Slashdot Mirror
Startup Offers Peltier-On-Chip
LowSNR writes "The South Carolina based startup Nextreme, Inc. is developing technology to put Peltier Coolers in chip packages, according to an Ars Technica report. The tiny coolers could be situated on top of local hotspots on the die and pump heat away through a package pin to the motherboard. Also, the Seebeck Effect allows the waste heat to be used to be harvested to generate/reclaim power."
That's cool.
Har har.
I think the heat recycling aspect itself shows some promise, and the design being built into the chip, but we have to consider that the same kind of designers that use these are just as likely to push the chip beyond the capabilities in terms of the total heat reduction for the system.
People who tend to overclock or use overclocked chips, will frequently push the envelope even further if they think they can get away with it.
-- Tigger warning: This post may contain tiggers! --
I imagine this would only solve the problem for the first millimeters , you'd still need a pretty big heat sink somewhere. Would this open up the door for really high frequencies or would it just optimize things a little ?
Call me when they put a miniature Sterling engine on a chip and use it to recharge the battery.
Can I use the excess heat from my P4 to produce the energy for my p4?
Only if you use IPv6 on the motherboard and IPv4 on the daughterboard.
-- Tigger warning: This post may contain tiggers! --
Someone let him out!
Gee I had this idea about 20 years ago, & I suspect many readers here also had a similar sense of deja vu when they read this! Oh well, more power to them...
Peltier devices on-chip have been used for a while, whenever temperature variations are intolerable. Some examples: Analog Devices AD595 thermocouple amp, which uses in-chip thermal calibration to ensure a cold junction of known temperature, and many voltage regulators and switching supply controllers that use temperature-controlled bandgaps as their voltage reference.
Nostalgia's not what it used to be.
In theory, yes; the chip is hotter than its environment, so you can put a heat engine between them and generate energy. The maximum theoretical efficiency of this process is given by Carnot cycle and depends on the heat difference between the processor and the environment and the temperature of the environment. With current processors you can't really exceed 60 degree Celsius, or 333 Kelvin, and the environment is typically at 20 degree Celsius, or 293 Kelvin, so the maximum theoretical efficiency is around 12%.
Of course, if you could find more durable materials, you could just insulate the processor, let it heat up to a thousand degree Celsius or so, and get nearly 77 percent efficiency. The hotter you run the processor, the more efficient the system becomes; a hypothetical plasma-state processor at 10,000K would give a theoretical efficiency of 97%.
It would also give a whole new meaning to "flamebait" ;).
Forget magic. Any technology distinguishable from divine power is insufficiently advanced.
SuperSoaker.... Make a nukulerized mini-soaker/quantum heat sink and call it a physical feat. Snatch defeat from the jaws of victory... shunt the heat to another dimension...
Previously: "Linux... Toward the Sunrise..." Now: "Linux... Toward the-- No, now, part of Every Sunrise"
The article says it's in North Carolina, not SC. Much as I wish it were here...
I confess this is not my area of expertise so I have my alabaster underpants in readiness for a good flaming.
But does this mean they can cool the chip without the heat sink/fan combo, or will they still need some method a pumping the heat around the chip to the areas that can process it. I mean comparatively the chip is quite big and we are only looking at one top layer of it surely?
I can hear my computer whirring as I type, anything that offers hope to get rid of that noisy thing gets my vote.
As cool as it is to use the peltier effect to cool chips, in no way is the "reclaimed heat" going to be enough to generate any significant amount of power.
I don't mind companies spouting marketing drivel, because that's what companies with marketing departments do. But this whole "fake green" thing that's going on recently has got to stop.
Interested in a Flash-based MAME front end? Visit mame.danzbb.com
http://media.arstechnica.com/news.media/thermalflow.png
The picture shows they're dumping the heat into the PCB
What I'm not really clear about is how that helps the cooling situation.
Sure, you're moving heat from the CPU/GPU die, but not to a heatsink...
So what's the point?
[Fuck Beta]
o0t!
The PCB is often used as a heatsink for SMD chips. It wouldn't work too well for a CPU, unless we adopt a new form factor with a surface area of 5,490 cm^2, or roughly a 2.5' square (and it would need a lot of heatpipes to make up for the extra distance between the CPU and the edges of the heatsink).
"In theory, no;"
fixed it for you.
I don't care how hot you get the chip, you can not get to 100%.
The Kruger Dunning explains most post on
Well, not really. Even an ideal heat pump could only push the energy from here to there. More realistic ones however produce some heat of their own.
That's right. Your fridge doesn't just move energy from inside to the radiator outside, it actually dissipate a little extra heat of its own. If you left the fridge door open, it would actually heat the kitchen a little. An air conditioner, ditto, that electricity it uses has to go somewhere, which means heating the outside air. It's no different for a Peltier. They actually produce quite a bit of heat in the process.
The only value of a heat pump is that it can pump it against the natural flow. It can move some heat from a cooler place (e.g., the inside of the fridge) to a warmer place (e.g., a warm kitchen while you're cooking.)
The Peltiers used for overclocking are, basically, used as a fridge. They can get the chip cooler than the heatsink (which actually becomes hotter as a result), and, with a powerful enough one, even below ambient temperature. Which is good for overclocking. Downside, if you get below the ambient air temperature, you get condensation, and electronics tend to not like that much.
Here they seem to want to use it just to move the heat somewhere else.
E.g., you have a really high-clocked piece of circuitry generating a lot of heat (e.g. the ALU), next to a piece of circuitry which doesn't switch as often (e.g., the cache), you can help more heat flow from the former to the latter. Downside, the chip as a whole _will_ become hotter.
E.g., at least according to the summary, they want to pump the heat to a pin to the motherboard, although, to be honest, I'm not sure exactly how much heat will a pin dissipate. If that was that great a way to get a lot of heat off the chip (remember, it must also remove the Peltier's own heat), we'd be doing that already instead of using big heatsinks. The pins aren't even going into the die anyway, but are wired to it via some more modest traces.
A polar bear is a cartesian bear after a coordinate transform.
... Cool Ranch chips?
That is for the circuitry itself. But you can maintain a much hotter zone away from the sensitive parts. In fact, isn't that exactly what a peltier cooler does?
"THERE IS NO JUSTICE, THERE IS ONLY ME." -Death
Let's see, "cooler" is slang for prison, Leonard Peltier is being held at Leavenworth, so this company is offering little Leavenworth models in potato chip packages? Excellent.
I once thought of tossing one of these thermoelectric (TEC) coolers in my system to try and help with waste heat, but the trouble with the seebeck effect is that it reaches a cutoff point fairly quickly. You see, the "cool" side of the tec doesn't remain cool, it heats up as well meaning that the temperature difference grows gradually smaller and smaller, which in turn reduces output of voltage and becomes a vicious cycle, building up more and more heat on the hot side.
TEC's are great when using the peltier effect since the electric side can be managed easily, so long as there is a way to pump away the waste heat. But the seebeck effect is not a solution to waste heat, not unless there is a breakthrough in efficiency i'm not aware of.
Do peltiers still suffer from occasional reversals where the sides flip cooling/heating?
I read somewhere many years ago that the problem with peltier coolers is that they are active devices and as such are susceptible to failure. When they do they basically become an insulator rather than a conductor of heat and you could fry your hardware. At least with heatsink + fans if the fan fails the heatsink will still perform some passive cooling and the chip will heat up gradually and your temp sensors would have a bit more time to react. In fact your fan monitors would sense the fan failure even before that.
You want fun, go home and buy a monkey!
Hmm, generating electricity from temperature differences between the hot end of the heat pump and the environment will make the hot end hotter than it would be with the system designed to simply dissipate heat as quickly as possible. This will make the heat pump consume more energy in the first place. Given that the actual efficiency of recapturing the energy is likely to be far less than the carnot cycle maximum, I wonder if this will ever be worth the trouble.
Isn't recycling heat called "thermal runaway" and the last thing you want in a processor :-) ?
Take Nobody's Word For It.
I was talking about a chip made from some hypothetical material which can withstand high temperatures and thus wouldn't need to be cooled. You would let it heat up as high as it goes and then let the heat run an electric generator to feed energy back into the batteries, thus reducing total drain. In fact, if the material can withstand high temperatures, it would make sense to put thermal insulation around it to get it as hot as possible, because this would increase the efficiency of the electricity generation.
Forget magic. Any technology distinguishable from divine power is insufficiently advanced.