login about faq

To prove you're not a spammer, email newuser.lgqa@gmail.com with the subject "Account Request" to request an account.


What's the Fastest Speed that a CPU can ever reach without breaking the Laws of Physics? / With Moore's Law increases in CPU Speeds, you would think computers becoming infinitely fast, but NO CAN DO!

  • Speed of Electricity - Overhead = Maximum Speed of Electric Computers
  • Speed of Light - Overhead = Maximum Speed of Quantum Computers
  • Faster than Light (FTL) Computing = Time Travel Computers

At the rate computers continue to accelerate like in the past, we'll max out in less than a century.

"... a time-computer would not actually fly through the centuries, of course. It would simply transmit information back in time to itself, allowing instant solutions to problems that would take an ordinary supercomputer billions of years to solve ..." says Todd Brun, a physicist at the Institute for Advanced Study in Princeton, N.J.

asked Jul 30 '10 at 15:29

r0bErT4u's gravatar image


edited Jul 31 '10 at 21:55

  • Would you mind translating that down to layman's term in Ghz, Thz, etc.


(Jul 31 '10 at 15:31) r0bErT4u r0bErT4u's gravatar image


(Nov 02 '10 at 07:24) r0bErT4u r0bErT4u's gravatar image

12next page »

Something I am good at answering:

Everything in the world is limited by the speed of light. in many cases electricity is assumed to be going that fast, however I believe it goes a little slower, and it vary depending on other things. Light moves at 3x10^8 meters/sec. we now have processors that are 32 nm (32x10^-9 m) between each component. That means the time it takes to get from one component to the next (minimum) is 1.07x10^-16 sec or .106 femto sec keep in mind for one instruction the processor has to go through 1000s of these components (ball park estimate) so it will take about 106 femto sec per instruction (once again ballpark)

Another thing you have to take into account is prorogation delay. Prorogation delay is the time it takes for a signal to arrive on one side of a component to the other, if you have a bunch of components in series you add up the times of each component. This is by far the longest time waist of the computer. In addition, I do not have any ballpark estimates on the time delay. The distance between each component on the chip is unlikely to get much smaller. In quantum theory there is apart that talks about "finite potential wells." a brief, and I mean brief expiation is that an electron can exist outside of where you would think it would. Imagine an open top box and there is a ball in there. Now like an electron it is impossible to know the balls exact position at any given time. So the best we can do is find the probability of the ball being in a certain location; however, the "probability function" says the ball has a small, but still relatively possible chance of it existing outside the box, and if it is too close to the next component or wire it can jump over causing errors.

In addition, we get to talk about Moore's law, which says every 18 months you will double the number of transistors on a chip, well that is breaking down now. We are starting to reach the limit of how many transistors we can fit on a chip. When you think about it, you cannot keep dividing anything in this universe by twos, when you come down to a single atom you cannot break it down to much more. (Yes, you can but for our purposes, it is useless) transistors need a set number of atoms to work.

OK as a summary: we are limited by the speed of light and the distance between each component. The closer the compounds are can cause an electron to cross to a different part of the circuit causing issues. Propagation delay also drastically limits the speed of a processor...

Conclusion: speeds are not going to get much faster with the current technology. In addition, as far as quantum computing, I am not holding my breath...

answered Jul 31 '10 at 12:10

trueb's gravatar image


edited Aug 01 '10 at 00:37

Excellent Answer! Would you mind translating that down to layman's term in Ghz, Thz, etc.


(Jul 31 '10 at 14:57) r0bErT4u r0bErT4u's gravatar image

well frequency is defined as 1 over the period, a period is the time it takes to complete one clock cycle. so neglecting prorogation delay. which is a huge part of the time it takes, and it honestly is not very realistic to neglect it turns out that if you have a period of 106 femto sec is 9 THz. however with propagation delay you will find that as far as clock speed it is right about where we are at now, which is why companies have not made processors much faster(clock speed), but add more cores.

(Aug 01 '10 at 00:45) trueb trueb's gravatar image

WOW! 9 THz would be awesome!! I've only heard of 10 GHz, but haven't seen any?!?

(Aug 01 '10 at 01:59) r0bErT4u r0bErT4u's gravatar image

You're wrong. Part of the technology involved with transistor count and nm tech is making those components work with fewer atoms and less space between the transistors. Therefore, we could eventually have zeptometer technology. Just maybe.

(Aug 01 '10 at 10:15) HHBones HHBones's gravatar image

Zeno's The dichotomy paradox

“ That which is in locomotion must arrive at the half-way stage before it arrives at the goal. ” —Aristotle, Physics VI:9, 239b10

Suppose Homer wants to catch a stationary bus. Before he can get there, he must get halfway there. Before he can get halfway there, he must get a quarter of the way there. Before traveling a fourth, he must travel one-eighth; before an eighth, one-sixteenth; and so on.

(Aug 01 '10 at 10:25) r0bErT4u r0bErT4u's gravatar image

HHBones - yes but you cannot split the atom and still have it work the same way as with the whole atom, a transistor has 2 different materials and are layered PNP NPN. without getting into specifics P is one type of material (impurities added to silicon), and N is another type (different impurities added) the order describes the layering of them. now today they are going away from silicon for high end chips because of some of it's limitations.

anyways what im getting at here is you can only shrink a transistor down so much before it is no longer a transistor.

(Aug 01 '10 at 13:44) trueb trueb's gravatar image

r0bErT4u - love the Zeno reference... :)

(Aug 01 '10 at 13:45) trueb trueb's gravatar image

HHBones - a helium atom is 31 pm in radius so 62 pm in diameter, and helium is the second smallest atom. a zeptometer (which btw i have never heard of until now) is out of reach. a electron is 2.8 fm in radius 5.6 in diameter and how can u have a wire so small it cant hold an electron? how can u have a wire and a transistor so small it cant hold an electron, and still have it carry electrons though it?

now dont get me wrong, quantum computing would help address being it will harness the state of the electron, but i'm not holding my breath for them to create a fully working quantum computer, let alone have one in every home.

(Aug 01 '10 at 13:55) trueb trueb's gravatar image

The fundamental limit on the rate of quantum dynamics: the unified bound is tight Lev B. Levitin and Tommaso Toffoli

for every unit of energy, a perfect quantum computer spits out ten quadrillion more operations each second than today's fastest processors.

ten quadrillion more than what? How many Operations Per Second do today's fastest processors spit out?

(Aug 02 '10 at 14:30) r0bErT4u r0bErT4u's gravatar image

Moore's Law hasn't ended, just lithographic processes and silicon chips. There are a variety of new technologies chomping at the bit to take over where silicon transistor chips left off.

Here's one:


(that guy may be more than just a little over-excited, but the technology is real)

MIM diodes and memristors may have a large impact on the amount of practical performance you can get out of a chip as well by shrinking and consolidating chip features (and even bringing in features that used to have to sit off the chip).

We have a long, long way to go before we hit performance limits of computing, and once those limits are reached undoubtedly they'll just start building bigger. Can't get any higher htan an exaflop? No problem, build a cluster of 12 exaflop chips. Most modern software is moving toward multithreading (you can even write multithreaded web applications in javascript on real browsers), so the performance advantages of multi-core CPUs and clusters is being realized on a consumer level.

(Dec 11 '10 at 13:30) Justen Robertson Justen%20Robertson's gravatar image

Justen Robertson - i beleve you are missinformed of what moore's law is and so is the article moores law only says that every 18 months the number of transistors will double on a chip, it has nothing to do with clock speed or the performance on benchmark tests.

two this question is not about moore's law directly. this question has to do with clock speed not FLOPS, flops and clock speed are different and seperate from each other, clock speed can impact the FLOPS a computers tests at, but is not necessarily a direct relation of how fast a computer can perform

(Dec 11 '10 at 20:43) trueb trueb's gravatar image
showing 5 of 11 show all

If light is seen as a wave, then who's to say there is a limit? Surely there must be a type of wave faster than light waves ;)

answered Jul 31 '10 at 11:49

Seb's gravatar image


no, i dont think there is. light is the fastest, unless you are in a vacum, where every form of wave travelss at the same speed

(Jul 31 '10 at 12:00) Tim Fontana Tim%20Fontana's gravatar image

light is the universal speed limit it can be thought of as a wave or a partial, it is kind of an odd beast. even a wave in the ocean takes time to get from one part of the ocean to the next. but the speed of a light wave is consent 3*10^8 regardless of the wavelength or magnitude.

(Jul 31 '10 at 12:23) trueb trueb's gravatar image

There is no limit. While we can use the speed of light in CPU's, that limit can by bypassed by using multiple streams of light. We actually do that now. 1 pipeline of light not fast enough, then use 2 or 3 or over 9000 :)

All that's needed is them being smaller so we can fit more stuff inside of the CPU cores.

So we can reach a point where the bottle neck becomes the speed of an electron or photon, We will go past that by simply add more cores and pipelines. For example in a videocard, a single chip will have hundreds of cores.

answered Jul 31 '10 at 15:14

Razor512's gravatar image


edited Jul 31 '10 at 15:15

IBM researchers have created a low-power device that can transfer information at high speeds using light.

"The device, called a nanophotonic avalanche photodetector, is the fastest of its kind and could enable breakthroughs in energy-efficient computing that can have significant implications for the future of electronics," IBM said in a statement, which included a video that details how the device works.

The device announced Wednesday, which was detailed in an article published by the journal Nature, is capable of transmitting data at speeds up to 40G bits per second, using a 1.5 volt power supply, IBM said. The light signals are carried over silicon circuits instead of the copper wires that are now used to carry electrical signals between chips.

The device, which is made using existing semiconductor technology, is particularly significant because it uses roughly 20 times less power than previous devices, IBM said.

The goal of researchers is to build an on-chip optical interconnect that would allow the construction of computers capable of exaflop performance, equivalent to 1,000,000,000,000,000,000 floating point operations per second (flops).

To put that in perspective, a 1 exaflop computer would be nearly 600 times faster than the world's most powerful computer, a Cray XT5 called Jaguar at the Oak Ridge National Laboratory in Tennessee, which is was benchmarked at 1.75 petaflops, or 0.0175 exaflops.

(Jul 31 '10 at 15:28) r0bErT4u r0bErT4u's gravatar image

cpus probably can become infinetley fast because of multi cores? im not sure if im right or not? or multi socket mobos

answered Jul 30 '10 at 15:40

Tim%20Fontana's gravatar image

Tim Fontana

I'm looking for the Maximum Speed a CPU can ever reach, but thanks for playing =0p...

(Jul 31 '10 at 10:17) r0bErT4u r0bErT4u's gravatar image

by your logic you would need an infinite number of cores... some thing we can never reach...

(Jul 31 '10 at 13:22) trueb trueb's gravatar image

didnt think of that, but we can keep adding cores, so there should be no limit, until we run out of space, and then cpus get bigger again

(Jul 31 '10 at 16:05) Tim Fontana Tim%20Fontana's gravatar image

Parallel processing is a difficult thing, you have to implement a parent process that will do the fork off, then you need locking... it is all very well having a million processes on a million cores... but at some point they have to write data somewhere. You have to make sure the processes children do not obliterate the data when they write to the same space at the same time.

(Dec 10 '10 at 10:11) Deadpan110 Deadpan110's gravatar image

Parallel processing isn't so much hard as it is different, to the same degree that object-oriented programming is different from procedural programming. You can't bootstrap existing software into parallel processing easily, but if you design the software from the ground up to take advantage of it, it's not an extraordinarily difficult concept to grasp.

(Dec 11 '10 at 13:35) Justen Robertson Justen%20Robertson's gravatar image

it is impossible to say....although the speed of light and such to come into play...you have to remember there is much more too that...for example..CPU even today can execute multiple instructions at the same time....so as the architecture of CPU changes...we can go faster and faster...the CPU or the future will probably have a much different architecture..which can execute thousands of instructions at the same time....that my guess

we wont max out anytime soon...plus we have multi core CPU..so yeah

answered Jul 31 '10 at 11:43

SJP's gravatar image


in addition to murphy's law breaking down, there's a limit as to how wide a transistor gate can be in a processor. at about 10 nm the electrons that make up the stream of electricity start to magically jump across the material (no matter what it is) and the processor doesn't work, eternally shI+ing on itself.

answered Jul 31 '10 at 13:50

samiam2013's gravatar image


Intel and chip-tech house Numonyx unveiled a new technology on Wednesday that the companies claim will enable non-volatile memory to break through NAND's 20nm barrier and scale down to process sizes as tiny as 5nm - and do so cost-effectively.


(Jul 31 '10 at 15:02) r0bErT4u r0bErT4u's gravatar image

well its not "magic," we know why it has to do with the finite potential well i mentioned, it is just odd

(Aug 01 '10 at 00:46) trueb trueb's gravatar image

I didn't say that it couldn't happen, we can make processors with gate widths smaller than imaginable, but if amount of energy travelling through processor doesn't change, it will shi+ on itself.

(Mar 08 '11 at 17:18) samiam2013 samiam2013's gravatar image

The maximum would be the speed of light. Let's say that for some reason the processor in question is a single-core that lacks optimization. It's a quantum processor. The die is about 15mm. Light travels at 3x10^8m/s. Break that down to millimeters and you get 3x10^10mm/s. Let's say that it's a 64 bit processor with 16-bit instructions. Let's also say that one C floating-point instruction translates to 4 16-bit instructions. So, for that instruction to reach the end of the die (going through processing) would take about 3*10^-9.999ish seconds. That info is translated into other code by the video card, but that's not what we're concerned with. This, put into FLOPS, is about 1 petaFLOPS. That is equivalent, roughly, to a modern single-core running at about 1.5-2 PHz, or petaHertz.

answered Dec 11 '10 at 14:18

HHBones's gravatar image


I probably messed up severely on the math, but that's my job.

(Dec 11 '10 at 14:19) HHBones HHBones's gravatar image

is it not meant to be nm?

(Dec 11 '10 at 14:27) Tim Fontana Tim%20Fontana's gravatar image

i think you did mess up on some of the math but overall concept is correct

(Dec 28 '10 at 10:58) trueb trueb's gravatar image

The die size is in mm. The chip is also measured by how precise the fabrication process is, i.e. how small they can make the transistors. The current smallest transistors are at around 32nm.

(Dec 28 '10 at 16:59) HHBones HHBones's gravatar image

I'm sure we'll be going to an era when we're trying to minimize electrical components, and replacing them with light or something else. Somehow.

I know we will be able to see back in the past, but I really don't think we'll be doing anything else with it.

answered Dec 10 '10 at 08:53

Jourei's gravatar image


Quantum computing is the be all and end all of modern day computing, it will render current methods of encryption obsolete. Put into simple terms, you feed it a mathematical equation and it will spit out every possible answer that every universe holds in every reality and every given time... you then filter the results. This is faster as it works at a quantum level but current day physics have yet to figure out a way to free it from interference (matter can be subjected to disturbance by the slightest of things i.e, a car driving past, someone switching on their household appliances 3 streets away or even a flick of a light switch): http://en.wikipedia.org/wiki/Quantum_computer As for relational speeds, that would depend on the filter mechanism and whatever medium is used to store and write data.

answered Dec 10 '10 at 10:36

Deadpan110's gravatar image


Re: cryptography, a new algorithm was released recently which is immune to all known (or theorized) quantum decryption methods. Crypto isn't going anywhere either, it's just going to change.

(Dec 11 '10 at 13:38) Justen Robertson Justen%20Robertson's gravatar image

Actually, with something called 'quantum teleportation', you can theoretically have an infinitely powerful computer.

Quantum teleportation

answered Dec 28 '10 at 10:43

HHBones's gravatar image


"can THEORETICALLY" that is the problem, theory breaks down in our world quite often...

(Dec 28 '10 at 10:59) trueb trueb's gravatar image
Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here



Answers and Comments

Markdown Basics

  • *italic* or __italic__
  • **bold** or __bold__
  • link:[text](http://url.com/ "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported



Asked: Jul 30 '10 at 15:29

Seen: 23,727 times

Last updated: Mar 09 '11 at 05:00