h a l f b a k e r yI CAN HAZ CROISSANTZ?
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Many things need cooling, including this computer and all
other computers in the world. Computers can be tricky in
this capacity as 30-350W can be generated in an area
equivalent to a couple of coins.
To dissipate this heat, a heatsink is used. This is usually a
block clamped to the processor
made of something
conductive, like aluminium, and with an array of fins to
vastly increase the area in contact with the air. This is all
fine and dandy until the air between the fins heats up. It
will set up convection currents and exchange with cooler
air, but this happens slowly. In response, engineers add a
fan. This blows cool air through the heatsink. Done. Except
this isn't perfect, there are dead zones where warm air sits
undisturbed and there is an insulating boundary layer of
warm air coating the fin surface. So, let's remove it.
Take a standard CPU heatsink, say an 80mm cube of
aluminium extrusion with fins set 3mm apart. The
conventional approach would be to clamp an 80mm fan to
the top blowing air down through the fins. This leaves a
dead spot under the fan hub and little air velocity right
down in the bottom of the slots created by the fins.
Instead, lets rotate the fan 90 degrees so the axis of it's
rotation is parallel to the fins of the heatsink. If we make
fan blades thin enough it can be arranged so that they
move through the slots between the heatsink fins sweeping
the warm boundary layer from the surface and carrying it
out of the slot.
It might also be useful to angle the blades so that there is
net axial movement of air and holes through the fins could
promote air movement down the length of the heatsink.
I've done a quick drawing <link> of the general concept on
the most default type of heatsink. I can imagine a few
alterations that might improve things. For example, a
general cage to enclose the blades and stop them eating
cables. Matching the radius of the upper surface of the
heatsink base to that of the rotor (difficult to make as an
extrusion). Two smaller adjacent rotors could work better
with overall lower volume.
Bad drawing of idea
https://imgur.com/a/sgcwciG
[bs0u0155, Dec 10 2020]
Slightly More Sophisticated, with added centrifugal flow
https://imgur.com/DH2Zj0k
[bs0u0155, Dec 11 2020]
[link]
|
| |
So, you're adding 80mm to the thickness of my laptop? |
|
| |
Take a look at how server CPUs are cooled. Multiple fans, horizontal ducting, high mass flow rates. Noisy - because the fans don't run in phase, and there's a lot of turbulence - but efficient. |
|
| |
That's a good bad drawing. |
|
| |
//how server CPUs are cooled.// |
|
| |
There's a lot going on there. The convention is front-to-
back flow through the whole unit, and vertical dimensions
are often constrained by 1U/2U etc form factors. The
front to back flow in a rack is already ducting, simply by
separating the room in two. That's not really possible with
standard box PCs |
|
| |
A quick peek into the Dell 5820 workstations we have
around here shows fairly basic ducting, there's shrouds
around each CPU heatsink with a fan sucking air through.
It's not great design actually, the front CPU exhausts
directly into the intake of the rear. The PSU is in it's own
compartment, and the RAM is shrouded with positive
pressure from the front of the case. There's a total of 8
fans, it's not particularly quiet or cool under load, but
that might just not be that big a deal for them. |
|
| |
In the enthusiast area, there are plenty of massive
3x120mm fans with multi heatpipe heatsinks, that this
would be similar in size to. |
|
| |
What I'm wondering is if selective mechanical sweeping of
the boundary layer might be an efficient system. If the
purpose of a duct is to supply the coolest possible air to
the hottest possible surface, then this operates as a
moving duct, the hottest layer of air is pushed in front of
the blade and behind it the cool air is directly introduced
to the fin surface. |
|
| |
Having the fans in phase would be a bad thing from
a vib perspective. |
|
| |
There's a temperature differential between the thing to be cooled and the environment. |
|
| |
While there's no "something for nothing", is there no way of employing this "waste" heat to actually enable the cooling ? |
|
| |
The CPU package needs to be no hotter than 30C (say). So, select a volatile coolant with a BP of less than 30C (Isopentane is 27.8). Use the vapour to drive a tiny turbine, which spins a fan, which pushes ambient air over a cooling labyrinth used to condense the vapour itself back to a liquid, which then is pumped (again, by the energy from the turbine) back into the reservoir that feeds the evaporator. |
|
| |
More of a "performance art" device than a practical way of cooling a processor, but it would probably function, because it's not really trying to extract usable "work" from the temperature differential. |
|
| |
//no way of employing this "waste" heat to actually enable
the cooling ?// |
|
| |
I'm surprised the ducts are not designed a bit better, they
seem to be constant area, except when you get to the fan
on the back where the round hole in the square duct
actually reduces area. You could get a little flow for free
using the "Meredith effect". But that would require a little
care and finesse, probably expense because constant sizes
are cheap. |
|
| |
Now, if you put shutter vanes on the cold end of the duct, and have a long, open tailpipe, is it possible to produce something akin to a tiny Argus pulsejet ? |
|
| |
That would be interesting. And very noisy ... |
|
| |
With the power (and therefore heat production) of
personal computers, I'm surprised that "proper"
refrigeration-type cooling hasn't become standard. Water-
cooled is becoming popular, but even that is only "getting
there" in that it's still only passive cooling.
On the other hand, in these days of high-speed wireless
connections, why not build the whole processor etc into a
multi-vane type shape, and spin the whole computer? The
only bits that don't need to move are the bits that the
human interacts with, which can be easily accommodated
by Wi-Fi or Bluetooth or... |
|
| |
> I'm surprised that "proper" refrigeration-type cooling hasn't become standard. |
|
| |
Condensation. It's possible to seal every bare wire, but expensive. Especially when you want to allow arbitrary expansion via slots and ports. |
|
| |
[8th of 7]'s and [neutrinos+shadow]'s ideas are just about crazy enough to work, and could even be combined ... just add a directional WIFI antenna and RADAR becomes a software problem. |
|
| |
// heat sink that used agitators between the fins // |
|
| |
Were they held in position by crocodile clips ? |
|
| |
OK smart alec so tell us when fins evolved legs. |
|
| |
Abut the same time that the other Scandinavian races did, presumably ... |
|
| |
An air bubbled, sand bed boxed there on the CPU, with neon LED? |
|
| |
Not a liquid but a lot of sweeping of boundaries. |
|
| |
The grain would have to be the right shape, mico shuttle cocks. Got to retain the fins. Although, Not too vigorous air flow or upgrades will be more frequent. |
|
| |
// snacked on some tourists // |
|
| |
If they were eaten by cartilaginous fish instead of reptiles, would that count as Shark's Finn soup ... ? |
|
| |