Slashdot Mirror
Microcoolers Could Change Processor Design
Skaven writes: "Nature.com is reporting about these nifty new microcoolers, tiny thermoelectric heat sinks that can be built directly onto CPUs. Using the new technology, scientists cooled a processor at 100 degrees C by 7 degrees. That's still a fried t-bird, but what this means is that if the technology gets good enough, cooling chips could soon be getting a lot easier. If anything, small 'hot spots' on the CPU could be avoided by strategic placement of microcoolers, thus helping all of us overclockers out. Heck, maybe even increasing the voltage to your CPU would make it run cooler...how weird would that be?"
the standard overclocking tool, exactly how?
This is really cool! (no pun intended)
This has some really far-reaching effects. Where heat was previously one of the prime concerns, it will become less so. I've heard stories of supercooled Pentium II's overclocked to around 1GHz. This could mean an instant increase in processor speeds, without any changes in the actual design. R&D, baby. R&D
Inconceivable!
I adds a lot of waste heat too. It would be funny to see the web farms have to upgrade their air conditioning plants because their chips require on-board heat disposal. A double whammy. Dissipate an extra 7C, but spend 200W to get it!
--
It sounds from the article (which was lacking in technical detail), like the microcoolers can chill the portion of the chip they're in contact with. Okay, I'm good with that. But where does the heat go?
Assuming that it's redistributed, what we're really looking at is a way to take that 1GHz+ CPU and let it run nice and cool while we fry everything else inside the case, right?
A few months ago I saw this article. It concerns making water run up hill so that micro coolers such as these can work in low gravity and zero gravity environments without the need of pumps.
Am I the only one amused that the lead researcher on the project to develop microcooling for electronics is named Xiaofeng Fan? It just happened to catch my eye. :)
If I could only live my life with my threshold at 4...
In a room temperature environment just about any thermal conducting material, even really small ones, will bring something that hot down that little amount. A circuit has to produce a lot of heat to stay at 100 degrees, even flowing air would drop the micro controller down a few degrees. Lets see some tests at closer to room temperature, and then I'll believe in the product.
It is interesting science, though, and makes me wonder if this will lead to efficient cooling devices for non-computing applications. For example, if this were made very efficient, chair-rail air conditioners could become possible (and low-noise too!). Me, I'll wait until the next breakthrough before shouting triumphantly.
--
We may not imagine how our lives could be more frustrating and complex—but Congress can. – Cullen Hightower
Actually, there was a whole story on this thing, I think, here:
http://slashdot.org/articles/01/01/23/1350208_F.sh tml.
Originally New Scientist had a story on it (here), and now it looks like it made it into Nature.
I guess it must be officially "cool" now.
but we will not likely see it next year... it will take a while.
"It is a greater offense to steal men's labor, than their clothes"
And if you read the very next sentence, you would see the part where it said "To be commercially useful, these devices will have to perform several times better than this; this should be possible with further improvements, the researchers estimate."
You mean like the Slot 1 and Slot A (Intel & AMD respectively) CPU packages? IT's been done, it's just much more costly. Although I do remember in the heyday of the Celeron 300A, quite a few overclockers would make "Celery Sandwiches" with a Celeron in the middle and a heastsink and fan mounted on both sides. It sire made for an interesting look.
If I could only live my life with my threshold at 4...
What would you do with a "Peltier-on-a-chip"?
1. Assemble a Beowulf cluster of them.
2. Leech more mp3s from Napster.
3. Bundle censorware with each one (in compliance with Texas law).
4. Leech more mp3s from Napster, but call it "hacktivism".
5. Wintel r00l3z!
6. Cowboy Neal.
Any sufficiently well-organized community is indistinguishable from Government.
Most chips manufactured are created to work at a certain maximum tolerance. If a chip won't test reliably at 1.5 Ghz, it's thrown in a pile of identical chips labeled and sold as 1.4 Ghz.
If these advances allow for reliable on-chip cooling, then you can bet that both AMD and Intel will keep these chips clocked as absolutely high as they'll go, thus eliminating the practice of user overclocking altogether.
The next Slashdot story will be ready soon, but subscribers can beat the rush and slashdot the links early!
Forest...trees... The amount of power being used completely unnecessarily by residential users is significant : Maybe it doesn't make a big difference when you consider one single home and you can laugh at initiatives for conservation, but when you consider an entire state it can be substantial. In 1999 there were 11,490,000 households in California. If every one of them replaced a single 100W lightbulb with a 15W compact flourescent, that is 976,650,000W of savings. Do you realize that most nuclear power plants only produce around 100,000,000W? So there you've potentially eliminated the need for >9 nuclear power plants by REPLACING A LIGHTBULB and you're talking about how individual users don't make a difference? Give me a break...
And you say that an extra 200W per PC, or >2,000MW over the state, isn't a big deal. Let me guess : You don't vote because your vote doesn't count, right?
If every one of them replaced a single 100W lightbulb with a 15W compact flourescent, that is 976,650,000W of savings.
You assume that the lightbulb is on all the time, which is incorrect. I hardly have any lights on ever at my place, and most people I know at most use bulbs for a few hours per day - and they're not going to spend a hundred bucks swapping bulbs - those 15W ones are expensive as hell. Telling people to buy them at an added cost to them - less beer, for example - without raising the price accordingly flies in the face of the economics upon which your country was built.
Not to say conversion isn't a good thing, but the reason people waste power IS BECAUSE THE PRICE IS ARTIFICIALLY LOW. If you want people to use less power, for god's sake, just RAISE THE PRICE. That's capitialism, aren't you guys the united states of america? The supply falls, the price rises, more people will want to build power stations - but oh, wait, you've gone and fucked yourselves with environmental legislation that flies in the face of reality. You SHOULD have several more nuclear power plants, or hydro, or coal, or whatever, if you want to sustain the current price to consumers.
You can buy all the power you want from us in Canada - it just isn't going to be cheap. Raise the price, and watch all those 15W bulbs fly off the shelves. Lower the enviromental regulations, and build some power plants. Just wait until people start using their A/C in summer - you have lots of people, well, you get lots of pollution to match.
..don't panic
You assume that the lightbulb is on all the time, which is incorrect. I hardly have any lights on ever at my place, and most people I know at most use bulbs for a few hours per day - and they're not going to spend a hundred bucks swapping bulbs - those 15W ones are expensive as hell.
Ah but therein lies the crunch : Most of the power system in place is to deal with momentary peaks because people do tend to all do the same things at the same time: Everyone cranks their ovens on at the same time, and generally at the same time AC powers up (and of course in warmer places like California every W of lighting turns into a W of heat that the AC has to remove from the air). At common times a good portion of the population has their hairdriers on in the morning, and their water heaters come on because they had a shower. Every W that is piled on top of that load is a W that has to be accomodated in the power grid.
Having said that a couple of quick points
I read a different article on this (don't have link, sorry) that suggested that these would be used to move heat between different regions of a chip. For example, cool a very hot ALU by dumping the heat elsewhere. Before you say that's stupid, in reality a more even heat distribution would decrease the maximum temperature of any given point on the chip, allowing it to run hotter on average (and hence faster). Also, it makes it easier to cool, as a greater region is in contact with the heat sink.
just right for when you need to carry that microbrew beer around but don't want the hassle of a regular sized cooler.
--
--
"It is now safe to switch off your computer."
It doesn't matter if it only moves the heat a small distance.. it still moves it. If you touch a surface with your finger, it doesn't matter if the material 1 mm down is at a thousand degrees.... if the surface you are touching is at 0 degrees, it'll feel cold. Period.
Peltier devices move heat away from one side to the other side.. and they also generate heat (which ends up on the hot side of course). That's why there is always a point where it's generating more heat than it can move, and becomes inefficient.
The point is, it moves heat away from the chip surface faster and more reliably.. that heat still has to be bled off with a heat sink/fan/whatever.