I noticed some time ago that the connection lights on the 24 port Gb switch in my office showed that one of my computers was not connected at the full speed of 1GB (1000Mb) per second. It took me a while to get around to resolving this but when I did, I found an interesting reason for this.
The hardware on the host is an on the motherboard Intel e1000e NIC.
When I finally got to it, I discovered using ethtool that the NIC was connected at 100Mb/sec and that there were no errors such as dropped packets at that speed. The /var/log/messages log file and the dmesg command both showed the following messages.
Feb 18 13:33:43 host1 kernel: e1000e: enp0s25 NIC Link is Up 100 Mbps Full Duplex, Flow Control: Rx/Tx
Feb 18 13:33:43 host1 kernel: e1000e 0000:00:19.0 enp0s25: Link Speed was downgraded by SmartSpeed
Feb 18 13:33:43 host1 kernel: e1000e 0000:00:19.0 enp0s25: 10/100 speed: disabling TSO
A Google search indicated that this type of problem can be the result of problems with hardware. The most common hardware problems for networks are defective cables. I switched to a new Ethernet cable between the host and the switch and the link came up at 1000Mb/sec (1Gb) and is now working fine.
After getting this working, I decided to test the cable. The standard cable tester I used showed that the cable was not defective in that all of the connections tested fine in a static continuity test. This is a short CAT5 cable and I typically am able to successfully use CAT5 cables for Gb networks with no problems so long as the cable length is less than 20 meters or so. In this case the cable was only about 2 meters.
I can only suspect that the cable was not manufactured to proper CAT5 standards and was therefore failing when used in an active, real world environment.