62

A network engineer professional is overseeing our town's installation of fibre and was explaining how much faster it was.

I pointed out that it wasn't a silver bullet as you had to pay the provider for a certain bandwidth. For example, I explained, I have 20 Mbit/s copper for about €30 per month, and 30 Mbit/s fibre cost about the same, however it wasn't going to be ten times faster. 200 Mbit/s would cost you €70 pm so there was a price.

He strongly disagreed and said the reduced latency meant 30 Mbit/s fibre was twice as fast as 30 Mbit/s copper.

Now, I get that reduced ping times and latency mean time to the first byte is faster, however if I'm downloading a 1 gigabyte file, 30 Mbit/s is 30 Mbit/s, right?

I'm not sure how it affects streaming, however was he right, or talking rubbish?

23
  • 57
    I would ignore anything that man says from now on :\
    – Ron Trunk
    Commented Feb 21, 2018 at 18:56
  • 9
    I would assume that since you have three network engineers form various parts of the world, all telling you the same thing (in different ways), that you understand you should just nod and smile when he opens his mouth, but double-check anything he says before you believe it.
    – Ron Maupin
    Commented Feb 21, 2018 at 19:59
  • 4
    I just have one word for the naysayers. Contention. Because of the potential greater bandwidth of the backbone feeding the street you are more likely to get the full 30mbps even at busy times. In the UK fibre to the cabinet is being rolled out pretty much everywhere and fir the lucky fibre to the premises is on the way.
    – Stevetech
    Commented Feb 21, 2018 at 22:16
  • 3
    Something that seems to be overlooked here is that long-distance fiber tends to have a lot higher capacity, even if it's only provisioned at 30Mbit. Commented Feb 21, 2018 at 23:46
  • 6
    The problem is what is the definition of faster?
    – jrtapsell
    Commented Feb 22, 2018 at 13:54

9 Answers 9

72

30 Mbit/s is the same speed, no matter if it runs over copper or fiber.

However, there are important link parameters other than link speed/pure bandwidth, so there may be differences. First, latency on fiber can be better than on copper depending on the line encoding - fiber requires much less elaborate encoding (see below) than e.g. xDSL. However, lower latency doesn't make the fiber faster but sensitive applications may respond faster.

Second, fiber's scalability is much better - in the future, you can just call your provider and order more speed. Speed on copper may be very limited, depending on the line length and quality. Speed on fiber is practically limited by your budget only.

Third, reliability or packet loss ratio is usually much better on fiber than on copper. Copper is generally susceptible to EMI (depending on cable type, cable quality, link length, and environment) while fiber is practically immune.

EDIT: in regard to "elaborate coding":

Fiber commonly uses 8b/10b line code with 20% line/bandwidth overhead, or 64b/66b line code with 3% overhead, but next to no time overhead or delay (less than a microsecond).

xDSL variants use OFDM/DMT and QAM encoding and modulation to cope with the channel's high attenuation/low signal-to-noise ratio. Reed-Solomon forward error-correction (FEC) is added to decrease the effective error rate, causing a transmission delay/added latency of a few milliseconds or a few thousand microseconds. Long lines also need to add interleaving for protection against burst errors, striping consecutive packets into each other - this causes significant, yet additional delay/latency in the order of 20 ms.

In a nutshell, voice-grade copper's low frequency bandwidth and its sensitivity to noise require elaborate encoding and FEC, which in turn significantly increase latency. Of course, when FEC fails and an error cannot be compensated, a retransmission is much worse than the usual 60 ms RTT for (long-line) ADSL.

13
  • 14
    30 Mbit/s is almost the same speed. Depending on what protocol layer it is measured on there can be a difference in the overhead for headers. But I cannot imagine scenarios where it would get close to a factor of two. And indeed latency and packet loss are going to be more important than the small variation there might be in overhead for the headers.
    – kasperd
    Commented Feb 21, 2018 at 22:58
  • 1
    Two other factors. One: This is an artificial cap. When the time comes for an upgrade, that fiber will likely scale up to 1Gbps. The copper will top out much sooner. Two: The copper is likely an asynchronous connection, where the upstream bandwidth is much lower. The fiber is likely fully symmetric. Commented Feb 23, 2018 at 15:43
  • It depends on what you do with it, too. In an online game for example, assuming 30Mbit/s is ample bandwidth, the fibre would be way better to play on!
    – Octopus
    Commented Feb 23, 2018 at 18:09
  • 1
    @Alexander The problem with the low-quality copper is its burst sensitivity and high attenuation. Fiber most often uses simple 8b/10b encoding with extremely low time overhead while xDSL variants use QAM and OFDM/DMT with reed-solomon forward error correction plus interleaving - the FEC and especially the interleaving delays packet transmission and cause latency. The reach of 8b/10b code on high-quality copper is about 10 m.
    – Zac67
    Commented Feb 26, 2018 at 18:08
  • 1
    Propagation over good coax or cat-7+ is actually faster than fiber (VFs: RG8: 77%, cat-7:79%, fiber: 69%), but compared to the encoding delay difference the propagation delay doesn't count much.
    – Zac67
    Commented Mar 25, 2018 at 8:00
28

30Mb/s is 30Mb/s, but ISPs usually sell you “up to 30Mb/s” because the speed of DSL technologies is highly dependent on the distance between your equipment and theirs.

With fibre, you are more likely to actually get 30Mb/s because the underlying medium is less sensitive to distance.

4
  • 8
    this is true, it means that when they sell 30Mb/s copper it isnt actually 30Mb/s. They have scammed us for long.
    – None
    Commented Feb 22, 2018 at 12:02
  • 1
    @Alexis_FR_JP They're not scamming you unless they sell you 30Mbps and fail to deliver it, which they don't do. If they sell you "up to 30Mbps" then you're getting what you paid for regardless of how fast it actually is (unless it's faster than 30Mbps).
    – Mike Scott
    Commented Feb 23, 2018 at 13:31
  • 1
    of course, I was just saying that the engineer wasn't totally wrong. That's legaly okay but for people up to 30Mb/s means near 30Mb/s. Everything else is just marketing, nothing else.
    – None
    Commented Feb 23, 2018 at 13:51
  • Even though fibre-optic cables do not lose that much bandwidth by distance, they still have a bandwidth limit. So when multiple customers share the same fibre and they all generate traffic at the same time, then they might not get the promised 30Mb/s.
    – Philipp
    Commented Feb 26, 2018 at 15:39
19

You might consider pointing out to your "network engineering professional" that the propagation delay in copper is LESS that that of fiber (in most cases).

The difference between the two is on the order of 0.1C. So in round numbers, that's 0.3 ns/m. If we imagine the distance between you and the provider is 10 km, that's an additional 3 µs delay. That's at least 3 orders of magnitude below whatever other delays there might be. You'd need very expensive equipment to even measure it.

This question might be helpful too.

8
  • 10
    You might want to be careful with "copper is faster than fiber" - only ancient RG8 and Cat-7+ is faster than fiber (Cat-5&6 are about the same). Additionally, copper requires more elaborate encoding which also costs time. ;-)
    – Zac67
    Commented Feb 21, 2018 at 19:24
  • 5
    A more efficient encoding would increase the bandwidth (more info per bit time), but would have no effect on propagation delay.
    – Ron Trunk
    Commented Feb 21, 2018 at 19:29
  • Well, I was referring to WAN: for a few 100 m to some km you can use fiber with 1000BASE-LX for instance which has extremely low overhead (GPON isn't much different) while VDSL and especially ADSL add noticeable encoding overhead = latency.
    – Zac67
    Commented Feb 21, 2018 at 19:36
  • @RonTrunk: No effect on propagation delay, obviously, but it can have an effect on latency, which can often have a big effect on perceived speed...
    – psmears
    Commented Feb 22, 2018 at 14:00
  • 1
    @Jim I didn't use either term. I spoke of propagation delay.
    – Ron Trunk
    Commented Feb 22, 2018 at 19:45
18

It doesn't appear that any one has explicitly addressed the reliability of transmissions over fiber vs copper.

It may be true that, for example, your router is throttled to 30 Mbps, but the transmission over copper may produce more errors than over fiber. This usually results in retransmissions (TCP will do this automatically), which will consume some portion of that 30 Mbps bandwidth.

From my own experience, the performance of a copper DSL line can be limited by environmental factors that reduce the lines capacity to provide reliable service. Your router may pump out 30 Mbps, and the router on the receiving end may not receive all of those bits correctly. For TCP, this results in requests for retransmission, which consumes part of the original 30 Mbps.

Fiber equipment is less susceptible to environmental factors, so I would expect it to be more reliable and thus not consume as much bandwidth for retransmissions.

Now you can decide whether 30 Mbps with few retransmissions is faster than 30 Mbps with a higher percentage of retransmissions.

I doubt that your network professional was thinking along these lines, but to answer the question in the title, I think you have to consider transmission reliability as a factor that can indeed make a difference.

3
  • 2
    There are other copper connections than DSL. For example, we demand, and get 100% reliability from T3, metro ethernet, etc. We have never used DSL.
    – Ron Maupin
    Commented Feb 22, 2018 at 1:00
  • 3
    There's a downside to fiber that could be categorized under "reliability": When there's a serious problem with fiber, it can be much harder to fix. I worked at a place several years ago that was served by fiber that was literally cut by a backhoe. The entire street had no internet for the better part of two weeks. Fiber just doesn't splice as easily as copper. Commented Feb 22, 2018 at 2:13
  • 4
    Todd: That's just the teleco not getting around the fix - not an intrinsic problem with fibre. When we have a fibre cut in in regional Australia (ie miles from anywhere), it is surprising for the fix to take more than 24hrs. Usually the repair teams are many hours drive from the problem.
    – Chris
    Commented Feb 23, 2018 at 3:45
12

Copper (ADSL or VDSL) specifies a maximum peak speed whereas fiber specifies a maximum average speed. My glass fibre (FTTH) connection is artificially limited to 50 Mbit/s average. At the start of a large download the peak speed is larger than that, and then I see it dropping until it hovers near the promised average. Without the limiter it could reach 200 Mbit/s or more, but I am not willing to pay for that if it would save me only 10 minutes of my time per month.

When we still had ADSL over copper, the promised peak speed may have been 8 or even 20 Mbit/s, but due to the distance to the exchange (3 km) it was only 5.5 Mbit/s, and the average speed was (for mathematical reasons) even lower. It may also depend on the activity of your neighbors, due to crosstalk between the copper pairs or due to sharing a common backbone (like FTTC, fibre to the curb). It even depends on how well your DSL modem adapts to the transmission quality (equalizer, error correction).

Also interesting is that the bits we are using for IPTV are not counted against the budget of 50 Mbit/s, whereas in the case of ADSL they necessarily are because then it is a hard physical limit. Fibre has so much peak reserve.

8

Latency can make a connection seem to be much faster.

As an example, typical HTTP traffic downloads a number of small documents that each fit in just a few packets. Even while typical page sizes have grown significantly, many of the largest parts of pages are often served via cache, where all you need to retrieve is the header confirming the cache is still valid (if even that much). If you download a total of 60K for a page, that's still only a few milliseconds at 30Mbps. A modest latency improvement can absolutely make the same 30Mbps seem to be twice as fast for common activities. This is especially true as web traffic is iterative: the base html document then asks for css or scripts, which ask for more css or scripts, when then ask for resources like images and fonts. So this extra bit of latency delay can be repeated several times to render a page.

Another example is video game traffic. Many games use frequent but very small packets with simple vector or action information about a character or object, where the total throughput is small but the latency is everything. Here, an extra few milliseconds saved in latency can make a world of difference and make the connection seem much faster.

But it probably won't. It's unlikely the latency will be enough different for you to notice this. Additionally, for http traffic server processing time tends to minimize the differences relative to the total page load time from click to rendered. Most important, for larger downloads and things like streaming content, it's the same 30Mbps either way.

Mostly.

There are some other factors that also generally tend to favor the fiber:

  • More stability and less environmental interference, meaning less bandwidth lost to TCP retransmits
  • The copper line is most likely asymmetric, where your upstream bandwidth is much smaller (not guaranteed, but at this speed asymmetric is far more common). The fiber line is almost certainly 30Mbps in both directions.
  • 30Mbps is just what you're paying for today. When the day comes you need more, you can probably turn the fiber all the way up to 1Gpbs with just a phone call to change your billing. The copper is likely to have a much lower ceiling.
  • The quality of service on the copper is giving you the max speed, where the average can be much lower. Fiber can generally deliver it's max speed much more consistently.

All of this together means the fiber is definitely the better link. By far. None of this, to me, justifies claiming a factor of 2x for what you have right now, but I sympathize with the engineer in your question, struggling to find a way to explain all of that concisely so managers will understand why it's important.

4
  • Thank you Joel for directly addressing the claim that was being asked about! I've heard people describe their high-speed satellite connections as utter rubbish for many tasks due to high latency, despite high bandwidth. The correct answer absolutely is that technically it all depends on just what is done. However, in practice, DNS lookups and much of web browsing favors latency over speed, so the "network engineering pro" is likely right in principle, but "twice" seems unlikely high in practice. See also David Schwartz's answer
    – TOOGAM
    Commented Feb 24, 2018 at 11:01
  • 1
    Even if two connections are "30 Mbps", the one with lower latency will be better. Copper (e.g. VDSL, ADSL, T1/T3) may be around 20 ms for the last mile whereas fiber is around 0-2 ms. For example, HTTPS 1.1 or lesser will require 3 round trips before transferring first byte of content and then browser needs to open additional paraller connections to fetch more resources (only possible after getting the first response after full handshake). A web page with HTML, external Javascript file and a small icon could take 3*2*3 ms = 18 ms over fiber but it could take 3*20*3 = 180 ms over copper. Commented Feb 25, 2018 at 18:26
  • For transferring a big file (requiring tens of seconds with full throughput) the "30 Mbps" is practically the only thing that matters. Copper may require a bit more retransmissions due interference but the difference should be less than 1%. Commented Feb 25, 2018 at 18:29
  • Cable ISPs aren't an extra 20ms to the last mile. For runs between the same two end-points, we're talking maybe 2ms latency difference at most. So it's more like 3*2*3 for 18ms total difference. That may be enough to matter playing counter-strike, but you probably won't notice it loading CNN.com. You might get lucky if the fiber ISP is able to give you a shorter run (that's gonna matter more), but then the copper ISP might do the same thing... it's down to who has the better right-of-way. Commented Feb 26, 2018 at 15:26
4

The link speed is the link speed. It is how many bits per second can be sent over the link. That is between your end of the link and the ISP end of the link. The latency involved in that link will be based on the distance between those two points. The difference in latency between copper and fiber on a link is so small that you couldn't tell.

You will have other factors that affect your throughput. A big factor is going to be the speed of the hosts/servers involved. Also, the network distance between the endpoints, e.g. how many hops from source to destination, and back, and how congested the path is will greatly affect your throughput.

3

All things considered a 30Mbit/sec link should be a 30Mbit/sec link no matter what the underlying technology.

That said there are several possible advantages to fibre based technologies compared to xDSL based technologies (assuming that's what you mean by copper vs fibre in this context).

Firstly it's possible to offer symmetric 30Mbit/sec connections with relative ease with fibre technologies, but extremely unlikely to be anything other than asymmetric with xDSL. So it's likely that all things being equal the downlinks would be the same on a 30Mbit/sec product with both, but the uplink of the fibre solution may well be substantially faster.

Secondly the overhead from encapsulation can be somewhat lower for a fibre based solution compared to xDSL. For fibre it can be as little as an Ethernet frame header (potentially per jumbo frame, not even 1500 bytes). For xDSL it could be as bad as PPPoEoA. Depending on the kit used and configuration chosen that could manifest as a reduced MTU, be totally transparent to an end user, or worst case the rate-limiting done could be applied at different layers meaning you would be sold a product with the same link speed, but slower TCP throughput in the xDSL case.

Thirdly an an extremely pedantic way that most likely isn't what was claimed the velocity factor (i.e. signal transmission as a percentage of speed of light in a vacuum) is higher in fibre optic as a medium, although over city wide distances the difference is unlikely to be noticeable. The speed of oscillation of the carrier waves in both technologies differs too. VDSL for instance is around 12MHz, but 1300nm wavelength light is many orders of magnitude faster. But I doubt that was what was meant :)

3

I currently have gigabit service over copper cable. My download speeds are typically around 250Mbps and my upload speeds are typically around 40Mbps. At my previous home, I had gigabit service over fiber. My download speeds were typically around 800Mbps and my upload speeds were typically around 550Mbps. These speeds are typical of the difference in the two technologies.

A fiber backbone can typically have a much higher capacity than a copper backbone. Even if you purchase 30Mbps service on both of them, and even if your line actually reaches 30Mbps on both of them, you are much more likely to be able to consistently reach 30Mbps of useful data over fiber. Why? Because a fiber backbone typically has a much higher capacity, ensuring that it is more likely that there will actually be 30Mbps available for your use.

Fiber is almost always implemented in a symmetric fashion, giving you solid upload speeds as well as download. Copper is almost always not, and upload speeds are often much worse than download. Like to make cloud backups? You want fiber.

The latency and line coding differences are typically negligible. It's the higher bandwidth of a fiber backbone and the symmetric bandwidth of a fiber last mile that really does make a difference.

4
  • At my company, all the WAN connections are symmetric, and almost all are copper. We test, and 30 Mbps is 30 Mbps, 150 Mbps is 150 Mbps, or 1 Gbps is 1 Gbps on copper or fiber. We would never purchase an asymmetric line.
    – Ron Maupin
    Commented Feb 23, 2018 at 1:04
  • @RonMaupin The problem is that the question is very vague and we don't really know what the fiber is actually being used for. Regardless, you don't have to go more than a hop or two from your copper WAN link and you will definitely want to find fiber. Commented Feb 23, 2018 at 2:01
  • The point in my answer is that the latency of the hosts, distance congestion, etc. are far more important than on what medium the service is delivered.
    – Ron Maupin
    Commented Feb 23, 2018 at 2:05
  • @RonMaupin But they're not separate things. A fiber backbone is going to have fewer issues with distance and congestion than a "comparable" copper backbone. Regardless of the speed of the end customer link, the quality of the network that connects to the end device really matters, and a typical fiber system is going to deliver a much higher fraction of link speed than a typical copper system. Commented Feb 23, 2018 at 2:47

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