The WORST CPU Cooler EVER – Sketchy Heatsinks Ep2
Linus Tech Tips
·Linus Tech Tips
·2017-05-06
·
1,547 words · ~7 min read
0:00
So, you might have seen our last attempt at making a heat sink where we took a
0:05
reciprocating saw to a block of aluminum with what could best be described as
0:11
limited success. But, of course, persistence is the name
0:16
of the game here at LMG. So, we went through the YouTube comments which were
0:22
full of suggestions about casting our
0:26
own heat sink from molten aluminum and
0:30
decided to set about doing it. But is
0:34
that really the kind of thing that you can do with $100 worth of supplies and
0:39
an empty warehouse? Well, there's only one way to know for
0:43
sure.
0:54
Browse privately and securely with Tunnel Bear, the simple VPN app. Try
0:59
Tunnel Bear for free at the link in the video description.
1:03
So, casting then. I guess we'll need a
1:07
couch to No, we're talking about the
1:11
kind of casting where a molten material is poured into a mold so that it
1:16
solidifies in the desired shape. There are a couple of important elements that
1:22
you need. A foam or plastic replica of
1:26
the intended finished design and a material that you can mold around it
1:32
that can hold its shape. The primary
1:35
design considerations for our heat sink
1:38
were how easy it would be to cast and
1:41
how much heat it could dissipate. We kept it as simple as possible in Solid
1:47
Works with the two pieces coming off the heat sink being to allow for the metal
1:52
to be poured into one end and the gases to vent out of the other. And the mold
1:58
had to be orientated in such a way that the bottom would be the most likely
2:03
surface to have a smooth, flat finish.
2:07
Although the fins ended up thicker than we'd have liked, when it comes to heat
2:11
sinks, size matters, and we calculated a
2:14
theoretical heat dissipation of over 200
2:17
watts. So then we went ahead and printed
2:21
our design using PLA plastic at 103%
2:25
size to account for the shrinkage of the aluminum during cooling. We chose PLA
2:31
because it's made from corn, so when it burns, it shouldn't poison us.
2:37
To improve our odds of success, we needed castingrade aluminum. But buying
2:42
that stuff new is pretty expensive.
2:45
Fortunately, there's a large source of casting grade aluminum, the scrapyard.
2:51
So, we headed there and picked out our victim, extracting the engine out of
2:55
this 2002 Toyota Yurus. Back in the
2:59
shop, we took apart the engine, cleaned off the dirt, and chopped it into pieces
3:03
small enough to melt. We used foam then
3:07
to extend our pouring and venting holes.
3:10
Then we mixed up clay, sand, and water
3:14
until it reached a consistency where it was pliable but able to hold its shape.
3:20
This mixture called green sand was packed around the 3D print and then
3:25
baked for a couple of minutes to partially melt the plastic and to harden
3:30
the clay. This whole process is called
3:33
lost PLA casting. We then placed the
3:37
aluminum scraps that we had into a graphite crucible, which is basically a
3:42
a large cup that can withstand really high temperatures, and set about melting
3:47
the metal in a crude furnace that we made out of fire bricks. One of the best
3:52
and worst things about aluminum is that when it's exposed to air, it almost
3:57
immediately forms an aluminum oxide layer on the surface. This is good for
4:02
preventing further corrosion, but it
4:06
makes it very annoying for melting. So,
4:09
it's important when casting then to remove the aluminum oxide layer that
4:15
floats to the top called dross so that it doesn't contaminate the finished
4:20
cast. With the dross removed, we were ready for our first pouring attempt.
4:24
Though it should be noted we weren't sure how much of the PLA had actually
4:29
melted on this one. Turns out
4:33
not much. Clearly the aluminum barely
4:37
made it into the mold and wasn't able to
4:41
melt much of the plastic in the process. So to troubleshoot, then we made a
4:46
different riser design out of foam so that the aluminum would be able to enter
4:51
across the entire top of the mold. Green
4:54
sand melt, purify, and pour again. And
4:59
we ended up once more with a face only a
5:03
mother could love. So, we moved on then from lost PLA
5:08
casting to lost foam casting, which is an extremely popular and much easier way
5:14
of creating molds. Due to the low melting point of foam, when the aluminum
5:20
is poured in, it immediately melts it away and fills the space. In theory, our
5:27
foam mold wasn't perfect by any means, but we were cautiously optimistic that
5:32
it would work. So fire melt and pour
5:35
again and we get to see what kind of a
5:38
result we were going to get.
5:42
Unfortunately, this wasn't a massive success either.
5:48
Although the foam did melt away and we
5:51
were able to create a good base layer
5:55
and even where the mounting plate would be was fairly well done, we simply
6:00
didn't have enough static pressure to force the molten metal up into the fin
6:06
shapes. We could solve this by creating
6:10
a much taller riser. But by this point,
6:13
the time crunch had become real, and our engineering department realized that he
6:17
needed to produce something usable or I
6:21
might melt him down. So, he made a clay
6:24
and fire brick sandwich and assembled a mold using a method that bears more
6:31
using a method that bears more resemblance to grade one art class than
6:35
to precision crafting.
6:39
To ensure on this the final attempt that
6:42
the metal would thoroughly melt, the furnace was remade and heated until the
6:48
entire crucible was red hot. Upon
6:52
pouring the aluminum into the mold, it started bubbling from the steam escaping
6:57
from the clay, making the metal of the finished product more porous, a problem
7:03
we hadn't had before when we were using green sand. So then with a bit of
7:09
massaging from the reciprocating saw to make it fit on a motherboard,
7:16
it was finally ready to mount.
7:19
The processor chosen was an Intel Q6600,
7:22
mainly because the old ASUS motherboard it was on was the only one where the
7:27
heat sink didn't interfere with the VRM.
7:30
And to help fill in the air pockets on the bottom of the heat sink, we added a
7:35
more than generous dollop of thermal paste. Topping off the creation with a
7:40
handful of zip ties.
7:43
And this was actually the first time I
7:48
laid eyes on what three of my staff had
7:52
been doing all week.
7:56
I don't know how anything could have been worse than what I was expecting,
8:00
but this may have achieved it.
8:07
I mean, the name of the series is
8:10
sketchy heat sinks, and um
8:14
I guess mission accomplished there. With
8:17
all of that said, testing an IDA 64
8:20
turned out better than expected. And sketchy heat sink number two was able to
8:25
keep the CPU reasonably cool for a fairly long time. Although this is
8:30
likely more to do with its large thermal mass than its ability to dissipate heat.
8:36
Ambitious but rubbish. 10 out of 10 IGN.
8:42
So yeah, in the end this turned out to be the no man's sky of heat sinks. Sure,
8:48
we had high goals set and we believed with the resources we had, we could pull
8:53
it off. But it took three times longer than originally planned and the end
8:57
result performs poorly, looks bad, and has almost none of the features that we
9:02
said it would have in the beginning.
9:05
So then, all that's left to say is thanks in advance to the metalworking
9:09
experts who by now have filled the comments with advice. We will bear it in
9:13
mind for the next installment then of sketchy heat sinks.
9:18
Freshbooks is the small business
9:21
accounting software. It's built from the ground up for how you work as a
9:27
freelancer or a small business owner. And it's the simple way to be more
9:32
productive and organized. Not to mention to get paid more quickly and on your own
9:38
terms. You can take deposits. You can get paid faster. You can create and send
9:42
professionallook invoices in less than 30 seconds. You can track your expenses.
9:48
You can even see when your client has
9:51
seen your invoice to put an end to that, oh, did you send it? No, I didn't see
9:55
it. Blah blah blah nonsense. And you can even take your payment directly through
10:01
FreshBooks. If all that sounds awesome and you want to stop dinking around with
10:05
complicated accounting software, then try FreshBooks for 30 days for nothing
10:10
at freshbooks.com/te tips and enter tech tips in the how did
10:14
you hear about us section. So, thanks for watching guys. If you
10:19
dislike this video, hit the dislike button. But come on, where's your sense
10:22
of fun? But if you liked it, hit the like button, get subscribed, maybe
10:26
consider checking out where to buy the stuff we featured. I mean, what would that even be? like scrap aluminum at the
10:32
link in the video description.