Slashdot Mirror
How To Build a Supercomputer In 24 Hours
An anonymous reader writes with a link to this "time lapse video of students and postdocs at the University of Zurich constructing the zBox4 supercomputer. The machine has a theoretical compute capacity of ~1% of the human brain and will be used for simulating the formation of stars, planets and galaxies." That rack has "3,072 2.2GHz Intel Xeon cores and over 12TB of RAM." Also notable: for once, several of the YouTube comments are worth reading for more details on the construction and specs.
Price, from comments:
Just under 750,000 Swiss Francs, or about $800,000
Slashdot, fix the reply notifications... You won't get away with it...
only seems to bother EE majors and everyone else seems to be immune to it.
Racist comments don't routinely stay modded up on slashdot as they do on many parts on YouTube. Use of dubious debating techniques such as the strawman usually gets noticed here. Unpopular viewpoints are often modded up to +5 Interesting if they are sufficiently well argued.
Slashdot ain't what it used to be, but it still has standards.
So these things happen on different (connected) machines if you will. Every neuron is a processor in itself anyway.
I think that's part of why computing power greatly surpassed humans long ago, and will not reach human levels for many years. The brain isn't digital. It holds an "infinite" number of analogue states, simultaneously. With massive errors and gaps filled in with guesses made from other parts, without even an minor error check that indicates that he information being determined to be "true" is 100% interpolation with 0% fact or actual memory. The very idea of an error check that was wrong more than right and kept no indication of where the result actually came from is so incredible that nobody would ever create a computer capable of operating that way. It won't be until we have computers many millions times more powerful where we can remake a "perfect" brain, until then, we'll never be able to match the capabilities of the human brain.
Learn to love Alaska
Doing equivalent computation to a human brain, and simulating a human brain, are two very different problems.
You have to solve a non-linear coupled differential equation at 1MHz to simulate a 5kHz sawtooth wave generator in Spice. It's about 30 MFLOPS. But, functionally, all you're doing is generating a 5kHz sawtooth wave, which is 15 kFLOPS of work. This is a 2000:1 efficiency difference between simulating an analog system and running a direct digital equivalent implementation.
So divide that 100 PFLOPS by the fundamental inefficiencies of simulating the analog domain in the digital domain, and you get a more reasonable figure for when a computer can functionally compete with the human brain.
I will ask the inevitable questions, as a system builder.
How many parts were DOA?
How many failed inside of the first month?
...
Boats and submarines don't "simulate" fish, but they still swim.
Airplanes don't "simulate" birds, but they still fly.
Artifical intelligence may not need to "simulate" the brain to reach human level.
Just sayin'.