Here are consecutive results from speedtest and iperf3. I am well aware that the iperf3 server is significantly further away. I have run speedtests to a server in the same Fremont, CA data center and seen results consistent with the Phoenix results, I just couldn't find the server ID to enter in Speedtest-cli for purposes of this comparison. I'll also now that today's iperf results are the best I've seen yet.
taylor in taylor-kbox in ~
❯ iperf3 -c iperf.scottlinux.com -4
Connecting to host iperf.scottlinux.com, port 5201
[ 5] local 192.168.0.153 port 32972 connected to 45.33.39.39 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 2.75 MBytes 23.0 Mbits/sec 12 52.3 KBytes
[ 5] 1.00-2.00 sec 730 KBytes 5.98 Mbits/sec 8 17.0 KBytes
[ 5] 2.00-3.00 sec 730 KBytes 5.98 Mbits/sec 2 26.9 KBytes
[ 5] 3.00-4.00 sec 730 KBytes 5.98 Mbits/sec 8 11.3 KBytes
[ 5] 4.00-5.00 sec 365 KBytes 2.99 Mbits/sec 5 11.3 KBytes
[ 5] 5.00-6.00 sec 365 KBytes 2.99 Mbits/sec 6 11.3 KBytes
[ 5] 6.00-7.00 sec 365 KBytes 2.99 Mbits/sec 6 11.3 KBytes
[ 5] 7.00-8.00 sec 730 KBytes 5.98 Mbits/sec 3 15.6 KBytes
[ 5] 8.00-9.00 sec 365 KBytes 2.99 Mbits/sec 5 18.4 KBytes
[ 5] 9.00-10.00 sec 365 KBytes 2.99 Mbits/sec 8 9.90 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 7.38 MBytes 6.19 Mbits/sec 63 sender
[ 5] 0.00-10.03 sec 6.64 MBytes 5.56 Mbits/sec receiver
iperf Done.
taylor in taylor-kbox in ~ took 10s
❯ speedtest
Retrieving speedtest.net configuration...
Testing from Cox Communications ([IP REDACTED])...
Retrieving speedtest.net server list...
Selecting best server based on ping...
Hosted by Cox - Phoenix (Phoenix, AZ) [DISTANCE REDACTED]
Testing download speed................................................................................
Download: 164.75 Mbit/s
Testing upload speed......................................................................................................
Upload: 10.63 Mbit/s
they do not artificially favor any speed test. however, the speed test they encourage (the one on the website) is to their own servers, which is one of the most direct node pathings. it is a bit disingenuous on their part.
Go to speedtest or ookla or dslreports and set the server manually to literally anything that isn't a Cox hosted server, and you will see the real world slowdown by distance and node hops.
"The problem is that these speeds are not holding up under sustained real-world loads, particularly on the Upstream side. I operate a Plex Media Server in my home and routinely stream from it throughout the day. In the past I was able to reliably stream at 8 Mbps, but in the last couple of months I have had severe buffering issues and had to bump down to 4 Mbps or less to get a stable stream. I have backed this up with a series of iperf3 tests, all of which show an initial transfer speed of 10 Mbps, which almost immediately drops to 4-4.5 Mbps after about a second of testing. Regular large transfers to other cloud storage services tell a similar story."
This sounds like your networking hardware is overheating, or more likely, a line in the ground got shifted somewhere and it cut one of the up lines. If you are running a media server then overheating is very likely, however.
Thanks for the suggestions. I have solid hardware in the chain, and I routinely monitored thermals and cpu usage on both my Media Server PC and my Router throughout my testing process, and neither were particularly taxed by the workloads or had any thermal issues to speak of..
As I mentioned in my follow up post a moment ago, I am considering this solved as I was able to achieve proper upload speeds to a local Cox hotspot, validating the performance of both my home network and the local Cox infrastructure.
