4,580 Posts served

11,093 Conversations started

The Speed of Molecular Computations

By Clay Breshears (Intel) (196 posts) on December 22, 2009 at 3:20 pm

I've been interested in biological systems for computation ever since the first DNA computation paper published by Adleman, "Molecular Computation Of Solutions To Combinatorial Problems," in Science (266, pp. 1021–1024; 1994) .  I've even blogged about it here and here.  Now researchers at the University of Kent have published a paper that calculates the limits for biological computations.  (See the press release at the university site.)

I guess I shouldn't have been too surprised that the press release didn't have too many details.  I thought there might have been some metric put forward like petaflops or megaflops or flops per fortnight.  Maybe those details are in the research paper.  Or it might depend on what biological entity is doing the computation and how many we can harness to execute on a given problem.

Not too surprisingly, though, it seems that the researches have shown that the rate of computation (and accuracy of results) depends on the amount of energy consumed by the biological elements.  Give them more energy, typically in the form of food, and the better the "devices" will do.  I thought it was going to be hard to keep multiple cores "fed" from memory.  In the near future, in addition to making sure the storage system is accessible fast and wide enough, with biological computers we will literally have to keep them fed.  I wonder if we'll see the day when processor energy consumption is measured in calories instead of watts.

Categories: Academic, Parallel Programming

For more complete information about compiler optimizations, see our Optimization Notice.

Comments (2)

December 22, 2009 3:50 PM PST jimdempseyatthecove Total Points: 77,429 Black Belt

Watts and Calories are different dimensional units. A watt is a unit of power (Energy per second). A calorie is a unit of energy. Therefore it would be a measure of calories per second. A watt is 1 Joule per second A calorie is 4.184 Joules Therefore 1 calorie/s = 4.184 watts The biological computer would likely be rated in calories per some elemental operation, with different elemental operations consuming a different amount of calories. The rate at which you can pump calories into the system will effect the rate of operations. There will likely be a waste heat issue or waste chemical waste issue and disposal issues restricting the performance of the system. I doubt if you will be performing FLOPS (floating point operations). More likely you will be performing agragations (consensis-like processes). While you could do FLOPS the current generation silicon devices do a wonderful job at FLOPS. The biological computer would likely incorporate a silicon part for doing mundaine tasks such as a flop.

December 24, 2009 8:43 AM PST Clay Breshears (Intel) Total Points: 30,566 Black Belt

Thanks for clearing that up, Jim. It's been too many years since I've had a look at Physics. As I was writing the artricle, I did wonder to myself what the unit of computation speed would be. Not all processing is done with floating point values. Heck, Adleman's original paper solved a non-numeric Hamiltonian Path problem. FLOPS might not be the best. IPS (instructions per second) seems generic enough, but how does one break down a biological process into instructions? Perhaps the type of computation will be different. Something like LIPS (logical inferences per second) that was popular at the height of AI, Fifth Generation, and logic programming years. Maybe we will use some kind of information/entropic metric ala Shannon: Petabits aquired per calorie per second (pBACS).

Trackbacks (1)

• Intel Software Network Blogs » The Speed of Molecular Computations ZB BU online
  December 22, 2009 9:02 PM PST

Leave a comment