How fast is the human brain?

If my CPU’s speed is measured at 933MHz, how fast am I?

 I imagine that the conscious part of the mind uses only a fraction of the overall processing power, the rest being used up by the various reflex systems, subconscious desires and repressing childhood memories. If this assumption is correct, would different parts of the brain run at different speeds?

The human brain is highly parrallel, having no internal clock to synchronise it, so any comparison with a PC’s processor is meaningless.

Right. But you can compare the speed of neuron firings to the speed at which transistors operate in a processor. Neurons become excited and unexcited much, much more slowly - on the order of fractions of a second (I seem to recall something like twenty times per second, but that could be wrong.) On the other hand, the brain is far more complex in its wiring, which more than makes up the difference. It’s not really possible to compare them, though. They operate in entirely different manners.

According to this hypothesis, if I read it correctly, consciousness runs at 30-70 hz.

No, that’s not really what they are saying at all - ‘oscillations’ (and evoked potentials, and such) do not allow one to say how ‘fast’ the brain is operating.

There’s a lot of data that indicates how quickly we can perceive successive tones, visual stimuli, touches, and so on - basically, how quickly input can be provided to the brain by our sensory transducers. There is also plenty of data on how long it takes us to becomes conscious of sensory events. However, such info has no direct relationship to the ‘speed’ or the brain or the ‘speed of consciousness’.

In terms of MHz, there is just no meaningful answer - there are things we can do that no computer is capable of, and things we are terrible at that the most puny computer can do with ease.

If you want to know more about exactly how the brain processes information (or at least, as much as we know about it) try reading Ian Glynn’s An Anatomy of Thought. As others have pointed out, the brain is a massively paralleled network of switches with no central clocking or control (though individual processes do seem to operate on a cyclic pattern, controlled by the hypothalamus, but that’s more of a hormone and neurotransmitter regulation function).

Neurons, though, are far slower than even the slowest diode. By that measure, the brain is enormously slower than any modern electrical computer. But you never have to stop, load Service Packs, and reboot your brain while you’re in the middle of a print job. :smiley:


I once heard a theory that our brains are actually so slow that it has a method of predicting the future, but by the time that prediction is processes it is in real time. If the prediction is wrong the mind has the ability to recreate the situation in your mind to make it ‘right’.

Basically we are living in a world a few split seconds behind, but our minds have extraploated a few split seconds ahead, so it is in real time. When errors happen our minds brainwash us to make us beleive the stuff that really happened.

Also we do have a very limited capacity for real time processing, but that is very focused.

sorry It was a very long time ago, in a galaxy far far away, and no cite.

I wouldn’t give up on finding an answer quite so easily. We can’t say exactly, but we can make some extremely rough comparisons.

From the May 2005 issue of Scientific American article on Neuromorphic Microchips

(any typos are my fault)

From this, it sounds like a brain may approach 1 million GHz. Now, I think that a computer is capable of doing much more processing during a single clock cycle than a brain does with a single neural connection. Thus, I would say that human brains could be slower than this by a factor of 10 to 100 at least. But, these numbers do give us some kind of starting point.

-Emerald Hawk, (who executed 3*10^18 instructions to bring you this post! :wink: )

I’d dissagee with that. A single neuron firing is the result of hundreds or thousands of different inputs from other neurons, each one applying an excitory or inhibitory threshhold which changes over time. A full mathematical model of a neuron is actually a pretty complex thing, hundreds of calculations needed to determine when a single pulse is fired. On the other hand, a computer during one clock cycle might be adding together two sixty-four bit numbers, or moving a register from one memory location to another, or a similar fairly minor calculation.

The Blue Brain project might shed some light on the answer to this subject:

They’re hoping to use one of the most powerful supercomputers on the planet to simulate the action of a single neocortical column (about 50,000 neurons). The human cerebral cortex contains about a million neocortical columns.

I don’t think it is that minor. Your brain has to fire a huge number of nurons to make a 64 bit add. It sounds to me like the nurons firing is pretty minor.

Except the a neron doesn’t in of itself do anything with an isolated purpose as opposed to a machine code instruction in a CPU.

I mean a really really rough analogy (but, at least to me, easier to understand) is cars vs. people:

Cars do one thing, pretty much: Move forward by converting chemical energy in gasoline. People do all sorts of things, but most of them have at least some mechanical component.

My car gets 25 miles to the gallon. How many miles to the gallon does a human body get?

Really, the question is meaningless. You could try to extrapolate the caloric value of gasoline and apply that to walking, but that’s stretching.

Similarly, the human brain does all sorts of non-linear non-discrete analog processing of things outside and (more importantly) loopbacked inside the brain. It’s meaningless to even talk about absolute processing power, since the ‘programming’ is intrinsic to the ‘hardware’ or the neural structure. If you were to attempt to use the neural structure for maximum processing power, you’d have to reconfigure it destroying things like personality, vision, etc.

Also a particular neural network (what our brains seem to be, really) doesn’t really
have a reasonable linear equivalent. You can get functional equivalence through approximating it with a linear algorithm (Rice theorem, or something like that, it’s been a while), but the very structure of neural networks goes agains the whole idea of sequential steps of execution (and hence operating frequency, or even processing speed overall).

I meant “meaningful sequential steps”. I mean, a nueral network can be executed sequentially, but the sequential steps don’t “make sense” to us semantically as steps.

Naah. The comparisons are still essentially meaningless. And:

Is a huge oversimplification of what a synapse does, especially if you think in terms of simulating synaptic behaviour. There’s a research group in the US that deals with computer simulations of synapses - neurotransmitter propogation, synapse morphology, the whole enchilada. To simulate a single synapse - IIRC - required a 64-CPU supercomputer. For ONE synapse. A brief overview here. ISTR that the researchers proposed that each synapse has a fair wedge of computing power - each synapse is not just a point where information is exchanged.

So, how fast is a human when they fail to do things that even a puny computer can do - e.g., taking logs to many decimal places, etc.?

All that tells us is that emulation is hard. It takes a computer running at close to a gigahertz to really do a good job emulating an NES, for instance, just because they’re different types of computers. But even at that, an NES and a PowerPC are far more similar to each other than either is to a brain. It takes a very good silicon computer to emulate even a very primitive neural computer, true. But then, it also takes a very good neural computer to emulate a very primitive silicon computer, too. I can go through algorithms in my head, and if I programmed myself right, I could probably even do things like take logs to nine decimal places. But I’d still do it much slower than my pocket calculator does. Does that mean that my pocket calculator is the superior computer, because it’s hard for me to emulate? No, it just means it’s a different kind of computer.