47

As the title suggests, Does a longer Ethernet cable slow your connection down?

  • 9
    Signal propagation time through cable is not significant. The real problem is package loss, and the stated max limit. The max rtt of 100mbps ethernet leaves us around 250m cable, which is just over 100m back and forth, and some time for the nic's to do some processing. – Filip Haglund Aug 18 '16 at 6:24
  • 1
    @peterh: That is a very optimistic estimate. If you assume 16,000 kilometers distance (which is certainly too little) and account for the approx. 30% increase due to the photons travelling zig-zag inside the cable (see physics.stackexchange.com/questions/80043/…), plus consider that c is only 2/3 of what it is in vacuum, you have 105ms one-way. Thus, upwards of 200ms, no routers. Now, the Univ. of Melbourne pings to an astonishing 166ms RTT for me (via 19 hops), but it turns out it's hosted in the Amazon cloud on US west coast... :-) – Damon Aug 18 '16 at 10:30
  • 1
    @Damon :-) Yes. But consider the packets should also go back. Australia is nearly on exactly opposite side of the Earth as Europe, so I think we can calculate with 2*20000km. With +30% zig-zag it is 52000km, with 2/3c it comes to around 250ms ping reply time. – peterh - Reinstate Monica Aug 18 '16 at 10:34
  • 1
    @Damon That question on physics SE assumes the cable is multimode, long distance communications fiber is singlemode, so the light doesn't zigzag. – Peter Green Aug 18 '16 at 20:35
  • 1
    @PeterGreen: Not like it really matters a lot (especially since the +30% zig-zag are ridiculous compared to the +100% detour in the Persian Gulf), but... do you have a reference which I could read for curiosity? Fibers are in the 8-10µm range. A photon does not really have a size being a wave-particle-thingie, but since traditional microscopes have a resolution of about 200 nano, surely a photon's "assumed size" must be definitely in the approx. 0.2µ range. Thus, unless a cable is perfectly straight, there must necessarily be collisions with the wall and reflection, and thus zig-zag. No? – Damon Aug 19 '16 at 9:30

13 Answers 13

5

The electric signal will be slowed down by a minimal amount (afterall it travels almost at light speed, exactly at 0,951c), how much time does light take to travel for 100 meters?

timeTaken  = 100/(299792458*0,951) = 0,00000035 seconds

So it just takes an extra 0,00035 milliseconds which is just 350 CPU cycles (on a 1 Ghz CPU). However the longer the cable the weaker the signal becomes, once the signal is weak enough it will starts to lose bits of information because of interferences, each time a bit is loss, something in the network layer sees that a checksum/parity check fails, and ask for that packet again.

Asking for a new packet will take a very long time.

So as long as signal is strong in the cable, the slowdown would be minimal (it is greater than I expected anyway).

Once you start losing information because cable too long, the slowdown would greatly increase.

Also note that certain communications protocols are timed, so if the cable is too long it may not even be usable because it would go out of sync (that's a by-design issue)

  • 1
    Good! You calculated! :-) Thus, it will be a packet loss. – peterh - Reinstate Monica Aug 17 '16 at 21:38
  • 5
    Note that the velocity factor of a CAT5 or similar cable is not 1. Simply dividing by the speed of light does not apply for most electrical media. – user2943160 Aug 18 '16 at 1:54
  • 1
    does not apply but provide an approxximate lower bound, so instead of 0,0003 milliseconds, the time is increased by "something more". It is not an exact computation of course, but gives an estimate – GameDeveloper Aug 18 '16 at 8:44
  • 1
    Signal speed in copper is around 0.951c where c is the speed of light. – wilsotc Aug 21 '16 at 15:22
  • Am I really criticized for a 5% error? XD – GameDeveloper Aug 21 '16 at 22:43
39

No, it will not slow down a connection, but you need to be aware of the maximum length of a copper connection which is 100 meters. This needs to include the length of your patching cable from the host to the data point and also patch frame to the switch.

However, when using Cat 6 with a 10 Gbit/s interface, you can only use up to 55 meters and would need to use Cat 6A to achieve 100 meters for this type of transmission.

So if you are going above the specified maximum cable length, you will start to see problems not just relating to speed, but also to loss due to the nature of electrical current running through the cable.

The 100 meters only applies to a single run without any intermediary network device such as a switch. If you have a switch in between, you can obviously extend this from port to port which the maximum would apply to for each cable run from device to device.

Using fibre connectivity, you can extend the range based on what type of fibre and light which is beyond the scope of your question.

  • 34
    Note that, due to the nature of TCP, data loss (e.g., from overlength wire) can cause a perceived slow down because the connection has to wait for lost or bad packets to be retransmitted. – Chris Bouchard Aug 17 '16 at 19:22
  • 9
    So just an FYI... electrons in a wire do take time to travel a distance. Electrons in a cat5e wire move at a speed of 0.64 * speed-of-light. So assuming a cable length of 100m the time it takes an electron to move that distance is: approximately 521 nanoseconds. Or time = distance / speed = 100 meters / (0.64 * 3e8 meters-per-second). – Trevor Boyd Smith Aug 18 '16 at 13:33
  • 1
    I was hoping for some more numbers, say, network devices expect to receive a 'successful' signal between these thresholds, and when transmitting data over copper, the signal 'strength' drops below the minimum after 100m, but more precisely. – Pysis Aug 18 '16 at 14:19
  • 14
    The electric field propagates through copper at .64C. The rate that an individual electron actually manages to travel along is more on the order of magnitude of human walking speed. – Affe Aug 18 '16 at 17:50
  • 1
    @Affe maybe in some high-voltage scenarios. Wikipedia says that one ampere squeezed through a 2 mm^2 wire gets us to 0.000023 m/s. Human walking speed is one or two meters per second. – John Dvorak Aug 18 '16 at 21:44
33

For all practical purposes, there will be no effect on the speed of your connection.

There will be a very insignificant amount of delay due to long cables. This won't affect the maximum speed of your connection, but it would cause some latency. pjc50 points out that it's about a nanosecond for every foot of cable length, which is a good rule of thumb used by many engineers when developing systems which are very dependent on latencies on those timescales.

In reality, you will never notice a difference. A "fast" ping time on the Internet is 10 ms, which is 10,000,000 ns. Adding even a few hundred feet of cable isn't going to have a noticeable effect at that point. In fact, nearly every step of the way involves delays which are more extreme than those seen from signal propagation. For example, most consumer grade routers will wait for the last byte of an incoming packet to be received and check it for errors before sending the first byte of the packet on its way. This delay will be on the order of 5,000 ns! Given that the maximum length of cable you can run (per the Ethernet spec) is 300 ft, the cable length could never cause more than 300 ns of delay due to the cable!

  • 2
    It's not the propagation delay which is a problem, but packet loss at very long cables. The speed will theoretically be the same, but the "perceived" speed can become much lower as packets are lost and have to be resent. – vsz Aug 18 '16 at 8:38
  • @vsz From the pages I have seen, its generally agreed that packet loss is very minimal up to the max cable length in the Ethernet specifications of 100m. – Cort Ammon - Reinstate Monica Aug 18 '16 at 16:14
  • 3
    Yes, but what if the cable is longer than what the specification permits? There will be a length where it doesn't work at all, so just shorty before that, might we find a length with a huge pocket loss but some still getting through? – vsz Aug 18 '16 at 16:16
  • 2
    @vsz I think that would be a very different question. I don't think the OP intended the question to explore non-spec usage of Ethernet. Anything will break if you misuse it sufficiently. – Cort Ammon - Reinstate Monica Aug 18 '16 at 17:54
18

Sort of, to a very tiny extent.

The longer your cable, the higher latency you experience - gamers call this "ping" time. However, the effect is about one nanosecond per foot of cable, which is unlikely to be noticeable in most cases. Especially as a single ethernet cable is limited to 100m.

This matters for high-frequency trading and occasionally for email.

It doesn't, of itself, affect the throughput or "bandwidth" of your cable.

  • 3
    +1 For the speed of light story. That made my day. With some protocols (like SMB, latency will affect throughput, as I learnt the other day)... – Aron Aug 17 '16 at 17:39
  • 1
    Yep, there's a reason some companies pay gajillions of dollars to have prime located server/network racks in specific interchanges with cables cut as short as possible. If you're that company, you'll know you need it. Everybody else, you don't care. :-) – Brian Knoblauch Aug 19 '16 at 18:49
  • 1
    It also matters (in the 10-100ns ballpark) for anything else that depends on accurate timing, such as telecom networks. (Source: my job is to make sure telecom networks have accurate timing :D) – Lightness Races with Monica Aug 20 '16 at 0:26
7

I believe it can, but not in the way most people are thinking about.

Most are thinking of the extra propagation delay through the cable itself. This is valid, but as people have already pointed out, so small that it's essentially always irrelevant.

There is another possibility though. Ethernet cables come in a few different varieties: cat 5, cat 5e and cat 6 are currently in (reasonably) wide use. Cat 5 doesn't officially support gigabit Ethernet, but with a short (e.g., 1 or 2 meter) cat 5 cable that's in good physical condition, you can often get a seemingly reliable gigabit connection anyway1.

With a longer cable, however, you could get enough signal deterioration that a gigabit connection was no longer possible. In this case, I believe you'd normally be a 100 megabit connection instead. In this case, you wouldn't just gain some irrelevant amount of latency--rather, you'd have lost a substantial amount of bandwidth.

This wouldn't have any effect on an Internet connection unless you happen to be one of the fortunate few with more than 100 MB/s bandwidth. Access to local resources could be affected much more drastically though.


  1. All of these use identical-looking RJ-45 connectors; the difference between cat 5 and cat 5e cable usually isn't obvious except by looking at the printing on the wiring to see which it says.
  • Yes. All other answers are theoretical but I've seen THIS happening in real-life. Even if you can get gigabit/100Mbit detected the connection will slow due to retransmissions because of errors due to signal deterioration – slebetman Aug 18 '16 at 5:56
  • 3
    If you are lucky a bad cable may link at a lower speed. If you are unlucky it may link at the higher speed and then start dropping packets like crazy. High packet loss will make your internet connection horriblly slow because TCP interprets packet loss as a congestion indicator. – Peter Green Aug 18 '16 at 20:43
2

The standard is 100m (~333.33 ft; 1m = 3 1/3 ft) before attenuation makes the signal unusable, but the direct answer to your question is yes, a long cable can slow your connection. Attenuation is caused by the internal resistance of the copper which humans perceive as lag/slow down of network connectivity. If the cable is under 100m, the slow down is relatively unnoticeable. It can cause issues if you're coming close to that 100m mark though. And keep in mind that the 100m length is measured from the point the cable plugs into the port on your computer to the point it plugs into a device that regenerates the signal, like a switch or a router. (I've personally had to change out a cable to a printer because the ~97m length caused sporadic communication.)

  • The standard has nothing to do with signal attenuation. The original reason was due to CSMA/CD, which is completely irrelevant in modern Ethernet installations. Today we almost exclusively use switches on Fast Ethernet installations, and GBe doesn't even HAVE CSMA/CD. – Aron Aug 18 '16 at 6:44
  • resistance do not slow down, (unless you start losing bit of information), distance that signal has to travel for slow down. The lag may be cause by 1000 other factors and failures at some points. – GameDeveloper Aug 22 '16 at 0:15
2

In theory, yes.

According to Shannon-Hartley theorem, maximum achievable capacity of a channel with additive white Gaussian noise is [1].

[1] bandwidth * log(1 + SNR)

Long runs of cable decrease both its bandwidth (as high frequencies are dispersed) and SNR (as signal amplitude decreases).

  • Some protocols are designd to rate adapt to varying channel conditions. Ethernet is not one of them. – Peter Green Aug 19 '16 at 20:36
  • But due to it being tight upper bound, under no conditions it will exceed it. – fghkngfdx Aug 22 '16 at 7:21
0

As a specialist in this line, I advice YES it does! But it is too little to affect you unless you had extended it too much. It also considers the quality of your cable line, connections, and others. But all of these are too minimal to get noticed. If you are talking of below 20 meters for a home, please do not bother to ask. These factors are for 100 meters and above. That is the reason why we have optical line.

0

There are two issues to consider, latency and signal integrity.

Latency is directly proportional to cable length. However, assuming we are talking about twisted pair Ethernet cables insidea a building the latency will be negligable compared to delays in equipment and in the long distance connections that make up the internet.

The other issue is signal integrity, if it gets too bad then the link will start dropping significant numbers of packets. TCP thinks dropped packets mean congestion and will drop it's speed accordingly.

If your cable is in-spec and your devices are in-spec and your distance is in-spec then packet loss should be negligable. However there is a lot of out of spec hardware out there, so I would be wary about operating right at the limit of the distance specification.

0

Yes. However,

  1. it is not called an Ethernet cable *
  2. it is not a connection, it's a transmission
  3. humans will not perceive delay introduced by cable length alone

*If you are speaking of a local area network you are probably referring to Category 5 or 6 cable. If you are speaking of a wide area link you are probably referring to single mode fiber optic cable.

0

The electrical signal propagation time for a 100 m maximum length Ethernet cable is only about half a microsecond. This is far less than the amount of time needed for your router, etc. to do their jobs.

This only begins to be relevant when looking at much larger distances: For example, from your computer to the server for a game you're playing; however that number is entirely in the hands of your ISP/its partners and the physical locations of you and the server itself.

0

Long cables will increase your latency since the signal has longer to go. This shouldn't matter much in your case since the signal propagates near the speed of light, the extra 10 meters will be imperceptible compared to the many miles to whatever server you are accessing. There will be some loss of signal over extremely long runs which will reduce bandwidth but shouldn't be significant over 20 meters, 100 meters is the point where you have to start worrying about the length of the run.

0

According to my own experience, it is a laughably misleading opinion.

In fact, it will hurt the latency since the longer distance, but not bandwidth. Supposing you are comparing the internet speed of using the 10m fiber and 20m fiber, you will find that electrical signals travel through an ethernet cable at the speed of light, or 299,792,458 meters per second. That means that it takes about 0.00000003 seconds for the signal to travel through your 10m cable, and 0.00000006 seconds to travel through your 20m cable. You cannot notice the difference here. Not mention if you compare to many miles to whatever server you are accessing.

However, longer distance transmission indeed have signal loss and noise issue, which might have an impact on the internet speed.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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