IBM Water-Cools 3D Multi-Core Chip Stacks

An anonymous reader writes "Water cooling will enable multi-core processors to be stacked into 3D cubes, according to IBM's Zurich Research Laboratory which is demonstrating three-dimensional chip stacks. By stacking memory chips between processor cores IBM plans to multiply interconnections by 100 times while reducing their feature size tenfold. To cool the stack at a rate of 180 watts per layer, water flows down 50-micron channels between the stacked chips. Earlier this year, the same group described a copper-plate water cooling method for IBM's Hydro-Cluster supercomputer. The Zurich team predicts high-end IBM multicore computers will migrate from the copper-plate water-cooling-method to the 3-D chip-stack in five to 10 years." Reader Lilith's Heart-shape adds a link to the BBC's article on these internally-cooled chips.

3D cubes are nice, I guess

[krog (25663)]

But they're really gonna rev up performance once they move to 4-cornered time cubes [timecube.com].

But the question is....

[CowboyNealOption (1262194)]

can it run Vista??

my favorite

[by Anonymous Coward]

mmm cool ranch centrinos

When will water cooling be feasible for ME?

[BadAnalogyGuy (945258)]

Water cooling is great for the bleeding edge enthusiast, but it's hardly an option for the workaday computer users. Laptops certainly could stand to use some better heat dissipation, and if water cooling through 50nm tubes is possible here, how long until it is both cost effective and size-effective for people who aren't interested in hardware for its own sake to see this type of thing offered to us, the average computer user?

And is stacking the chips better than laying them flat and in a strip (like Pentium M)?

Re:

[wattrlz (1162603)]

... And is stacking the chips better than laying them flat and in a strip (like Pentium M)?

Sure. The interconnects could be shorter and thus impose much less lag. Core one wouldn't need to go through core two to talk to core three, etc.

Re:

[nategoose (1004564)]

Laying them out flat is better for cooling because it has more surface area, but the cube can be faster since the maximum distance from any 2 points within it is reduced from what it would be if the same chip area were laid out flat. This is why it NEEDS water cooling.

Re:

[colmore (56499)]

How is this insightful? Water cooling may never be feasable for you. Unless you count the datacenters that you use for networked applications (like... readings slashdot) or the large numerical processors that enable the science and engineering behind the crap you use every day.

Water cooling wasn't invented by overclockers. Cray used it in many of their production systems in the 70s and 80s and its use with CPUs goes further back than that.

The stack of chips is to increase the connectivity between the mul

Electrolysis

[mrbluze (1034940)]

To cool the stack at a rate of 180 watts per layer, water flows down 50-micron channels between the stacked chips.

I wonder what reactivity of water with the surrounding surfaces will do to the life of the chip. AFAIK pretty much anything that uses water has an inherent limitation to its life, owing to the presence of superoxide radicals and free hydrogen ions.

Re:

[iamdrscience (541136)]

superoxide radicals

Sounds like a sweet name for a band.

Re:

[wattrlz (1162603)]

If the inside of the system is all made of one material couldn't you just put in deionized water and hope for the best? Copper, silver, and silicon are pretty water-resistant when there isn't anything in there with them to catalyze the reaction.

Re:

[chthon (580889)]

I once saw a demonstration (mid 80's, I think), on an exhibition, of a water purifying system. The demo consisted of a tank of water with in it a playing television. The backside was removed to demonstrate that all the components where effectively submerged under water.

Re:

[rahunzi (968682)]

What about FREEEZING???? you limit chip to an environment where water is liquid - also size of molecules is finite and chips sizes decrease, water will not... I guess a PLATE or MANIFOLD would work and simplify connectivity... this is also SCALABLE into some supercooling/conductivity

Re:

[drinkypoo (153816)]

Perhaps they will utilize a sacrificial anode (another great name for a band...)

This will never work

[iamdrscience (541136)]

How can IBM be this stupid? You can't cool a stack of chips with water, they'll just get soggy. I know it's hard to be patient, but if your chips are too hot to eat, you're better off just waiting for them to cool down.

Re:

[ptbarnett (159784)]

Insightful?

Moderators are on crack this morning, again.

Re:

[iamdrscience (541136)]

It makes sense, you see, although modding a funny post as "funny" may be more accurate, modding a funny post as "insightful" instead is definitely more funny, so it's a much more appropriate moderation. You see, the moderator is making a joke about the joke. I believe this is called "metamoderation" -- if you have an account you may have noticed Slashdot encouraging you to metamoderate from time to time.

180 Watts per layer

[javilon (99157)]

Sounds like too much, with typical numbers around 60 watts per processor this days.

Re:

[kcbanner (929309)]

These are not you average Centrino procs, these is server hardware running at close to 100% load probably. It gonna get hot :P

3D CPU structure

[Lord Lode (1290856)]

I always liked the idea of a 3D CPU with all the cores and memory interweaved through each other in a way to have the optimal short path for its purposes. A LOT of memory could be there right next to the CPU. It would be fast even without clocking it very high, so not even have to consume that much watts per layer. It's a crazy amount of watts per layer mentioned in the article btw...

Yes, that was God's idea...

[mosel-saar-ruwer (732341)]

...when He created us:

human-brain-vis304784-ga.jpg [nationalgeographic.com]

Great minds think alike.

Upgrades to cotrol systems needed

[kcbanner (929309)]

Right now, if the pump is off, or if the flow isn't flowing, the processor is none the wiser and happily starts up. I've seen my Core2Duo hit 100C when my pump died, my only warning was when the comp just shut off when it hit the temp cap. There needs to be some sort of control system that is actually linked in to the processor, so that it won't start if the flow of water through the block (or now the CPU itself) is below a certain rate. Most people who do use watercooling, however, know what they are doing and this usually isn't an issue, it would just be nice to know the server rack won't melt itself when someone blows the pump breaker.

Re:

[conureman (748753)]

With the rapid self destruction time inherent in this system, I would like to think the engineers have addressed this issue.

Re:

[ColdWetDog (752185)]

IMHO they should use a fluorinated carbon of some kind - something that won't react with the processor nor participate in significant electrolysis.

You mean, like this [jupiterresearch.com]? More likely they'd use something like commercial antifreeze solutions seen in vehicle radiators. But you still need to maintain the fluid (draining, replacing and what not.)

Next up, oil changes for your PC (cue stupid car analogies).

Re:

[ColdWetDog (752185)]

Err, stupid copy buffer. How about like this [wikipedia.org].

Wanders off to get more coffee muttering about the lack of an edit function in slashcode.

Everything that's old is new again

[holviala (124278)]

We just finished removing all the water cooling tubing which the old mainframes used.... But hey, don't tell anyone that watercooling big computers isn't a new idea :-).

Risky

[Tribbin (565963)]

If the water gets 100C. it will boil and leave the processor in an isolating bubble.

Imagine the mistakes of the future

[kiehlster (844523)]

I can see it now, "IBM struck with class-action lawsuit after several incidents of computers being left out in the cold of winter cause the processors to explode due to the natural properties of water expanding into ice. Other incidents with water contamination in liquid nitrogen-cooled 3-D processors have resulted in a similar lawsuit."

Peltier-Seebeck

[conureman (748753)]

I am not an engineer, but I've been kicking this concept around in my head for a while, short paths FTW. I always thought of thermoelectric cooling solutions, water-through-the-chip... wow

P.S.

[conureman (748753)]

BTW, I was thinking of 16 exabytes of RAM for each processor core, on the same chip, so the Bus only feeds the peripherals.
  Like I said, IANAE.

Again?

[K. S. Kyosuke (729550)]

Funny, I thought that this thing has already existed for a while at IBM? They called it Multi-Chip Module-Vertical (MCM-V) at that time. But maybe just the cooling had to be redone for those power-hungry modern cores.

IBM have done this before

[mad zambian (816201)]

IBM and water cooling of chips is not really new. I remember reading of some research they did back in the 80's when they etched micro channels on the back of processor chips, and forced water through them. IIRC, they reckoned they could eventually dissipate almost 1KW per square centimeter.
You want to drive bipolar chips fast, you apply more power. And end up with a piece of silicon dissipating way more heat per unit area than an electric fire. Mind you, so do Athlons.

Multicore resource portal

[by Anonymous Coward]

If you like to read more information on multicore processors, go to http://www.multicoreinfo.com/ [multicoreinfo.com] .

plumbing always leaks

[johnrpenner (40054)]

plumbing always leaks eventually - what a mess - my system melted down, and there's coolant all over the cpu -- blech. :-P

CMOS = Power Efficient???

[Cassini2 (956052)]

Does anyone remember the good old days when Metal Gate CMOS represented a power efficient process? We have went from CMOS devices consuming milliwatts and microwatts to processors with a 125W+ Total Power Dissipation. This announcement is talking about 180 Watts per layer!

How long will it be before my computer heats my house while I browse the internet? When does the first combined datacenter and heating cogeneration system get installed?

Re:

[Yetihehe (971185)]

How long will it be before my computer heats my house while I browse the internet? When does the first combined datacenter and heating cogeneration system get installed?

About two months ago. http://www.ecofriend.org/entry/ibm-manages-to-warm-pool-water-with-its-heat-emissions/ [ecofriend.org]

Re:

[necro81 (917438)]

CMOS is still a whole lot more power efficient than the TTL logic (i.e., bipolar junction transistors) that they replaced. Ideally, a CMOS transistor only requires power when switching states, whereas a BJT burns power continuously. Per transistor, they are a much better way to go.

The problem with high total power dissipation is the result of several interrelated trends, all of which can be related to Moore's Law. More transistors got crammed onto a single chip (a linear increase in power dissipation

Cray?

[CompMD (522020)]

Isn't this just miniaturizing what Cray used to do? They had processor/system boards in fluorinert, IBM has processor cores in water.

Alcohol cooling is a bad idea.

[Lilith's Heart-shape (1224784)]

You don't want to cool with alcohol. The boiling point of most alcohols is between 60 and 80 degrees Celsius, as opposed to water's boiling point of 100 degrees Celsius.

It's quite the opposite

[mangu (126918)]

You don't want to cool with alcohol. The boiling point of most alcohols is between 60 and 80 degrees Celsius, as opposed to water's boiling point of 100 degrees Celsius.

Actually boiling removes much more heat than conduction. This is the principle used in heat pipes [wikipedia.org], where you want a low boiling temperature, because that will be the temperature in the hot side.

Re:

[somersault (912633)]

Don't you have to conduct the heat to the liquid before it boils?

Re:

[somersault (912633)]

Because it's in 3D chips? One of the main issues with layering up circuits like that is heat dissipation.

Basic Physics of Thermoconduction

[mrbluze (1034940)]

I thought you would want something with a low boiling point so you can move the heat as far away from the source as possible?

Something with a high specific heat is what's needed, which is why water is good. You can have any boiling point you like, depending on what pressure you apply to the liquid (boiling point is when vapour pressure = atmospheric pressure). If you are going to compress and decompress something to drive heat away, then use a gas.

Re:

[snl2587 (1177409)]

Of course, all of this is assuming this is some extremely pure water. Otherwise fouling will occur, and in 50 micron tubes I'm fairly certain it will be hard to clean.

Re:

[Sandbags (964742)]

Much more importantly, 50 micron tubes will be extremely fragile. To move water in sufficient quantity to avoid such small samples reacking critical temps while inside the CPU is going to be REALLY difficut. with tubes only a few mollocules in diameter, extremely consistent pressure will need to be maintained. A bubble forming due to boiling would shatter the substructure of the CPU and destroy it. Too high pressure and pipes burst. Too low pressure, and water won't move fast enough to avoid either boi

Re:Alcohol cooling is a bad idea.

[jeffmeden (135043)]

Only if you are making a wet/dry system, such as one that relies on phase change. If that's the case, it's refrigerant you want, and not alcohol (there is no real benefit to the vaporization unless the pressure swing is high). If you are doing closed loop all liquid, you want something that stays a liquid since vapor can't carry as much energy as liquid can given the same space. See automotive liquid cooling and refrigeration phase-change cooling for plenty of high-efficiency examples, none of which use alcohol or any similar substance.

Re:

[Lumpy (12016)]

Propane works well for this. and YES there used to be guys that did their own R12 old car ac repair by adding propane to the AC system.

It works, but is one hell of a fire hazard.

Re:

[Lilith's Heart-shape (1224784)]

I figure that when it comes to vaporization, less is better. Vaporized coolant has to be contained, lest it condense in inconvenient places (like your mobo's capacitors).

Re:

[Gewalt (1200451)]

Actually, what you want is a fluid (fluid!=liquid) with as high a specific heat capacity as possible, while maintaining an acceptable viscosity in the expected thermal operating range. Mercury would actually be awesome in this application, but of course, its "dirty" so we couldn't have that.

Re:

[jank1887 (815982)]

helium doesn't cool things to low temps. you cool the helium down to low temps, and then pump the cooled helium against your heat source. it takes a lot of energy to cool the helium in the first place, and would take up a lot of space.

Somewhere...

[Petersko (564140)]

"When it's 1024 processors in a water-cooled solid block of silicon"

Somewhere there's a geek who has already accomplished this goal. He's using it to run Crysis at 4800x3600 with full detail, at 1600 frames per second, and no matter who he shows it off to, he still can't get laid.