Supercomputers 'will fit in a sugar cube', IBM says (BBC)

[eppur_se_muova]

By Jason Palmer
Science and technology reporter, BBC News, Zurich

A pioneering research effort could shrink the world's most powerful supercomputer processors to the size of a sugar cube, IBM scientists say.

The approach will see many computer processors stacked on top of one another, cooling them with water flowing between each one.

The aim is to reduce computers' energy use, rather than just to shrink them.

Some 2% of the world's total energy is consumed by building and running computer equipment.

Speaking at IBM's Zurich labs, Dr Bruno Michel said future computer costs would hinge on green credentials rather than speed.
***
more: http://www.bbc.co.uk/news/technology-11734909

[ishaneferguson]

1. IBM Nobel Prizes

This looks like it draws upon the work of IBM (Zurich) Nobel Laureates Gerd Binnig and Heinrich Rohrer.

[veganlush]

2. but won't the water melt

the sugar?

[RC]

4. That is why they have to be fast, to get the program run before the cube dissolves.

[Roma]

3. I'm still amazed that my Droid has

a 1GHz processor, 512MB RAM and 24GB storage. In a "phone".
I ran a 5 store retail operation on a Compaq 8086, twin floppy, suitcase portable for crying out load.
How much computing power did NASA have when they landed on the moon?

[XemaSab]

10. I found an article about this (one of many, I'm sure)

Take Intel's venerable 8086 for example -- you might know it better as "x86". Released in 1979, just a decade after Apollo 11's trip to the Moon, the 8086's cousin, the 8088, formed the basis for the IBM PC we all know and love. When the IBM PC "XT" was released in 1981, the lowest end configuration had 8 times more memory than Apollo's Guidance Computer -- 16k, vs the Apollo's 2k. The read-only storage of the AGC was 32k,

The IBM PC XT also ran at a dizzying clock speed of 4.077MHz. That's 0.004077 GHz. The Apollo's Guidance Computer was a snail-like 1.024 MHz in comparison, and it's external signaling was half that -- actually measured in Hz (1/1000th of 1 MHz, much as 1 MHz is 1/1000 of 1 GHz).

Internally, the 8086 had 8 16-bit registers available to work with -- for those not familiar with the internals of a processor, a register is much like the numbers you'd keep in your head while doing math, and "memory" is more like scratch paper where you write things down for later use. The 8086 could keep track of 8 of those, the Apollo Guidance Computer held just 4. (The AGC also had a host of other non-general purpose registers, ranging from 1 to 16 bits in width, but it's difficult to compare those with the architecture of the 8086)

Regardless, we've established that we're dealing with some pretty archaic hardware. In many ways, the AGC was half the IBM PC XT you could buy off the shelf just a decade or so later.

http://www.downloadsquad.com/2009/07/20/how-powerful-was-the-apollo-11-computer/


------------------------------------------------

It's my understanding that your camera, your watch, your non-smart phone, or your calculator is more powerful than the AGC, and your smart phone is orders of magnitude more powerful.

[OnlinePoker]

17. Remember the movie Apollo 13

There was one scene where all the rocket scientists used slide rules to make their calculations. Maybe not as accurate, but they were certainly faster than the computers of the day that often had to have programs loaded into them before you could do anything.

[spin]

5. I wonder if in the future you will have to put Preston anti-freeze ...

in your computer in case the temperature drops below freezing.

[aquart]

6. Will they expect us to eat them?

[struggle4progress]

7. "water cooling channels not much more than a hair's breadth in thickness"

Dunno. Manufacturing defects and irregularities won't give you uniform channels, and the pressures required to move water through those tiny pipes might be enormous: I should expect some steam damage to chips unless the channel networks were intelligently designed with some redundancy. And I doubt if water would be the ideal coolant: if you could really shrink the transistor network to sugarcube size, it might be worthwhile to use liquid helium or something like that as a coolant -- you'd reduce the temperature substantially and so could run the clock much faster

[XemaSab]

8. I would expect that the water would move via capillary action

:shrug:

[struggle4progress]

9. capillary forces may help wet the channels initially but won't maintain flow

under most circumstances

[XemaSab]

11. If there's evaporation or some other pull at the end it'll work

Like in trees. :D

[struggle4progress]

14. I suppose you could design a chip that spewed steam from one side

I might expect oxidation of the silicon by steam to be a problem, so I'd expect that a design like that would work better with a coolant other than water and a refrigeration system to capture the gas

[Confusious]

12. Why would you use liquid helium?

Silicon isn't a super conductor, and chilling it below a certain temperature won't make it able to run faster.

After a certain GHZ, around 3, the problem becomes the silicon itself.

It starts leaking between the junctions, requiring multiples of power for even a small speed improvement.

Why do you think they've gone to multi-core? It ain't 'cause it's neat. It's the only direction they can go until they get a new material, like diamond.

[struggle4progress]

13. It seems to me the 3ghz limit is thermal; hobbyists get beyond 4ghz regularly

with liquid cooling

A 3D processor chip will have more serious cooling problems, which explains the built-in cooling-channel design; running in a lower temp environment might give more leeway with the thermal problems

I expect good quality 3D chips will be very very expensive for the first few generations. If one has the resources to purchase machines built on such chips, I would expect good cooling to be an add-on expense within budget. Hobbyists seem to get 5ghz easily from off-the-shelf chips with crude liquid nitrogen cooling: I can't imagine one could get comparable cooling with water, even if the chip has built-in cooling channels

I'd expect some advantage for microchannel flow from the low viscosity of liquid nitrogen or helium. Proposing liquid helium was just shooting from the hip: maybe you are right that liquid helium isn't necessary, and maybe there would be some real disadvantages to it; I just can't see doing this with water

[Confusious]

19. The main disadvantage of liquid helium or nitrogen is that it doesn't last long

I used to work for a company that made infrared cameras that were cooled by liquid nitrogen and helium. It's not for the regular consumer.

Ice plugs can form, the gas can't escape, and BOOM! shrapnel everywhere.

I also got the silicon part wrong. The speed issue is a limitation of current manufacturing techniques, such as the mixture of the silicon and the size of the transistor.

Mixed with other materials, such as germanium, it can go faster.

http://www-03.ibm.com/press/us/en/pressrelease/19843.wss

[struggle4progress]

20. These proposed 3D sugar-cube sized processors won't be for the regular consumer either:

they're going to be incredibly expensive for the first few generations, and they'll be for organizations who need supercomputers, since that's what they will replace. At 10-15 years for the first generation, it'll be 20-25 years before any version could possibly hit the high-end enterprise market

[Confusious]

21. My point is, liquid helium and nitrogen require more precautions

Then just running water through them. There's also the cost issue. You don't get the liquid nitrogen back. Once it turns to gas, it's gone.

[bananas]

15. This thread should be moved to the sewing forum

Because it clearly has nothing to do with Energy/Environment

[joshcryer]

16. Reducing chip energy use has nothing to do with E&E?

I find it more relevant to E&E than a significant number of copy paste jobs here.

[Nihil]

22. He's just bitching ...

... because his earlier trolling cross-post from GD ended up being moved back there!

:rofl:

[hunter]

18. How many virtual computers will it run?

That's where much of the energy savings come from. Instead of having multiple racks of computers you just have one.

[Javaman]

23. One lump or two?

first you get the sugar, then you get the money, then you get the women.
-Homer Simpson

We were talking about computers right?