The new forums will be named Coin Return (based on the most recent vote)! You can check on the status and timeline of the transition to the new forums here.
The Guiding Principles and New Rules document is now in effect.
Supercool Computers & Legos
Mad_Morlock
Registered User regular
Registered User regular
I've got this design for the basics of a new solid state cooling system.
I had to build it out of legos, but it's designed with the peltier effect in mind.
Thermal Processing Unit
It would beat in time with the clock of the CPU like a heart. It literally pulses heat away from the core to maintain a stable ambient temperature. Search for 'peltier current pulse' on Google for more information on how to clock a peltier for maximum heat transfer over distance vs cooling provided by a steady state current flow.
Anyone out there know enough about cooling technology to tell me why this isn't feasible?
-M
I had to build it out of legos, but it's designed with the peltier effect in mind.
Thermal Processing Unit
It would beat in time with the clock of the CPU like a heart. It literally pulses heat away from the core to maintain a stable ambient temperature. Search for 'peltier current pulse' on Google for more information on how to clock a peltier for maximum heat transfer over distance vs cooling provided by a steady state current flow.
Anyone out there know enough about cooling technology to tell me why this isn't feasible?
-M
Mad_Morlock on
0
Posts
That being said, looks neat.
edit: Thread signing is frowned upon.
Edit edit: Looks a bit overly complex to me. It's difficult to control heat flow over a length of pipe that long.
And the size isn't the issue. I had to built it out of legos just for the layout.
You've seen the layout of traditional peltier coolers? Like the ones found in the mini-fridges?
why dont you post more supplemental material (photos/info) for us to understand what you're talking about here
The peltier effect was discovered about 150 years ago. Its a basic quantum effect that moves heat from one side of a junction to another.
Traditionally, peltier devices have been constructed with all junctions facing the same direction. This creates one device with a hot side and a cold side.
There are several problems with using this for cooling solid states systems, but most revolve around design issues which I think I found the solution to... provided you can view the entire process from another angle.
However... water cooling, from a physics standpoint, is a step backwards. Using a mechanical cooling solution to cool a solid state system is a mismatch. Solid state is just suppose to be that. Solid. No moving parts. Electricity in and results out with the only real movement taking place on the quantum level. Thats what solid state is supposed to be. I'm saying I think I found a more elegant solution that the very physical laws of the universe seems to have built into the design of transistor based computers.
I thought about patenting... but I'm poor. So I figured I'd give it away anyways.
As for people thinking about it already... I've been beating bushes for the last year trying to get somewhere. I've even been talking to people that used to work for Intel. There's nothing out there like it.
Yes... you still need a heatsink to dissipate heat. But the peltier effect of the circuits moves the heat to the outer shell of the device, keeping the CPU cool, while providing for a larger radiant surface area. This means that heat will dissipate faster. Use an ionizer instead of a fan and you're completely solid state.
Seriously though, why would you want to try and remove heat from the sides of a CPU?
[edit] I mean, why not just use a conventional peltier and fit a bigger heatsink? How are you getting the heat from the top of the chip (largest surface area, and cools the chip equally rather than say concentrating heat in the centre) to your distribution device?
The problems usually lie in that peltiers take a lot of energy to power (read: really inefficient), and that the needed extra power supply with the size of the system itself usually means you have to have at least a full-sized tower to hold it all.
It's a cool idea (hurr), but it usually isn't much better than just throwing on a better fan and making better contact with the heat sink via thermal paste.
I've already been over the math of it all down at the college. But can't get anywhere with it in Labrador.
If you've got a heat source with a radius of 1, and you translate that heat outwards to a structure with a radius of 10, your radiant surface area increases by a factor of 100. That means you've got 100 times more space to dissipate heat from.
Jeak: That's what heatsinks do. Remove heat from the side of a CPU. I'm just saying use all the sides. Remove heat in every direction you can.
Jeak it's a model.
But in 2002 it was discovered that if you use a current pulse (like a clock for your CPU) you can maximize the efficency rating of the junctions.
However... this wouldn't do anything for the current device layouts. So something new has to take advantage of this discovery.
Peltier Current Pulse
xzzy: Yup. Complete encasement. Box within a box. Each box designed to move heat out to the box that surrounds it.
Or you could just use a standard finned heatsink and get an even bigger dissipation area without going to all the extra trouble? Besides which, if you just cool from the sides, the height of the average silicon chip is a couple of millimetres, so you're getting significantly less surface area to remove the heat from the chip initially. If you use the top as well and just cool radially, you'll get a heat gradient in the chip, which can't be good for it. And if you cool vertically as well as radially... why bother? The extra cooling from the edges is going to be close to zero, it's still going to produce a mild temperature gradient and with a decent heatsink you're probably dissipating all the heat you're generating comfortably anyway.
You can't use all sides of the CPU. The CPU is flat. You'd have to make a heat sink to attach to the top of the CPU, then remove heat from the heatsink.
My comments:
(a) wouldn't this work better if it were completely round instead of square? You have to work off a heat sink anyway, why not start round?
(b) maybe for added insane factor, you could create a heatpipe, and then instead of a radiator at the top, you could attach a series of these peltier rings along the length of the pipe?
My idealized design of the completely assembled unit would look more like a sea urchin.
But if you're basically just slapping a block of copper on top and then attaching peltiers round the edge, you're still going to get uneven cooling and I still don't see how it's significantly better than just using the standard setup with a good heatsink...
But in applications where you couldn't have moving parts, or where moving parts are just too unreliable.... special applications, computers in extreme environments.... If I'm interpreting it right... he might have something here.
I'M A TWITTER SHITTER
First... stop thinking about it as heat. Instead, consider it an atomic vibration.
As a processor 'vibrates', a heat sink will dampen this vibration passively by allowing the atomic vibration to travel up it's fins and into the air beyond.
This is a purely passive system though.
Completely encasing a processor in peltier junctions would thermally isolate it from the heat sink. However, each junction operates as a vibration dampener. They move the energy of the atomic vibration to the opposing ends of their own junctions for an actuve net dampening effect instead of a passive one.
From the outside, heat would still be dissipated through the heat sink, but internally the CPU would be kept at a stable operating 'vibration'.
One question: Is there anyway to keep this outside of the field of patents? I tried declaring it ART, but the forum nazis in the Art forum shot me down. I've put out too much information about what I've been working on over the last year to ever be worried about secrecy. But I wouldn't want some corporation running roughshod over everyone else because they were the first to the table with a patent for this very basic idea.
Uh... right. I know how heat works and I know how peltiers work. I just don't see how the system you're proposing is a useful improvement over a normal peltier and a large efficient heatsink.
A normal peltier device produces a temperature difference between the two sides of the element. Like an on and off switch. Reverse the current and you reverse the switch. It relies on a steady state current flow that is always on. While operating, the current flow of the system forces electron evolution across each junction in a mirror image action. As the electron-hole flow created by the current alternately evolves and devolves across the junction, each become hot-carriers for the energy that will be dumped at their respective ends at the normal conductor as 'heat' or atomic vibration. However, during the transition of the junction, its not really what I personally would define as heat. Because of the way it evolves and devolves through application of electric current I'd be more inclined to call it vectorized entropy.
(Successfully generating a special case that thoroughly violates the 2nd Law of Thermodynamics.)
Rotating the peltier into orientation with the transistor would let you work with a new geometry. Even the circuit I've laid out is incredibly primitive compared to what I see it becoming. I can imagine it as the seed in a fractal tree that should grow not only into a heat circulation system, but could possibly provide the basis for quantum computing as well.
For solid state physics, transistors have become an analogy for electricity. It's either on or off. You need an on, off and a superposition of both to generate the basic three states for quantum computing. Electricity can't be on and off at the same time though.
But using transistors and peltier junctions together in the same lattice should let you generate a much larger variety of states. You could have an ambient or off state, and on state, an on twice state, an on three times state, up to the design tolerances of the processor. Moving and reading energy vibrations over time in the solid state lattice itself.
But thats just my qubits.
Ahem... back to the previous point. Modern peltier devices operate inefficently using a steady state current flow. Their maximum efficency levels are reached using a current pulse instead. Taking this into account it can be inferred that they are better between at moving a packet of energy over a distance during a unit of time than they are at maintaining a continuous energy difference over time. Designing your peltiers to move work heat as far away from the processor as possible as soon as it's formed seems like a better idea.
You seem to be talking about two different things at once. Yes, ok, pulsing peltiers makes them work better. What's that got to do with your crazy taking-heat-out-of-the-sides idea?
Right now peltiers are just designed to take care of one surface area. Intergrating them on both sides of the processor and around the perimeter would maximize the dampening effect by using it on the entire surface area of the hot body. Instead of only 100 square mm of space, your peltier would be acting on 240 mm^2. Moving heat outwards at that point by 1 cm would provide you with a block 11mm*20mm*20mm for a total radiant surface area of 1680mm^2.
Beyond rotating the peltier 90 degrees, it's just simple math.
Putting aside the intractable problem of how on earth you'd dissipate heat off the bottom of a chip, particularly given that that's where the pins tend to be, I'm still utterly unconvinced that the extra area you're gaining by using the edges is going to make a big difference, and I certainly don't understand why you're not just using fins to increase your surface area.
Another problem you will have from creating a multistage TEC like that is that the outer stages have to handle the heat from the CPU itself plus each of the layers' self generated heat below it. Essentially the extra thermal load reduces the total temperature differential across the device. In your design the lower layers are also smaller in surface area than the upper layers. Therefore, the amount of heat transfered from the CPU is limited the the amount of heat "throughput" that the lowest layer is capable of.
These are complete separate issuses from whether pulsed TEC's are more efficient and ignoring arguments against the eventual 3D design.
Er, the last time I looked, the majority of the underside of modern CPUs is kind of
full of pins. You want to stick heat sinks there as well?
OK. Here's an extreme environment
http://www.nsbf.nasa.gov/
The computers on board these payloads go into effectivally vacuum. No air
to conduct/convect away the heat from the CPU. So they use fanless, low power
single board computers, for example,
http://www.ampro.com
Some heat sinking is necessary, and they get warm (~70 deg C), but these computers
survive in this environment without moving parts.
I'm not sure what advantage over a big heat sink and or copper piping to a radiator
this method gives, apart from being a power hog.
The last time I checked the, pins didn't come out of the bottom of any processor that plugged into a socket on a motherboard. They come out laterally first from the processor before becoming pins or bearings which plug into the socket.
Having said that, this kind of redesign would necessitate a redesign of the pin-socket assembly. You'd be heading back to something closer to the old slot designs to move the processor off the board to cool both sides. Thats nothing new.
As for the power requirements for a peltier to function, they move a larger amount of heat than is required to power each circuit, provided there is an active heat source to move heat from. You're not trying to achieve anything lower than ambient. You're not trying to generate the 70 degree temperature difference in one stage that you'd see in a traditional solid state cooler. But then if the unit is completely sealed you also never would have to worry about condensation issues. If each successive stage moving outwards from the processor is more powerful, it should be able to handle the load from the innermost stages more easily. Like building an inverted pyramid of peltiers that create the larger radiant surface area. The whole processor would have to be designed from the ground up to be a self-contained unit.
This should make the whole system work in atmosphere, yes. An ionizer would make an apropriate replacement for the fan and would make the entire apparatus completely self contained. A solid state brain, heart and lungs.
So how big does this need to be? Is it going to bulk up latops even more?
If you want to convince people, then actual numbers and calculations will help a lot.
Clearly laid out, with temperature drops, voltage/power requirements etc.
I built the thing out of legos on a theory just to give a 3 dimensional representation of the circuit.
But based on what I know, and everything that I've read, it should require less power to move heat from the active source using this method that it would to create the temperature drop you see in current peltiers.
A peltier on each side would already halve the power required by each junction to transport the heat because it's being moved in two different directions. The peltier units just need to be used more efficently without attempting to drop the core temperature below ambient.
I really don't know how well it'll work with laptops.
But then what would you do if you had a quantum laptop?