Questions & Answers:
Actually, this isn't just one Web page, it's two Web pages, each with a hidden data file in the header area of the page. When either of these pages loads, it notes the time just before and just after the data file loads, then uses this information and the size of the data file to calculate the rate at which the data arrived at your computer.
When the first of these two pages was loaded, it did a rough calculation using a small data file to provide an estimate of how long it will take the second page, with its much larger data file, to download. The second page, with its larger data file, allowed the second page to calculate more accurate figures for the Your Line Speed: box and Speed Test Thermometer above.
Like any major highway system, the Internet information highway has many roadways and interchanges, each with their own capacity and speed limit, and, like highways for cars, sometimes you get traffic delays. Just as you have to wait in a your car while other traffic goes through at a traffic light, data sent to your computer has to wait while other data passes through routers, the Internet equivalent of an intersection, on it's way to you.
You are more likely to run into slow Internet traffic during peak use hours than those times when fewer people are online, such as after 11 p.m. and before 7 a.m. When this page was being tested, I found that the results on my 144Kbps DSL line results varied from a consistent 120 to 135Kbps mid-morning and afternoon, all the way down to 64Kbps during evening peak-use hours. If you really want to find out what your best possible Bandwidth Speed Test result is, try loading this page at 3:00am on a weeknight, when almost everyone is asleep when Internet use is at it's lowest. (Friday and Saturday nights are not a good choice.)
No Internet connection ever performs at 100%. When we test loaded this page direct from our test Web server across a two computer 10Mbps local area network (LAN), our best result was 8.2Mbps (8160Kbp). There is always some loss, and the faster your connection is, the greater that loss is going to be.
Next, this page can only measure the time it takes the applications data--the actual data file--to reach your browser. Wrapped around that data is some overhead that can range from 2 percent to 25 percent of the total data sent. There's no way for the program built into this Web page to control or discover exactly how much overhead was used to send the page's internal data file, but generally the percentage is small, and this page adjusts its figures up by 2 percent to compensate. Nevertheless, the actual overhead may be higher than expected, resulting in a depressed value for the Kbps figure.
Also remember that your connection to your ISP is just one part of the system that gets data to your computer. Between your computer and the server that sent this Web page there are probably a dozen or more routers, communications links, and other network components this page had to travel through. Each of these components have a set capacity and speed at which they can operate, and most of them handle network traffic for thousands, even millions of computers every day. The inevitable result is that all Internet traffic encounters some sort of delay as it transits across The Net, and that reduces the amount of data that gets funneled into your connection in the first place.
The delay could be at your ISP, or it could be elsewhere. (See the explanation above, but if you're consistently getting Bandwidth Speed Test results that are substantially below expectations, the root problem is most likely your ISP's fault.
While your dial-up modem, DSL, ISDN or other Internet connection may be a dedicated line, all of an ISP's connections get combined into one or more shared connections. In most cases, these shared connections have less capacity than the combined total of all the customer connections they serve. Done judiciously, this works better than you probably think. Since most Internet users spend more time reading their email and Web pages than they do downloading them, they're only using a fraction of their connection's actual capacity. Overbooking allows an ISP to combine several customer connections into a single link that's smaller (and less expensive) than the combined total of all the connections they serve, without reducing the amount of data sent to a customer when they are downloading data.
The problem is that some ISPs, cable modem companies, and DSL providers take the overbooking concept too far. They funnel so many connections into a small combined connection that normal customer demand overwhelms the capacity of the combined connection. This is a particular problem during peak use hours, when line speeds can slow to a crawl.
Unfortunately, there is no remedy for this problem. As competition in the high-bandwidth Internet connection business heats up, you'll have more options and your ISP will have more incentive to maintain more reasonable overbooking ratios. Until that time, however, your only options are to complain to your ISP or switch to another Internet provider with a better track record.
There could be several reasons--and most of them aren't your ISP's fault.
First of all, static electricity caused by radio signals, power lines, and other sources interfere with most 56K modem signals, forcing them to fall back to 42-50Kbps.
56K modems also require a clean, straight through telephone connection to the telephone company's central office switching center. Phone company line amplifiers that boost a telephone signal over a long distance, PBX switchboard systems, and other phone equipment alter the phone signal and force 56K modems to fall back to speeds of 33.6Kbps and lower.
Finally, the FCC doesn't allow 56K modems to use the full range of signals that phone company equipment can generate. They're concerned that it'll cause static interference to other phone lines.
So no 56K modem in the United States ever connects at 56K. Most 56K modem users seem to connect at speeds of 44-48Kbps.
If you're just dying to see the raw data used to calculate your download speed, here it is: