Handheld Supercomputers in 10-15 Years?

An anonymous reader writes "Supercomputers small enough to fit into the palm of your hand are only 10 or 15 years away, according to Professor Michael Zaiser, a researcher at the University of Edinburgh School of Engineering and Electronics. Zaiser has been researching how tiny nanowires — 1000 times thinner than a human hair — behave when manipulated. Apparently such minuscule wires behave differently under pressure, so it has up until now been impossible to arrange them in tiny microprocessors in a production environment. Zaiser says he's figured out how to make them behave uniformly. These "tamed" nanowires could go inside microprocessors that could, in turn, go inside PCs, laptops, mobile phones or even supercomputers. And the smaller the wires, the smaller the chip can be. "If things continue to go the way they have been in the past few decades, then it's 10 years... The human brain is very good at working on microprocessor problems, so I think we are close — 10 years, maybe 15," Zaiser said."

Re:Why supercomputers?

[Anonymous Coward]

More to the point, Supercomputers are not called "Supercomputers" because they are simply faster than other machines. Supercomputers are large-scale vector machines designed for number-crunching capacity. They're great at scientific modeling and simulation, but aren't exactly something all that useful to the average person. (Unless you somehow think that the Cell in the PS3 was the smartest idea ever.)

Also, like most things in computing, "Supercomputer" is a moving target. Today's supercomputers tend to be large clusters of inexpensive machines running OSes like Linux, Mac OS X, or Solaris. (Windows supercomputing clusters probably exist as well, but I doubt that many organizations are willing to pay the software licensing fees.) So unless we can have a 500 processor distributed computing cluster in a Palmtop in 10 to 15 years, I seriously doubt we'll have "a handheld supercomputer". And if you want to go by the supercomputers of yesteryear, technically we already have that power in our handhelds. e.g. An iPhone's SIMD-equipped 625 MHz ARM [engadget.com] processor could probably hold its own in vector calcs against some of the earlier supercomputer installations.

Sooo.... I call sensationalist headlines. Do I win a prize?

Re:10-15 years?

[maeka]

A quick google search appears to show modern PDAs [ocforums.com] competing nicely with a mid-80's Cray. [wikipedia.org]

Re:Why supercomputers?

[stonecypher]

Isn't a super-computer a relative term?

No. Supercomputer is a specific term with a specific speed attached, and has been since the word was coined in the 1970s. The word is backed by law, because of export restrictions. A supercomputer can perform a trillion floating point operations per second (one teraflop,) which was a goal that was difficult at government scale in the 1970s, and is now not all that big a deal. You remember when that North Carolina State professor made a supercomputer out of eight PS3s? He couldn't have done that if "supercomputer" didn't have a rock solid meaning. It's one of those things that only old people seem to know anymore, like that a byte is not necessarily an octet, that bits per second and baud aren't the same thing, or that bandwidth and storage - and indeed everything but ram - is measured base 2 instead of base 10.

Surely the measure of what is a super-computer and what isn't must be based upon what the fastest machines are in the world at that time.

Nope. That would mean that something that's a supercomputer in year 1 might not be in year 2, which would reduce supercomputer to a marketing term. Believe it or not, computers are measurable. Some terms have actual meanings. This is one of those.

I mean, I don't know the exact figure but I would that my Dual Core Intel box at home is probably a good deal faster than a super-computer from the 80s.

Nope. Home PCs will likely cross the teraflop threshhold around 2012. All supercomputers from every era have the same processing threshhold. A current quad-CPU dual core box would be enough.

Perhaps what he means is that what we currently do with supercomputers today will be able to be done with low cost computing.

Nope. He means a teraflop.

However, the improvements in computing speed will also apply to super-computers.

Generally speaking, once a computer has been manufactured, technology improvements do not alter it. There are exceptions, especially in computers which are limited by temperature, but not many. A supercomputer from the 1970s is still a supercomputer today. Please stop attempting to argue with an article on grounds of metaphor structured around words of which you don't know the meaning.

In short, I can't really understand the super-computer slant of the article. Why not just talk about general-purpose computing instead?

The interviewee is old enough to know that supercomputer means something fixed, and that therefore there is a threshhold to be crossed in the fashion of getting a supercomputer into a specific form factor. The interviewer doesn't understand geeks well enough to know that they won't know what a supercomputer is, and fails to explain, probably expecting people to go read the deeply wrong article on Wikipedia. That help?

Re:We already have handheld supercomputers

[stonecypher]

No, really. An iPhone is much more powerful than the Cray-1, and probably significantly more powerful than a Cray X-MP.

I'm not sure why you believe this. I'll assume you mean the Cray 1A, since the Cray 1 is just a specification; it's a bit like talking about the 386, since the 386 ran at about a dozen different clock speeds. The Cray 1A was the first actual implementation of the Cray 1 spec, and was initially installed at Los Alamos. SCD's Cray 1 was installed about six months later, and ran at 160 megaflops [ucar.edu]. (The Los Alamos Lab one almost certainly ran at the same speed.)

Gen3 IPods use a pp5002d as a CPU. I'm not able to track down its actual performance, but in several places I see a Rio engineer saying that Vorbis is just at the edge of its performance capabilities. Tremor, a Vorbis implementation, runs just fine on the Nintendo DS - it eats about 40% of your CPU time if you're running it on the Arm9/75. Sony cites their UX50 - an Arm9/125 - as performing 2.51 megaflops. yCPUbench quotes 2.44, suggesting Sony has a slightly better tuned test set for that architecture, which isn't surprising. If tremor needs 40% of a 75mHz arm9, or ~30mHz, then it needs 24% of the UX50, or about 0.6 megaflops. This suggests that the iPod has a bit over 0.6 megaflops to bring to bear. Considering that all it does is play music, it should be no surprise that it has less CPU than a Nintendo DS, which needs to do many things in parallel with playing music.

What is surprising, however, is that you believe that it's faster than a Cray 1A. 160 to 0.6 - the cray from the 70s is approx. 265 times as fast.

Now, the Cray X-MP ran at a huge range of speeds, because it was a modular design; there are deployments that were several thousand times as fast as the base install. But, if you check that same SCD history PDF as above, their X-MP/48 ran at 0.91 gigaflops, or about one point five million times as fast as your iPod. Still, that was kind of a lower end X-MP, because SCD was saving up for a TMC CM-2. The X-MP is about half as powerful as an XBox running untuned linux [uh.edu]. The iPod is nowhere near that ballpark; it's only about twice as fast as a Gameboy Advance.

The iPhone certainly has much more RAM and storage than they typical early Crays

Storage, yes. RAM, not even close - your iPod has 96k, and in 1970, the Cray 1A at SCD hat 8 meg. Please stop making things up.

Maybe you should try doing the math before getting on the soapbox. When someone fills in the numbers you thought you could pull out of the air, and you're wrong by an average of six orders of magnitude, you start looking pretty bad.

Re:Why supercomputers?

[stonecypher]

the first TeraFlop computer didn't appear until 1997. Does that mean that there were no supercomputers before this date?

You are correct. On checking, the number is gigaflop, not teraflop. My mistake: I misremembered on which line the term hinged. The first supercomputer appeared in 1961 - the IBM Stretch. But, in response to the intent of the question, yes, there is a specific date on which we crossed the supercomputing barrier.

Re:Yes, it will run linux

[pclminion]

Throwing things on the floor go much faster than 9.8 m/s^2.

No it doesn't, at least once the object leaves your hand. Then it's back under the influence of good old gravity, at 9.8 m/s^2, regardless of how fast you may have thrown it.