hi,
i just switched over to a cable modem type isp, and it's my understanding that my 6mbps is not necessarily allocated to me, but is available when i use it, and that network traffic can also slow down my transfer speed as a result.

it's also my understanding that as a result of the shared bandwidth when i use a site that checks your internet speed, it can be inaccurate, and show a higher rate than what you're actually getting. this hasn't been the case for me in the past, but with current provider that was hooked up yesterday, according to speedtest.net, i'm getting 22mbps, which is crazy fast, and i don't see how it would be possible, but at the same time my browser is pretty sluggish, and doesn't feel like even what i had b4 which was 5mbps (current is supposed to be 6).

How can i test for accurate dl/ul speeds with cable internet?
thanks,
rlr

Yes, the bandwidth is shared in Cable. But you shouldn't need to worry about it much as you'll usually get most/all of the speed you paid for.

That is pretty fast. Perhaps the site or your ISP did some caching and that threw your results off?

You can try some other speed testing sites to see how the results vary.

http://www.google.com/search?q=related%3Awww.speedtest.net

As far as actual speed vs garauntee speed, it can only be truly tested within the ISPs network. Normally this is tested by uploading a file and downloading a file via FTP to another computer. This other computer must be accessable (see traceroute) without leaving the ISP's network. If you use CLI FTP, turning on bin and hash, it will give you the speed. Files should also be large, like 5MB or so. Online speed test sites use very small files to do the tests, and they are not accurate this way. You need to know your average speed, which requires a larger file. You will notice if you download something, it starts out pretty fast and then slows down to a certain speed. This is why small files should never be used for speed tests.

Actually, the only shared part of cable (the DOCSIS kind, like your ISP is using) is the upstream. But in periods of heavy congestion (anywhere on the network), chances are you'll get less.


No, exactly the inverse. With heavy congestion, speeds may go down but sharing any part of the network won't ever make something appear as faster than it really is. Either ways, the speed reported is what you're actually getting *from that particular server*, and that is perfectly accurate (just not a measure of what you were exactly expecting)


With every cable ISP I've been, I've been getting at least the advertised speed. Being over isn't that surprising. Also, your particular ISP uses SpeedBoost which makes the beginning part of every transfer faster (like doubling your speed for the first few seconds -- making every transfer much faster, save for really large files e.g. over 100MB)


And that often has little to do with your max speed, unless you're on a page with a lot of large files and that the server at the other end is fast (e.g. youtube videos may download lots quicker). Web pages these days include a lot of other files (tons of images, cascading style sheets, external javascripts, favicons and so on), and for every single file you have to make another HTTP request. The network latency just adds up. Try ping'ing a popular server, see how are the round-trip times. Max download speed and network latency are unrelated (much like max throughput and seek times on a hard drive), and varies quite a bit. In fact, just changing which DNS servers you're using may make this a fair bit quicker as most pages have files across different domains (and sub-domains too), like say ads and visitor statistics stuff, and to make HTTP requests to load each of those you have to make a DNS lookup first (latency adds adds up quickly yet again).


There is no way to "accurately" measure any internet connection. Using your ISP's test server (like Tripredacus mentioned) is a way to make sure your test won't return a lower speed because of network congestion elsewhere on the internet. But 99% of internet speed tests sites have been ridiculously inaccurate for me. Like speedtest.net, that used to say I had like 700KB/sec speeds, whereas I could easily download from say, Microsoft's website at 1250KB/sec the very second after that test completed, or a relative with a "basic" cable plan, that said he could get 70KB/sec, yet could get 125KB/sec steady downloading from any website anytime we wanted...