Because I don't like Comcast enough to give them money for nothing
October 3, 2011 6:41 PM   Subscribe

So Comcast came by and installed their super-fast internet (up to 35mpbs), but in real use, it doesn't seem any faster than my old U-Verse (6mpbs). And sometimes not even as fast.

I can see from the various speedtest results that I'm getting the speed they claim, but my pages don't seem to load any faster, and in fact when I open 6 or 8 at a time, it frequently seems slower overall. Is it the case that by the time you factor in server speeds, DNS, and <other stuff that I don't know about>, that there's really no effective difference for casual use, or is my prejudice against Comcast just skewing my perception?

If it makes any difference, I'm running Windows XP, and use Chrome browser. I do realize that for big downloads, the speed would make a difference, but I almost never do big downloads.

I'm willing to pay Comcast more than I would AT+T, but only if their service is appreciably better. I've still got AT+T hooked up, so if there are some tests anyone could suggest I try, or good ways to compare, that would be helpful as well.
posted by still_wears_a_hat to Computers & Internet (15 answers total) 8 users marked this as a favorite
 
Do you own your cable modem and is it an older one? We got a letter from them today telling us that our 10+ year-old motorola was inadequate.
posted by ghharr at 6:43 PM on October 3, 2011


Response by poster: ghharr, no, it's their newest modem, at least according to the installer.
posted by still_wears_a_hat at 6:59 PM on October 3, 2011


Best answer: For browsing the web, bandwidth is not the big issue once you have broadband, at least in my experience. For a IIRC statistic, I have heard that web designers try to set up their sites so that they load within 4 seconds on a budget broadband connection. Bandwidth is really more useful for downloading large files, or streaming things like video. If your speedtests are coming out at the right speed, there's really not much evidence of Comcast being at fault.

Also, nthing alternative DNS servers. Google's are very good, and they are easy to remember if your computer somehow loses them and you need to get online and auto-DCHP is being a pain. OpenDNS offers some fine-tuning settings which are useful, too. IIRC, Chrome can be set to prefetch DNS queries, too, which speeds up loading the next page.

I think you should consider switching back if you don't do bandwidth intensive things like torrents, streaming videos or other large downloads, since that's what higher bandwidth gets you.
posted by mccarty.tim at 7:00 PM on October 3, 2011


Remember, you can't download data any faster than the remote server is willing to send it. Few (as in pretty much no) servers are willing to send data to you at those kinds of speeds.
posted by Chocolate Pickle at 7:09 PM on October 3, 2011 [5 favorites]


35mpbs is a measure of bandwidth.

bandwidth="how much stuff comes down the pipe in a given amount of time"

latency="how quickly does that stuff reach me"

In terms of perceived speed for typical Web browsing, latency is much more important. There are various ways of measuring it, but the "ping" utility (available on Windows, Mac OS X and any UNIX system) is probably the easiest.

+1 on what the others said about DNS, as your computer cannot fetch anything from a site till it knows that site's IP address.
posted by tomwheeler at 7:27 PM on October 3, 2011 [1 favorite]


Here is an analogy for my previous explanation of bandwidth and latency. Having the ability to carry more stuff in a single trip from Point A to Point B doesn't mean you'll get there any faster.
posted by tomwheeler at 7:30 PM on October 3, 2011


As others have said, bandwidth is a measure of how fat the Internet Tube running into your house is, latency is a measure of how quickly the Internet Juices are flowing (more accurately, from how far away they're coming).

Speed tests are designed to use the entirety of your bandwidth. They have to, otherwise they wouldn't be able to test it. They're not faking it.

If you have a speed test that claims you're getting 35Mb/s, you are, and comcast is holding up their end of the bargain.

Are you going to see the difference between that and 5-10Mb/s while you're browsing the internet? Yeah, maybe a little on places like youtube, but on a standard HTML page that's just full of text and images? No, doubtful.

Gaming, streaming audio and video, and downloading (larger) files will all be (positively) affected by your increased bandwidth. So, too, will doing any of those things at the same time as you are using the internet for something else (gaming while you stream movies, for example).

So, yes, it's entirely possible that you are getting the 35Mb/s and that you'll never notice the difference.
posted by toomuchpete at 7:44 PM on October 3, 2011


Chrome is a modern browser, but if you have a slow computer it could take a while to convert the HTML et al. into what you see on screen. You can tell the difference between a top-of-the-line processor and something in a midrange desktop, and definitely a netbook-style processor. Borrow a friend's laptop.
posted by wnissen at 7:50 PM on October 3, 2011


I can see from the various speedtest results that I'm getting the speed they claim, but my pages don't seem to load any faster, and in fact when I open 6 or 8 at a time, it frequently seems slower overall.

An XP era machine (pentium? single core? slow core2duo?) is going to have issues rendering 8 pages at the same time. You're probably CPU bound. Or if your machine is especially beefy, you're network latency bound or simply at the point where Chrome isn't going to show you a human measureable difference. I have a fast internet connection and a brand new i5 processor with an SSD and web surfing just tops out at a certain point.


The web has all sorts of architectual issues that may never be resolved like the ancient http protocol or the overly handshakey SSL negotiation. Its also full of practices like page rendering hampered by embedded objects or remote javascript. If google ads or google analytics runs slow, you'll notice it on any site using those services, unless they're coded in a way which loads them last.

The real benefit of fast internet speeds isn't browsing, but large file exchange, streaming HD, etc. If you just want a fast web connection, switch back to DSL.
posted by damn dirty ape at 7:58 PM on October 3, 2011


Remember, you can't download data any faster than the remote server is willing to send it. Few (as in pretty much no) servers are willing to send data to you at those kinds of speeds.

The first sentence is true but the second is not. I work for a company that delivers a large amount of web traffic (enough that our servers alone count as more than "pretty much no") and I can tell you that we will send you data at those speeds when we can.

I regularly pull from the internet at 7+ megabytes/second, and approximately 2 megabytes/sec at home.
posted by tylerkaraszewski at 8:30 PM on October 3, 2011


Seconding episodic here, especially in regard to Comcast's speed test. Had some problems with my modem, did a speed test via speedtest.net and was getting 20-30% less speed than what Comcast's speed test was telling me.
posted by kuanes at 4:49 AM on October 4, 2011


How many days ago did you have your new Internet installed?
posted by dgeiser13 at 5:53 AM on October 4, 2011


Look at the internet like the postal service. How long does it take for the sender to write the letter and drop the mail in the mailbox? How long does the post office take to deliver it? How big can the package be? Once you get it, how long does it take you to actually open the mail once it's at your house?

Your ISP is only your local letter carrier. And based on the speedtests, he is doing his job.

So, back to technical-land. If you are getting correct speedtests, then you know the infrastrucure is right. What I have found to be the cause of slow page loading is websites that use affiliates for various things. When you click on a site, your computer gets the html file and starts rendering it. As it is rendering it, it sees calls to other websites, and has to go to those websites (ad servers, google analytics, the flickr photostream, etc.). All of that adds delay, especially if the html is written in a way that doesn't let anything else finish until it gets the answer it is looking for.

Further, there are rendering issues within the browser. How long does it actually take for the browser to draw the pictures on the screen?

And finally, as others above said, you can only get data as fast as the originating site can send it. If the webserver has a gigabit connection, but is serving 2000 simultaneous users, then you are only going to get 500 kilobits.

> Remember, you can't download data any faster than the remote server is willing to send it. Few (as in pretty much no) servers are willing to send data to you at those kinds of speeds.

But tests like speedtest.net require at least 100 Mbps for any test host on their network.
posted by Horselover Phattie at 8:26 PM on October 3


And that is verified by the asker getting the advertised speed when they run the speedtest.
posted by gjc at 6:09 AM on October 4, 2011


Speedtest.net

It will tell you what your actual download speeds are.

Most websites don't serve up info as fast as you can download it unfortunately. But when you do something like downloading files from usenet or the like, you do. A speed test will set you straight.
posted by darkgroove at 7:22 AM on October 4, 2011


You've probably found your answer, but just in case . . .

Cable modem should be DOCSIS 3.0 and you should be using a GB ethernet card. (Your tech says that it's the latest. What's the make and model? Check it out yourself; cable modems are cheap.)

After that, it's the 'net that's slow. Do what you can inside the perimeter.
posted by Man with Lantern at 9:40 AM on October 4, 2011


« Older Songs only a historian would love?   |   Help me have a nice day trip with my girlfriend in... Newer »
This thread is closed to new comments.