[kwlug-disc] Reliable Broadband speed test
unsolicited at swiz.ca
Thu Mar 3 22:47:14 EST 2011
Raul Suarez wrote, On 03/03/2011 9:34 PM:
> I used to believe that all broadband speed tests were created equal.
No, depends upon time of day, amount of traffic, number of hops
between, etc., etc.
Since you are talking broadband, not ADSL, I will assume you are on
As you no doubt know, the only speed you can reliably test is between
you and your next hop (ISP). Heck, even that isn't reliable as other
internal equipment may be getting their fingers in the mix. So even
then the only reliable test is with all other equipment in the house
disconnected. (An advantage, therefore, to routers that have speed
test capabilities within them, to take this variance out of the equation.)
For your purposes, assuming Roger's, use Roger's own speed test site:
Even then, who knows whether this is even a reliable benchmark of
perceived speed. e.g. Roger's may unfilter, or even boost, traffic to
that site, for all I know. So results show fast speeds, but actual
browsing gets perceived differently.
> A few minutes ago, it felt that my Internet connection was getting slower, so I
> went to one of my trusted broadband test sites and it showed a very slow result,
> then I went to another, and showed a surprisingly fast result, then yet another,
> and showed a plausible speed.
> Is there a more reliable way to test broadband speed?
> http://reviews.cnet.com/internet-speed-test/ First 2.3 Mbps, then 20 Mbps,
> then 10 Mbps
> http://www.whatismyip.com/tools/internet-speed-test.asp 10 Mbps, then 20 Mbps,
> then let it rest and 10 Mbps again
> http://speedtest.primus.ca/ Between 10 and 20 Mbps.
> Could it be that Rogers is cheating and identifies traffic to test sites to
> burst the transmission?
> After the checks the performance improved. but is not constant, There seems a
> lot of stuttering (fast, slow, fast) even while loading the same page.
> So in the end, browsing feels sluggish even if the speed test says 20 Mbps.
AFAIK, Roger's filters and limits all encrypted traffic - so an https
page will be slower than an http page.
And there's no way to know I know of to know that Roger's -doesn't-
play with speeds depending upon where you're trying to go.
Add in time of day, moment to moment, congestion, and all the other
variances, I suspect all one can know is: (a) is my connection speed
to my ISP not diminished by anything on my end; (b) is it fast enough
for me. And the only thing one can do if it isn't is change providers
- and there's no guarantee that anything will be any better when you do.
Sad, but AFAIK, ce la vie. Particularly as all ISP speed promises are
phrased as '... up to ...'.
In the end, AFAIK, all one can do is confirm that the promised speeds,
to the provider's own speed test site, if they even have one, are
Having said all this, I have had success in the past, if you can pry
out of the ISP what benchmarks -they- use, with getting problems
addressed. In that case it was execulink, and probably only because I
was talking to their excellent (golden) support people on Victoria -
and I don't in general often get to such non-execulink support people
that both know what they're doing, and can directly measure anything.
e.g. In the past I've gotten to such knowledgeable support people
elsewhere, but they had to submit a request for a speed test not just
run it - so their ability to detect or do anything is stupidly limited.
Execulink, according to my bookmarks, is
YMMV & GL
More information about the kwlug-disc