- Joined
- Nov 21, 2020
That should never happen. The cards may be crappy but the actual link makes no difference. Check the error counters, if you're seeing any hardware level errors(CRC) then there's too many.I have some short runs where copper can only get 7~8Gb and the fiber does a solid 9.5Gb in iperf.
Here's my post from the hardware thread.
To interrupt the benchmark slapfight with some benchmarks.
I finally, well, almost, have all my "PC" shaped boxes running on 10G Ethernet. I have an ITX system and tried out the new Realtek 8159 USB. My gaming PC I "upgraded" from an AQC card that dropped every hour to a Realtek 8127. Here are the benchmarks.
iperf3 -c HOSTNAME -t 20 plus the options listed, -R to reverse the flow, -P2 to try 2 streams. All tests run from the named device to a server with 2x 10G Broadcom in a LACP.
I finally, well, almost, have all my "PC" shaped boxes running on 10G Ethernet. I have an ITX system and tried out the new Realtek 8159 USB. My gaming PC I "upgraded" from an AQC card that dropped every hour to a Realtek 8127. Here are the benchmarks.
iperf3 -c HOSTNAME -t 20 plus the options listed, -R to reverse the flow, -P2 to try 2 streams. All tests run from the named device to a server with 2x 10G Broadcom in a LACP.
Code:
[ ID] Interval Transfer Bitrate Retr
USB RTL 8159 on a USB 3.2 x2 connection.
[ 5]...
Reactions:
x 2
The Realtek suffer with single stream, the AQC has a glitch on dual stream, maybe flow control, but they'll all hit line rate solidly. The Intel is fiber connected. Even the shitty USB-C can do line rate on a single stream.