Radically Faster Simulations

Computers tend to get about ten times faster in about five years. In ten years, 100 times faster; in fifteen years, 1000 times faster. This is called Moore's Law, and it's been quite steady and predictable for several decades.

Recently, Nanorex simulated a 25,374 atom machine called a worm gear. The simulation took about 370 hours of computer time. We might expect that five years from now, this would take about 37 hours--still painfully slow.

But a new thing is happening: computer time is being sold on the Internet. Amazon.com is selling computer time by the hour for 10 cents apiece. I doubt that the price per hour will rise as time goes by; meanwhile, the hour will get more and more powerful. Five years from now, the full simulation will need 37 hours and cost only $3.70.

NanoHive already is designed for @home (distributed computing) type use, so I'd expect it to be portable to the Amazon "cloud" which is designed to provide standard computer resources. In that case, a user running large simulations would be able to buy 100 CPUs in parallel for the 22 minutes the simulation would take on a 100-CPU cluster. If the user bought 1000 CPUs in parallel--I'm guessing that Amazon's computer network would support this--they still would pay only $0.10 per CPU-hour, and the simulation would take a bit over two minutes.

There are two ways to look at this. One is that private users will now have inexpensive access to a supercomputer. The other way to look at it is that for just $10 per clock hour--affordable even to hobbyists--researchers can get a ten-year jump on Moore's Law.

The rapid advance of Moore's Law is one of the reasons why we expect molecular manufacturing to get rapidly easier to develop. Now, as soon as a bit of software infrastructure is put in place, the widespread availability of supercomputing will advance by ten years or more.

Chris Phoenix

Tags: nanotechnology nanotech nano science technology ethics weblog blog

Comments

[NanoEnthusiast]

Chris, what do you think are the chances that quantum computers will arrive soon enough to help out? From what I've seen progress is happening faster than I used to think.

Also, how do you address the skepticism that many have with the results of simulations? Ideally, you could always use quantum chemistry techniques, but that is not practical for large systems. I know that NanoHive does allow you to use DFT and it uses distributive computing. Theoretically, if enough people donate their CPU cycles you could get a supercomputer to do DFT calculations for free. Of course, a practical quantum computer would be best.

[Chris Phoenix, CRN]

There's a company that claims to have a quantum computer for atom simulation nearly demo-able. Sorry, I can't find the link at the moment. I don't know enough about it to know how much help it'll be.

As to skepticism... I'm inclined at this point to ignore the skeptics and focus on the people who know how to use simulations within their limitations to advance our understanding of the nanoscale. As time goes on, the skeptical point of view will become irrelevant.

Chris