11

According to the notes I'm reading:

10/100 Mbit/s Ethernet refers to the standard that can autosense which speed it needs to run between speeds of 10 Mbit/s or 100 Mbit/s.

Why would autosensing be required? Wouldn't it be best to result to 100 Mbit/s or will this impact the network in a negative way?

  • 11
    If you had been building networks at the time 100 Mbps Ethernet was introduced, the answer would be clear: so you can incrementally upgrade a network without tedious manual configuration of every port. – Todd Wilcox Jun 9 '16 at 3:51
  • 2
    ask yourself: why not jump to 1000 Mbps directly? The same answer will apply to your question. – njzk2 Jun 9 '16 at 15:15
  • I don't see that as duplicate, as it is asking about using half-duplex/duplex when autonegotiation fails and why the difference between 10/100 and Gig operation in that case. – YLearn Jun 13 '16 at 13:58
27

Some devices could only run at 10 megabit/s, so the device at the other end would autosense the speed to match. If a device that has a maximum speed of 10 Mbit/s is connected to a 10 Mbit/s / 100 Mbit/s switch, the switch needs to lower its speed on that particular port in order to effectively (efficiently) communicate with the device.

These days, most devices will autosense between 10 Mbit/s, 100 Mbit/s, and 1000 Mbit/s, but back in the days of "fast Ethernet" the choices were 10 Mbit/s and 100 Mbit/s.

  • 2
    I don't think i've ever heard of a device that was only 100/1000 , the common combinations seem to be 10/100 10/100/1000 1000/10G and 100/1000/10G . – Peter Green Jun 9 '16 at 1:57
  • Yes, I guess you're right. I just meant, "a modern device is likely to be 100 or 1000." – Ted Quanstrom Jun 9 '16 at 2:16
  • 3
    Actually there are devices that only support 100/1000, such as a Cisco Nexus 2248TP Fabric Extender. They are used in datacenters, where it is still common enough to find legacy specialized hardware that handles only 100M, but 10M is completely gone. – Jeremy Gibbons Jun 9 '16 at 3:56
  • 1
    100mbit isn't only used for legacy servers. Low end embedded systems (ex Raspberry Pi) only come with 10/100 ethernet ports. At that level it's not just that an extra dollar of hardware would be a significant increase on the BOM (like with some cheap laptops from a few years ago); but the CPUs on the older models weren't fast enough to make good use of that level of bandwidth. – Dan Neely Jun 9 '16 at 15:38
  • 1
    Some new lab devices still use 10 Mb/s cards even today. For example oscilloscopes, spectrum analyzers, function generators can all be found today new with just a 10 Mb/s port for connecting to network. – AndrejaKo Jun 9 '16 at 16:11
14

Remember 10Mbps came first, then 100Mbps, then 1000Mbps.

The advantage of supporting multiple speeds and automatically switching between them is you can upgrade your network gradually without having to worry about what speed each device supports or replacing everything at once. You just plug a device in and it connects at the highest mutually supported speed. There are a few devices that are single speed but they are uncommon enough that it's not usually too much of a problem (though you do come across the occasional person wondering why thier Raspberry Pi won't talk to their media converter).

I think this is a large part of the reason twisted pair Ethernet has remained the dominant networking technology. It has been able to massively increase performance while keeping (almost) everything compatible.

Unfortunately with 10G the waters have got a lot muddier, afaict no devices support both 10M and 10G and some older ones don't even support 100M. I imagine the situation will be similar with the new 2.5G and 5G stuff. Fortunately newer 10G devices tend to support 100M and most 10M devices are probablly out of serivce by now so hopefully the breakage shouldn't be too painful.

  • I haven't see a 10 Mb network card in a very long time. I haven't even seen a 10/100 NIC in maybe ten years (although I still see things like firewalls with 10/100 interfaces). I would almost expect trouble finding drivers for a 10 Mb only card. – Todd Wilcox Jun 9 '16 at 4:18
  • 4
    Woah there, remember 1Mbit came first! Preferably on coax-piercing transducers... Honestly, you kids these days with your pac-mans, your 10mbit/s, your hula-hoops and hippedy-hop music... – John U Jun 9 '16 at 11:12
  • 2
    @ToddWilcox O'Rly? Have you seen network "shields" for embedded CPUs, like Arduino? The cheapest ones use bit-banging and those tiny CPUs can barely made it to 10Mb. like open-electronics.org/low-cost-ethernet-shield-with-enc28j60 There is plenty of 10Mb in IoT. – Agent_L Jun 9 '16 at 11:56
  • 2
    @JohnU AIUI early experimental coax Ethernet was 3Mbps and starlan (an early vendor-specific attempt at twisted pair ethernet) was 1Mbps, but everything that made it into an actual standard was 10Mbps or higher. – Peter Green Jun 9 '16 at 12:34
  • 2
    @Agent_L I doubt anyone is bit-banging Ethernet, Even at 10 Mbps that would be a very tall order. They are using 10 megabit only controllers though (notablly Microchip's ENC28j60 ) – Peter Green Jun 9 '16 at 12:39
0

It's all obsolete, in the current era. Which means whatever you are reading is obsolete.

Anything less than 1000Mb/s is ancient history. Fortunately for folks with antique equipment still running, myself included, most 1000 Mb/s ports will autonegotiate down to whatever the obsolete equipment is running at.

The point of autonegotiating is being able to speak to an obsolete device at a speed it can handle, without having to specifically configure a port to talk that slowly.

  • 5
    ...unless you are running a cluster of Raspberry Pis. They are limited to 100 Mb/s. – Nathan Osman Jun 9 '16 at 5:34
  • 6
    There is still a sizeable price gap between full-Gig enteprise switches and those providing FastEthernet ports (at least for low-end branch-office style switches for connecting PCs and phones). For many corporate environments, there really is no reason to have Gigabit to the desktop, and FastEthernet is more than enough, despite it's antique status. Egoistically I would prefer Gig to avoid the usual nightmares of duplex mismatches and the like, but unless the suppliers just stop making FE switches, we're probably going to keep buying them for a while yet. – Jeremy Gibbons Jun 9 '16 at 5:45
  • 1
    I work in a very budget-sensitive environment (school) and have not bought a Fast-E device in over a decade...There are a few still running in less-critical roles, but when they are replaced, it will be with Gig devices. If RasPis are 100Mb devices they are "newly manufactured as obsolete" which goes with them being cheap, but does not make 100Mb non-obsolete, IMHO. – Ecnerwal Jun 9 '16 at 11:02
  • 5
    Most management ports (on enterprise devices) are 10/100. iLO and DRAC on servers, most other IPMI-type interfaces. So, your "obsolete" characterization" is wrong. If you haven't bought anything with a 10/100 interface, you're not playing in enterprise IT, at least in the server room. – mfinni Jun 9 '16 at 12:40
  • 1
    I would agree that 10 mbps is essentially gone, but fast ethernet is not. – Ted Quanstrom Jun 9 '16 at 15:24
0

10/100 autosensing is for backwards compatibility with old hosts that are 10M only.

10M was the only ethernet for a long time. When 100M was introduced, it was extremely expensive and only used for backbones or servers (and not all servers either, just the important one or two that could justify the cost). Therefore it was assumed that 10M devices would be around for a long, long time.

Interestingly enough: 10/100M NICs were so expensive that using one to talk to a 10M device was initially considered wasteful. Why spend all that money to talk to a 10M device unless that 10M device was scheduled to be upgraded. If it was decided NOT to upgrade

Why was 100M either so expensive initially? Besides the fact that new tech is usually more expensive at first, at the time the only other network technology that was 100M/s was FDDI which was extremely expensive and extremely difficult to manage. So, the 100M/s ethernet vendors could charge high prices and it would still be cheaper than FDDI. If bread cost $100/loaf and you found a way to make it for $10/loaf, you'd charge $99/loaf for as long as you could, right?

I remember a Cisco salesperson in 1996-ish saying "we expect 10/100 switch ports to drop to $100/port eventually, so in the future we won't offer 10M... just 10/100M." At the time this seemed like crazytalk.

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.