Hacker News new | past | comments | ask | show | jobs | submit login

Yes, and when a US cable company advertises 25Mbps, they mean they will give you 25Mbps for 5 minutes which they take from the whole neighborhoods' 50Mbps, and then throttle you down after that to god knows what.

If it's not fiber, I assume there's tons of overhead/lying with the numbers.




I pay for 200/10 Mbps and out of hundreds of tests, I have never gotten less than 235/11 and usually get 260/12 on hybrid cable (fiber backbone, DOCSIS 3.0 32x8 local loop) from Comcast.


Are these tests akin to speedtest.net for a few minutes, or iperf over wireguard for a few hours?


Those are speedtest.net, but being careful to choose a test endpoint not under Comcast's control.

I just did one again and got 323/11. https://www.speedtest.net/result/11034876082

Assuming the setup time on my end is reasonable, I'm happy to run other tests for the thread's benefit.


Speedtest.net isn't a great a measurement because cable is notoriously bursty. Cable ISPs will give you the full (neighborhood) pipe for a minute, and then throttle you back to some set speed. This is helpful to speed up web browsing, but not when measuring bandwidth.

I'd recommend timing any big upload that takes at least an hour. Ideally encrypted and somewhat unique, so that it's not likely to be prioritized or served by an edge box.


I wasn't willing to go for an hour, but I planned an upload of one that would take about 15 minutes (to get well past any initial burst rate). The upload rate started at 1.3 MiB/s, quickly climbed to 1.4 MiB/s and stayed there the entire upload.

This is a shot on my personal iPhone family video, so there's zero chance that it is cached anywhere.

If you compute using a file byte as being 8 network bits (the reality is slightly higher, of course), you get:

  jsokoloff@Jim-Macbook Downloads % time aws s3 cp iPhone_opening.MOV s3://XXX-XXX/photos/ 
  upload: ./iPhone_opening.MOV to s3://XXX-XXX/photos/iPhone_opening.MOV
  aws s3 cp iPhone_opening.MOV s3://XXX-XXX/photos/  12.84s user 7.66s system 2% cpu 14:38.47 total
  jsokoloff@Jim-Macbook Downloads % ls -la iPhone_opening.MOV 
  -rw-r--r--@ 1 jsokoloff  access_bpf  1265329653 Aug 22  2020 iPhone_opening.MOV
  jsokoloff@Jim-Macbook Downloads % echo -ne "bps: " && echo "1265329653 * 8 / (14 * 60 + 38.47)" | bc          
  bps: 11523031
  jsokoloff@Jim-Macbook Downloads % echo -ne "Mibps/100: " && echo "1265329653 * 8 / (14 * 60 + 38.47) * 100 / 2 ^ 20" | bc 
  Mibps/100: 1098
So, 10.98 Mibps (11.52 Mbps) for an almost 15 minute period, which matches the above extremely closely.


That's a much better test and thanks for doing it. I usually use a factor of ~10 (9.5) to go from advertised Mb/s to actual MB/s, so I agree 1.4MB/s is decently better than the 10Mbps quoted.

Of course the ultimate test is hopping on a popular torrent ;)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: