Supposing that neurons function similarly to transistors: A neuron able to fire $200$ times per second and transistors can be switched on and off more than $100,000,000,000$ ($10^{11}$) times per second. Let's say it fires 1 out of 2 times in average. We have $86,000,000,000$ ($8.6 \cdot 10^{10}$) neurons in a brain, and $4,000,000,000$ ($4 \cdot 10^9$) transistors in medium CPU.

$\text{Units count} \cdot \text{firing probability} \cdot \text{firing rate} = \text{total fires per second}$

A brain's total fires per second: $(8.6 \cdot 10^{10}) \cdot 1 \cdot 200 = 1.72 \cdot 10^{13}$.

A CPU's total fires per second: $(4 \cdot 10^9) \cdot \frac{1}{2} \cdot 10^{11} = 2 \cdot 10^{20}$.

A CPU is faster than a brain by $\frac{2 \cdot 10^{20}}{1.72 \cdot 10^{13}} = 1.16 \cdot 10^7$, or about 12 million times faster. I gave the brain an advantage that every neuron is firing non-stop instead of just 1%, and that they fire at the rate of the fastest neurons.

Why isn't this argument valid?

  • 8
    $\begingroup$ The brain is not a computer, and neurons are not similar to transistors. $\endgroup$ Commented Jul 18, 2014 at 14:53
  • 2
    $\begingroup$ @ChuckSherrington It's an analogy between them. "and neurons are not similar to transistors" They are similar that they both need specific condition to fire. (They both fire too) Anyway, you mean those scientists are wrong even when they are trying to compare those two devices? $\endgroup$
    – KugBuBu
    Commented Jul 18, 2014 at 15:49
  • 1
    $\begingroup$ Check out: biology.stackexchange.com/questions/19767/… $\endgroup$ Commented Jul 18, 2014 at 19:06
  • 4
    $\begingroup$ @KugBuBu This blog post covers it pretty well: scienceblogs.com/developingintelligence/2007/03/27/… The (biological) neuron is a lot more complicated than you think. $\endgroup$ Commented Jul 19, 2014 at 22:20
  • 4
    $\begingroup$ Incidentally, while I seem like an old curmudgeon, what I'm really saying is learning more about the neuron itself will be a better step for your own edification than getting hung up on the analogies. Also, take a look at the enteric nervous system. It's got almost as many neurons as the brain does, but there's a key component missing and that's the number of synapses present. It's be quite a CPU as well, but it's main function is to propel food through the gut. $\endgroup$ Commented Jul 20, 2014 at 1:34

5 Answers 5


There is a basic epistemological problem here that was only touched upon by Chuck Sherrington - everyone is making the assumption that the brain processes the same kind of information as a digital computer. There is no real evidence to suggest that it does, in fact. A digital computer is an instantiation of a Turing machine, which is equivalent to certain kinds of automata. In order for the "processing power" of the brain to be compared to that of a digital computer in the first place, one needs to show that the brain employs representations (discrete entitites/states like bits) and rules (well-defined transitions between states/bits). Nobody has even come close to doing this, even for a subsection of the brain. This would be done by showing that the brain implements some digital computation - David Chalmers' famously [1] explains how this needs to be done. According to current the state of research, the brain seems to be a complex biological system, operating at multiple levels of measurement, and does not process information in discrete terms! And yes, Chuck Sherrington says it, neurons are not simply on/off!

[1] Chalmers, David J. "A computational foundation for the study of cognition." (1993).

  • $\begingroup$ Wait, I calculated how many times per second they can potentially fire, not about on and off. But still need to admit that they are different. (While it's not impossible to make transistor work like neuron) $\endgroup$
    – KugBuBu
    Commented Sep 29, 2014 at 11:10
  • $\begingroup$ Okay. However, your original question made the assumption that we are talking about discrete information (see en.wikipedia.org/wiki/Information_theory). If the information is continuous, we need to analyze it differently (i.e. en.wikipedia.org/wiki/Differential_entropy). Besides, it doesn't just matter how fast a system is able to calculate, it matters how and what it calculates! $\endgroup$
    – user6682
    Commented Sep 30, 2014 at 6:17
  • $\begingroup$ The question is about how fast is able to calculate. Brain is recursive self-improvement device. You can't compare them without removing this feature. The thing I am comparing here is hardware, not software. (Brain's software is hardware-builtin, but still exclude that) $\endgroup$
    – KugBuBu
    Commented Sep 30, 2014 at 10:09
  • $\begingroup$ The point is that the hardware of the brain may not be comparable to the hardware of a computer. $\endgroup$
    – user6682
    Commented Sep 30, 2014 at 10:12
  • $\begingroup$ "and does not process information in discrete terms!" Ultimately it must, because of quantum mechanics... $\endgroup$
    – Calmarius
    Commented May 5, 2016 at 15:08

Have you ever seen IBM's Watson? Watson is composed of a cluster of ninety IBM Power 750 servers, each of which uses a 3.5 GHz POWER7 eight core processor, with four threads per core. In total, the system has 2,880 POWER7 processor cores and has 16 terabytes of RAM. It must be kept in a (very) large refrigerated room.

Watson is a question answering (QA) computing system that IBM built to apply advanced natural language processing, information retrieval, knowledge representation, automated reasoning, and machine learning technologies to the field of open domain question answering. According to John Rennie, Watson can process 500 gigabytes, the equivalent of a million books, per second.

Watson's official debut was on the game show Jeopardy, where it was pitted against two of Jeopardy's best players. It won, but the humans gave it a run for its money. Most believe the win was more due to Watson's faster response time (it was electronically keyed into the buzzer; Watson could activate the buzzer within about eight milliseconds, whereas the human response time to the go ahead light signal is several tenths of a second. In addition, to access memory more quickly, content was stored in Watson's RAM for the game because data stored on hard drives are too slow to access.

Watson was devised as a medical diagnostician (as well as other applications). With many years, scores of technicians, who knows how many billions of dollars, and access to medical encyclopedias, books, medical journals and the internet, it can outperform third year medical students in only one area: oncology (and even then, only lung, prostate and breast CA).

So, one average medical resident (4 years of medical school, and, say, second year of residency) can outperform a very, very fast medical supercomputer 99% of the time. I specialize in Primary Care medicine: Family and Emergency Medicine. With a few exceptions per year, within minutes of initiating a conversation with a patient, by watching the patient, reading his vital signs, and hearing answers to perhaps a dozen questions, I already have narrowed down my diagnoses to 2 or 3 top candidates and several other lesser considerations. I am one individual with one brain which accesses only to the information to which I've been exposed, yet I will outperform a medical supercomputer >99% of the time. In under 10 minutes. (This does not mean that I only spend 10 minutes with a patient. Medicine is about more than the correct diagnosis.)

This is Watson:

enter image description here

My brain is about the size of the "W" etched on the room's window.

You tell me: Which is "faster", a supercomputer or a human brain? It really does depend on your definition of "faster", doesn't it?

Putting Watson to Work: Watson in Healthcare. IBM.
IBM's Watson is better at diagnosing cancer than human doctors
IBM Watson's impressive healthcare analytics capabilities continue to evolve

  • $\begingroup$ Searching answer in "massive" data vs. keep firing until it "rings a bell". It isn't the same task at all. I meant in hardware level, rather at a level that brains can optimize their task. (Which they are good at) $\endgroup$
    – KugBuBu
    Commented Jul 19, 2014 at 21:51
  • $\begingroup$ To chime in on the subject of medical informatics, I think it can be good to have a computer assist diagnosis, because doctors are subject to the same cognitive fallacies as the rest of us: Doctors & statistics $\endgroup$
    – user6682
    Commented Sep 16, 2014 at 7:24

I'm not sure the math checks out in the question (the CPU cycles per second seems awfully high), but I think there are some useful principles to keep in mind regardless of the details of the math.

So let's assume that we do have a computer that can perform more operations per second than the combined sum of all action potentials in the brain per second. Is the computer faster than the brain?

The answer is it depends on what the question is. There are certainly types of information processing that computers are much faster at than human brains. A cheap calculator can solve the problem 2854 x 239 much faster than the average human brain. But brains tend to be much faster than computers at pattern recognition type problems.

The main point is that looking at the speed of transistors and neurons is the wrong level -- or at least it provides an incomplete picture -- to be thinking about the speed of information processing. Not that the 'hardware' doesn't matter, but other factors matter too. For a computer, the software matters. For a brain, the network of connections between neurons matters.

  • $\begingroup$ It's the cycles of a singel transistor. The score of a CPU is the potential effort of the CPU. (Same for the brain) $\endgroup$
    – KugBuBu
    Commented Jul 19, 2014 at 21:42

Anology Taking the analogy and calculations directly, you are assuming that the fundamental computing unit of the brain is the neuron; we do not know if this is true. It could be a cortical column, a group of several neurons, the neuron, a dendritic branch (a fascinating review paper!), a synapse, receptors or neurotransmitter vesicles (how about glial cells?). This paper describes the somewhat accidental fascination with the historical view of a neuron being 'the unit'.

So, to the numbers. If one treats a singular neuron's action potential as a 1/0 like a transistor, sure you're numbers kind of make sense. (Not sure where you got the 1% figure from?) However this is making the assumption I eluded to above. If you were to run the numbers with synapses, the brain might win.

You are further assuming that a 1/0, spike or no spike, representation of a neuron is all that the brain does. We do not know how the brain encodes information -- as others have said, it might not work in the same serial 1/0 fashion that we have engineered computers to operate. Information/processing in the brain could be via individual neurons firing rates, population firing rates, spike timing, etc. (sections 1.5, 1.6 and 1.7 review some). Imagine that the timing of when a transistor flipped compared to another was important! Moreover, this only considers action potentials (or EPSPs if one redoes the numbers with synapses). Neurons (and synapses) are extremely non-linear and a lot of information/processing might be undertaken before or after an action potential (EPSP). This is the age old question of analogue (brains) vs. digital (modern computers).

Serial vs. parallel I think an important point missing from previous answers is the fact that the brain is massively parallel. That is, there is not one processor serially undertaking instructions.

A typical CPU (core to be specific) in a computer may be very fast in terms of calculations per second -- arguably a great deal faster than a single neuron -- however, there are very few of these individual units when compared to the brain. This is why supercomputers take advantage of hundreds of CPUs (and cores). However, the numbers of these cores is extremely low when compared to the brain (if we are assuming that a neuron is the fundamental unit, see above), where an analogy might be that each core/neuron is a lot slower, there are however millions (trillions if 'the unit' is a synapse) of cores.

Yes, you have considered that a CPU has many many transitors on it, however the CPU itself it still only processing information in a serial fashion. It does this operation, then this operation, then this etc. As I said above, this is why supercomputers use many CPUs as this allows them to do a somewhat parallel process. However this is more complicated as in reality, a computer's parallel programme is typically just a serial one split a few times. For example, you want to calculate what A+b is where A is a huge matrix and b is a constant. One can do this in serial by iterating through all the elements in A and adding b; or, one can 'parallelise' this by using 10 CPUs and splitting the matrix in to 10 subsets and having each CPU do a serial operation on their own subset and finally concatenate the resulting subsets back in to A again. This isn't really parallel computing but rather smaller serial operations (parallel programming is hard). The brain is truly parallel. (The quote in the answer to this previous question has a great analogy in it.)

To compare the two is fundamentally flawed as the way we quantify the performance of a CPU does not apply to the brain. It might be possible to compare them if a common metric could be devised, but this would mean we would need to understand how the brain processes information in the first place, which is a long long way off.

As a final note, have a look at this article.

(Edited to include more detail as there was interest/comments.)

  • $\begingroup$ I calculated the calculation that the whole neurons can do together (While actually you can fire only 1% of them at once) [not a singel one], and one 1 billion cpu. $\endgroup$
    – KugBuBu
    Commented Sep 29, 2014 at 10:48
  • 1
    $\begingroup$ Citations for backup & further reading would make this a great answer @James! $\endgroup$
    – Krysta
    Commented Sep 29, 2014 at 12:17

The computer analogy fails on several counts. Most notably, the status of a "bit" in a computer system is completely determined at any one time by the running program. Neurons and synapses are quantum entities, whose state at any one time is a superposition. Both neurons and synapses are living things, to some extent acting independently. Each synapse is a quantum nanocomputer, vastly more powerful than any existing computer and performing a "calculation" that cannot, even in principle, be carried out by a computer.


  • $\begingroup$ not even a quantum nanocomputer? $\endgroup$
    – faustus
    Commented Jan 3, 2018 at 10:29
  • $\begingroup$ Popular focus on "quantum computing" has focussed on entanglement, which is just one quantum effect. Google, the leader in this field, is also interested in quantum effects in biology, specifically in the brain (amazon.ca/Quantum-Effects-Biology-Masoud-Mohseni/dp/1107010802/…). The synapse is, in effect, a quantum nanocomputer but entanglement has little to do with the issue. $\endgroup$ Commented Jan 4, 2018 at 16:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.