. To our sponsor. The simplest way to cool your CPU from both sides is to just

00:00:55.840 --> 00:01:01.280
point a fan at the back of it and that's been done before. My first PC case was the Antec 302,

00:01:01.280 --> 00:01:06.240
which had a mount for a fan right behind the motherboard tray. Did I spend 30 of my high

00:01:06.240 --> 00:01:11.680
school dollars on a Noctua fan to put it there? Yes. Did it do anything from what I remember?

00:01:11.680 --> 00:01:16.480
Kinda. So first things first, let's see if Antec was cooking or not. Noctua has this handy magnetic

00:01:16.480 --> 00:01:22.080
mount from their desk fan that we're just going to use just as a quick little preliminary test here.

00:01:22.080 --> 00:01:27.680
We're seeing about a degree or two difference going from like 51 point something to more

00:01:27.680 --> 00:01:33.360
closer to 50, 49.7 kind of area. And that lines up with the preliminary testing that we did

00:01:33.360 --> 00:01:38.640
taking measurements from the CPU and hardware info. But we all know that proper cooling requires

00:01:38.640 --> 00:01:44.640
contact. So let's see how easy or hard it is to mount a cooler on the back here. First things first,

00:01:44.640 --> 00:01:48.960
any mounting design that we make will need to make sure that the CPU cooler on the front

00:01:48.960 --> 00:01:54.160
isn't affected because spoiler alert, it's doing most of the work. And mounting hardware for CPUs

00:01:54.160 --> 00:01:58.640
is made with relatively tight tolerances to ensure that you can get even contact across the CPU's

00:01:58.640 --> 00:02:03.040
integrated heat spreader. And pretty much every part of the mounting hardware plays an important

00:02:03.040 --> 00:02:08.480
role, including the parts that go on the back of the motherboard, which are in our way. Let's get rid

00:02:08.480 --> 00:02:14.400
of it. And you know what else is in our way? The Intel socket backplate that holds the CPU

00:02:14.400 --> 00:02:19.360
retention ARM on the front. And that also prevents our second cooler from making direct contact with

00:02:19.360 --> 00:02:23.760
the back of the motherboard. So not only can we not get away with using out of the box CPU cooling

00:02:23.760 --> 00:02:28.960
parts for this, we're going to have to remove the backplate as well, which means that we're going

00:02:28.960 --> 00:02:35.760
to need to make a custom backplate. Yes. Originally, what I was thinking of doing is we could probably

00:02:35.760 --> 00:02:43.120
just modify the original backplate to add the cooler onto the back. However, as you can notice,

00:02:43.120 --> 00:02:48.320
there isn't really a whole lot of material on the ends to put the bolt holes through. I'd basically

00:02:48.320 --> 00:02:52.800
be drilling right along the edge here. So we're going to have to make a new one out of metal.

00:02:59.120 --> 00:03:03.360
Well, that's one thing I noticed. Oh, is that going to be a problem? They probably put this

00:03:03.360 --> 00:03:09.920
so there's a small thin layer insulation layer to probably make sure just in case

00:03:10.880 --> 00:03:15.360
that none of these traces or contacts. Oh yeah, because there's definitely open views there.

00:03:18.720 --> 00:03:23.200
So we plain just won't be able to get the same degree of contact on the back as we have on the

00:03:23.200 --> 00:03:28.160
front. First of all, we're on the wrong side. But second of all, this is not a flat surface like

00:03:28.160 --> 00:03:33.760
the cold plate on your cooler or the IHS on your CPU. There's a bunch of tiny electrical components

00:03:33.760 --> 00:03:37.920
that are going to prevent this cold plate from sitting flush with the back of the motherboard.

00:03:38.000 --> 00:03:41.280
What are we going to do about that? Well, we're going to need a cushioning layer that's a lot

00:03:41.280 --> 00:03:49.280
thicker than your typical thermal paste, like this one millimeter pad. Sadly, it has a worse

00:03:49.280 --> 00:03:54.160
thermal conductivity than any thermal paste. But hey, it's much better than air, which is what

00:03:54.160 --> 00:03:59.040
the fans are working with. So I'm just going to estimate it because like, is this got to be good

00:03:59.040 --> 00:04:03.520
enough? You know, that's what I always say is just got to be good enough. We decided to go with a

00:04:03.520 --> 00:04:09.600
mechanism that's just mounting into these corners that are already part of the Intel bracket because

00:04:09.600 --> 00:04:14.400
it just makes our lives easier. To mount in here, we had to make some custom length screws because

00:04:15.520 --> 00:04:20.320
I don't even actually know exactly what these holes are for on the Noxua coolers because I've

00:04:20.320 --> 00:04:25.360
never used them. I'm assuming it's compatibility with much older hardware, but I've never actually

00:04:25.360 --> 00:04:30.880
had to use them. So now I do. And it's what an exciting time for me. A new experience. Thankfully,

00:04:30.880 --> 00:04:35.440
there's Robertson screws in the precision screwdriver kit, which has an adapter that lets

00:04:35.440 --> 00:04:40.880
you lose any of those precision bits with the LTD screwdriver. Technically, the length of these

00:04:40.880 --> 00:04:46.640
screws should mean that nothing actually will screw into the motherboard, but there is a chance that

00:04:46.640 --> 00:04:53.680
like things can bow and bend a little bit. So I am going to be careful. I'm kind of trying to stop

00:04:53.680 --> 00:04:59.920
when I feel resistance out of just sheer terror because I am legitimately worried about this

00:04:59.920 --> 00:05:06.800
board existing. So let's, let's boot. If this seems simple, it's because Justin did all of the

00:05:06.800 --> 00:05:11.120
hard work, honestly. I mean, looking at it right now, we're still looking fine. It's not better.

00:05:12.240 --> 00:05:16.560
It's not worse. This is kind of interesting. So when we're looking through the thermal camera,

00:05:16.560 --> 00:05:24.720
we can see that our front CPU cooler is just much, much warmer overall than our back CPU cooler,

00:05:24.720 --> 00:05:29.440
which implies that it's moving away a lot more heat to be expected. But honestly, it feels like

00:05:30.560 --> 00:05:36.880
we haven't done much because the base of our CPU cooler is 20 degrees. So I don't think it's,

00:05:36.880 --> 00:05:41.360
it's doing much at all. So why doesn't the cooler perform better? Well, the motherboard is acting as

00:05:41.360 --> 00:05:44.640
an insulator, which slows the rate of heat transfer. And it doesn't matter how thermally

00:05:44.640 --> 00:05:50.080
conductive our CPU cooler or our thermal pad is, it can only remove the heat that reaches the other

00:05:50.080 --> 00:05:54.800
side of the motherboard. Think about it like this. It doesn't matter how many lanes your highway is

00:05:54.800 --> 00:05:59.920
if everyone has to file through the same single lane bottleneck. Support public transit. This cooler

00:05:59.920 --> 00:06:03.920
can only pull away as much heat as it's being given. But once you pull all of the heat away,

00:06:03.920 --> 00:06:08.480
you're going to be rate limited by the motherboard. So it doesn't matter how much extra cooling power

00:06:08.480 --> 00:06:12.160
you put in here unless you decide to go sub zero. And if you're going to go sub zero,

00:06:12.160 --> 00:06:17.680
just put it on the front. Don't put it on the back. I'm just going to take the fan out. I wonder

00:06:17.680 --> 00:06:22.000
if it will even like, the cooler will even saturate. Like that would be a good indicator

00:06:22.720 --> 00:06:31.920
if, if this thing is actually doing anything. 21.3. It's going up very slowly. So it is actually,

00:06:31.920 --> 00:06:34.960
we're transmitting some heat away, but you know, it's going to make this heat up faster

00:06:34.960 --> 00:06:38.960
as if I just touch it with my hand. Let's compare that to running the front without the fan. So

00:06:38.960 --> 00:06:42.400
we're looking at it right now. It's sitting at 27 degrees. Now Reese is going to hit

00:06:43.120 --> 00:06:54.720
Cinebench. And we're going to see how quickly this thing heats up. 28, 26, 29, 30. It's like,

00:06:54.720 --> 00:06:59.680
it took us like literally minutes to go up a degree before. And this is climbing

00:07:00.400 --> 00:07:03.760
a degrees per second. And eventually it will start getting to the point where it's probably

00:07:03.760 --> 00:07:08.720
thermal throttling. This just illustrates how much more important this is than whatever we're

00:07:08.720 --> 00:07:13.360
doing on the back. Keep watching, please. We decided to use our environmental chamber to

00:07:13.360 --> 00:07:18.560
have a perfectly controlled temperature for the most precise of results. But when we tested in

00:07:18.560 --> 00:07:24.400
there, every permutation of our system performed the exact same. Back panel, back fan, back cooler,

00:07:24.400 --> 00:07:29.520
stock cooler, it did not matter. Was all of our preliminary testing wrong? No, it's the environmental

00:07:29.520 --> 00:07:33.600
chamber who was wrong. To maintain temperatures within a tenth of a degree, the chamber needs to

00:07:33.600 --> 00:07:37.760
constantly blow conditioned air. And that conditioned air blowing was enough to cool the

00:07:37.840 --> 00:07:42.560
motherboard on our control system, which means that we saw no improvements on our cooled system.

00:07:42.560 --> 00:07:47.600
And that blows. We needed a more realistic airflow scenario. So we found a spot in our

00:07:47.600 --> 00:07:52.080
office that was roughly equidistant from a bunch of ventilation. We set the room temp to a tight

00:07:52.080 --> 00:07:57.600
22-ish degrees and tried again. And would you look at that? With the custom back plate and a fan,

00:07:57.600 --> 00:08:01.840
we see CPU temps drop one to two degrees. Heck yeah. And with our cooler mounted to the back,

00:08:01.840 --> 00:08:09.280
we see a further improvement of one or less degrees. It's not nothing, but it is almost nothing.

00:08:09.280 --> 00:08:12.720
Oh, and we've kind of been neglecting to tell you about the results that make this whole endeavor

00:08:12.720 --> 00:08:17.520
look even more foolish. When we compare CPU temps with our front mounted cooler on its

00:08:17.520 --> 00:08:21.840
stock mounting hardware, instead of our custom back plate, it performs better than our back

00:08:21.840 --> 00:08:26.880
mounted cooler setup. So all the haters at the start of the video who said this won't work,

00:08:26.880 --> 00:08:31.040
shut up. It does work. It just works poorly. Maybe you could get better performance out of

00:08:31.040 --> 00:08:34.800
a more precise custom mounting mechanism. But why bother when you can just do this?

00:08:37.040 --> 00:08:41.120
Or maybe you could just segue to our sponsor. Thanks for watching. What do you think? Was

00:08:41.120 --> 00:08:44.640
there something that we missed, something that you could have done differently? Let us know in

00:08:44.640 --> 00:08:48.240
the comments. And if you liked this video, why don't you watch a cooling solution that did work

00:08:48.240 --> 00:08:49.680
like our AC cooler?
