The human brain uses 200 to 400 kilocalories per day, which equates to between 10 and 25 watts of power. For comparison's sake, that's about 10 to 25 percent of the power that it takes to run a 100-watt light bulb. A regular computer performing the same amount of calculations the same way the brain does would take more than 40 million times the energy that the brain uses.
More facts about the human brain:
- The brain can perform about 1016 synapse operations per second using only that 10 to 25 watts of power. The most efficient supercomputer out of the top 500 computer systems in the world as of 2011 could do only about 2,000 millions of floating point operations per second (MFLOPs) per watt. That means that the supercomputer would need to use 500,000 times the energy that the brain uses to perform in the same way.
- The 10-watt number might be the source of the myth that humans use only 10 percent of their brains. Actually, almost all parts of the human brain operate to some degree at all times.
- The brain has more than 1,000 trillion synaptic connections. The most advanced computers as of 2011 had only about a million silicon neurons.
Follow wiseGEEK
More Info: http://blog.discovermagazine.com
Discussion Comments
The largest chips in 2011 would have had billions of transistors, not a million. We're up to what, 8 billion or so per chip now (max). Heck, the current gen CPU/GPU in the iPhone has like 2 billion.
In terms of numbers of transistors versus numbers of neurons, computer chips are within spitting distance, though obviously they don't work the same way, and there's insane complexity in brains from bajillions of connections all over the place. (Although also computer chips run way, way faster than brains.)
Is it really correct to compare MFLOPs with brain's synapse operations?
Post your comments