3

I am trying to install a 1000Mbit segment that is about 285ft long. It works perfectly fine at 100Mbits, but it won't do 1000Mbit. The testing devices (2 computers, 1 switch) have 1000Mbit ports. All ports were verified at 1000Mbits using a 175ft test cable.

I have been using iPerf3 to test the connection speeds. The 175ft test cable averages 945Mbits/sec and the 285ft cable averages only 97Mbit/sec. Both cables are cat 5e made from the same spool.

It appears that the 285ft segment cannot detect the 1000Mbit ports between the two computers, so it drops down to 100Mbits, and it clearly handles 100Mbit without any problems.

It's my understanding that 10\100\1000 all support segments up to 100meters, thus if a segment works fine at 100Mbits, why would it not work at 1000Mbits?

6
  • Have you checked all four twisted pairs for continuity, shortages and correct pairing?
    – Zac67
    Mar 11, 2022 at 11:33
  • I created the 285ft segment myself and used an inexpensive Ethernet cable tester that matches the 8 individual wires on both ends (ie straight through cable). If there is failed connection, or the wires are crossed, the tester indicates those problems. It does not do anything more than that. I was using iPerf to test the transfer rate, which was pretty good with 175ft cable at 945Mbit\sec
    – nrnoble
    Mar 11, 2022 at 12:05
  • 1000 uses a lower signal voltage on all four pairs. If the line loss is too great, it obviously won't work. This is a common failing of desktop NICs, and many cheaper switches.
    – Ricky
    Mar 11, 2022 at 12:21
  • 1
    @nrnoble Cheap cable testers only test continuity and shorts. They can not check correct pairing which is essential for anything beyond very few meters.
    – Zac67
    Mar 11, 2022 at 13:11
  • 1
    Did you use solid-core cable and follow all the requirements of maximum pull tension and minimum bend radius, etc.? Did you perform all the required tests in this answer and they all passed?
    – Ron Maupin
    Mar 11, 2022 at 17:30

1 Answer 1

2

In contrast to 100BASE-TX, 1000BASE-T requires all four twisted pairs to work. As with all variants, the used pairs need to be matched correctly - swapping conductors between pairs significantly increases noise/crosstalk and makes the link fail for anything beyond a few meters.

Many gigabit devices automatically fall back to 100 Mbit/s when a pair fails to work (Broadcom's "Ethernet@Wirespeed"), but that is not a given - some devices simply fail to link.

In addition to a simple discontinuity, a split pair will also make the link fail for any longer run. When Auto Negotiation has established mutual 1000BASE-T support, the PCS sublayer tries to link each lane (=twisted pair) separately, so that any pair swaps are detected. Any split pair causes severe crosstalk, causing the involved pairs fail to complete the lane-level link. Accordingly, you must take care not to swap any wire between pairs. And of course, changing the twist rate or reversing the twist direction can also cause significant crosstalk or attenuation.

Also, for completeness: the nominal 100 m range of twisted-pair Ethernet the cabling requires 90 m of rigid solid-core cable and a maximum of 10 m flexible patch cable. Using extended lengths of patch cable - with a much higher attenuation per length - can make the connection unreliable or fail outright.

100BASE-TX is more tolerant to those cable faults than 1000BASE-T, so the former might work (barely or even stably) while the latter fails outright.

Your Answer

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

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