Lewis
Member
- Aug 5, 2005
- 15,483
- 621
Find the Fastest ISP
By John Brandon
Broadband, shmroadband. Surfing the Web is still sometimes more like wading against the current. Images crawl up the screen, chunks of text appear . . . slllowwwly . . . at your ÂÂfavorite portal. What if Web browsing were much faster? What if the entire page popped up instantly? In this sceÂÂnario, you'd probably get your browsing done faster, visit more ÂÂpages, find time for friends and family, even leave your house on occasion.
Faster browsing is certainly possible, but there's a wide gap between theoretical broadband speedâ€â€the one your Internet service provider lists on your bill, such as 3 megabits per second (Mbps)â€â€and actual browsing speed, which might be more like 200 kilobits per second (Kbps). Two critical factors cause these dramatic slowdowns while browsing the Web. One is the fact that Web pages are made up of lots of little elementsâ€â€includÂÂing the underlying HTML code and text, plus all the images, embedded Flash, and other contentâ€â€and downloading in this way is just less efficient than grabbing a single object. But another big factor is that, despite claims that often sound alike, not all ISPs are equal: Some simply run faster than others.
The question is, which is the fastest? There's never been a bulletproof, scientific method to test actual surfing throughput. Sites such as BroadbandReports.com, PC Pitstop, and Testmy.net send one large file to and from your PC to test upload and download bandwidth. These can be good tests if you often transfer large files, but they don't do a good job of measuring the browsing experience, which is the dominant Internet activity for most of us.
To find out whether our suspicions about slow browsing speeds are true and to report which ISP provides the best surfing bandwidth, PC Magazine created SurfSpeed (go.pcmag.com/surfspeed), a free utility that tests actual page download speeds for ten of the most popular sites on the Web. From mid-May to mid-June, more than 10,000 readers downloaded the utility from PCMag.com. SurfSpeed uploads performance data to a central database, so users can easily compare their lines' performance to that of others in their area and to their ISP's performance in other areas. (The utility spells out what data is collected.)
A few ISPs are beginning to roll out fiber-to-the-premises services, promising much faster speeds than cable or DSL. Fiber optics transmit data using hair-thin glass tubes. Verizon's FiOS (which stands for Fiber Optic Services) is by far the largest fiber-based broadband service, but it's available only in select areas of 16 states. Still, we had more than 200 SurfSpeed users who report that they're already using fiber. At an average of 271 kilobits per second, fiber was 35 percent faster than cable and more than 60 percent faster than DSL. (By default, SurfSpeed reports results in kilobytes, which is equal to 8 kilobits, but the industry standard is to report in kilobits.)
Verizon was the only fiber service with sufficient responses for us to analyze its results, which, at 293 Kbps, were even faster than the ÂÂaverage fiber number. If you can get FiOS, you'd be crazy not to go for it. Verizon currently charges $44.95 a month for an advertised 15 Mbps download rateâ€â€about the same as you're probably paying for cable and perhaps slightly more than DSL. Quadruple that price to jump up to a 30 Mbps plan.
Even if you can't get fiber to your house, the closer the fiber comes to your residence, the faster your connection is likely to be, because you're connecting to the Internet backbone with fewer hand-offs. The New York/New Jersey/Connecticut area has a heavy concentration of fiber optic lines. Optimum Online, the broadband service from Cablevision, which serves many parts of the tristate region, averaged 235 Kbps on the SurfSpeed test, giving it a 12 percent speed advantage over Cox Cable, the second-place cable-modem finisher, and is 40 Kbps faster than AT&T Yahoo! (formerly SBC Yahoo!), the fastest DSL provider. These speeds have helped Optimum Online win PC Magazine's Readers' Choice award for ISP satisfaction in our annual survey for the past two years. (For more on our survey, see go.pcmag.com/sr.) On the other end of the scale, Qwest was, by far, the slowest ISP; its DSL service averaged only 109 Kbps on the SurfSpeed test.
Cable tends to be faster than DSL because cable operators optimize channel bandwidth frequencies over the coaxial cable, and because DSL has to live within the limitations of legacy telephone lines. It's worth noting, however, that DSL is typically cheaper than cable access, and according to a recent JupiterResearch report ÂÂreleased in June, it will become more popular than cable by 2010, thanks to lower prices.
Other Findings
SurfSpeed gathers data about the user's CPU, RAM (both total and available), and operating system. Although we suspected that faster CPUs and more RAM might affect download speed, they turned out to be nonfactors. In fact, the results are all over the map: SurfSpeed ÂÂreported faster access with slower processors in some cases, and higher speeds with less RAM. But don't take this to mean that you won't benefit from a faster PC. SurfSpeed focuses on how long it takes to grab bits off the Internet, not how long it takes to display those bits. A faster computer will render Web pages more quickly once the images and text arrive in your browser.
What's the best time to go a-browsin'? See the results chart. SurfSpeed time-stamps all its tests, so we used that data to see when you should set your alarm clock to get up and go online. Not surprisingly, performance does peak in the middle of the night, when fewer people in the U.S. are using their PCs. For cable customers, the best time was 1 a.m. eastern. DSL's best time was 5 a.m. eastern, but there was very little variability between 1 a.m. and 7 a.m. In fact, overall, there was only a 7 percent difference between DSL's slowest hourâ€â€3 p.m. eastern (lunchtime for the California crowd)â€â€and its fastest. Interestingly, cable bottomed out at 8 p.m. in New Yorkâ€â€but again, the difference wasn't great. The difference with fiber was much greater, but we'd like to see more data before drawing conclusions.
The broadband ISP business is still fairly regional. See how the ISPs are spread across the country. When we divided the country into four geographic regions, Comcastâ€â€by far the best-represented provider in our SurfSpeed dataâ€â€was the only ISP to show up in all four. It speeds slightly ahead of Adelphia and Cox in the West and runs slowest in the Northeast.
The Best ISP
So which ISP should you choose? Would that you really had options! The types of broadband you can get still depend on where you live. You probably have only one cable company to choose from and maybe a few DSL providers. Which technology should you throw your lot in with? If you can get it, the answer is clear: fiber. And in general, cable will run faster than DSL, though that's not always the caseâ€â€AT&T Yahoo! outpaced four large cable providers. How should you choose? Grab a copy of SurfSpeed at go.pcmag.com/surfspeed and get up-to-the-minute data about all the ISPs in your area.
Urban Stats
Fastest: San Francisco/Oakland/Fremont, averaging 227 Kbpsâ€â€11 percent faster than the New York/Northern New Jersey/Long Island metro area.
Slowest: Denver/Aurora, moseying along at only 150 Kbpsâ€â€the Bay Area is 51 percent faster.
Happiest: Minneapolis/St. Paul/Bloomington, despite barely passing Denver's poky connections, reported the highest satisfaction level. (It's fast enough to stream A Prairie Home Companion. What's to complain about?)
Most unreasonably grumpy: Despite near chart-topping speeds (203 Kbps), residents of Silicon Valley's own San Jose/Santa Clara/Sunnyvale area had more gripes than anyone except those contending with the St. Louis area's slow connections (164 Kbps).
How Surfspeed Tests
When you download large files, high bandwidth generally translates into faster file download speeds. But remember, a file download also relies on the site that's serving the file to you. If the site has a small pipe (limited bandwidth) or it's overloaded with users, you're unlikely to see the speed that your provider touts on its Web site.
With that in mind, Bruno Sonnino, the author of many PCMag utilities, developed SurfSpeed for us as a test of real surfing speed. It's a free download (registration required) from go.pcmag.com/utilities, and we urge you to give it a try. We hope to revisit this analysis in the future.
How SurfSpeed Works
Bandwidth testing sites download a single large file, initiating a single connection. Instead of downloading one large file over a single server (with lots of bandwidth), SurfSpeed grabs pages (and page elements) from multiple sites with varying amounts of bandwidth. In addition, just like a real Internet browser, SurfSpeed initiates multiple connections to get all of the page elements. Each connection takes time, and that becomes a part of the SurfSpeed equation.
SurfSpeed
SurfSpeed behaves more like a real browser than like a bandwidth test, and in doing so gives a unique measurement that more accurately reflects how you spend 90 percent of your time on the Web, whether you're surfing or sitting and waiting for a page to load. By default, the utility runs once per hour, "browsing" to top sites AOL, Apple, eBay, Go.com, Google, MapQuest, Microsoft, MSN, MySpace, and Yahoo!. You can run it on the sites of your choice, but we considered results only from these preset sites.
Once you run SurfSpeed, you can begin comparing your results with others. Each time you run a test, the results are sent to the PC Magazine servers. This data is then used to compare your speed with that of others in your ZIP code, state, country, and the world. You can also see how your ISP performs compared both with other ISPs and with itself from state to state.
SurfSpeed
How we got the results
For this story, we considered all the tests run from May 12, 2006â€â€the day after we unveiled SurfSpeedâ€â€to June 8, 2006. In all, we had data from more than 12,000 users in 126 countries, and 128,000 test runs. Since each test hits ten Web sites, we had over one million data points to consider. We restricted our analysis to users in the United States.
Anyone who has spent any time online knows that, no matter how fast your connection, occasionally sites won't be available; pages won't load properly; or downloads will take an inordinately long time to ÂÂarrive. So we worked with Datatechture, a New Jersey-based research company, to help us identify outliers and clean the data so we could do our analysis.
We considered only those ISPs and connection types for which we had at least 50 users. We further filtered out data from users who didn't produce enough test runs over the period that we analyzed.
Fiber Lust
Envy It's a rare day when people regard me with such, but ever since I got fiber-to-the premises (FTTP) installed, it's been a daily experience.
I've won the broadband lottery. No one else at PCMag's offices has FTTP (Verizon calls it FiOS), and scarcely a day passes that someone doesn't ask, "So, how is it?" The answer is always the same: It's great!
Fiber's most obvious benefit is speed: You basically get as much as you're willing to pay for. I bought the second-tier service, for roughly $45 a month, to get 15 Mbps. Interestingly, my cable company offers the same speed for the same price, but my actual cable broadband speeds were far slowerâ€â€45 minutes for a 1.7GB file download on cable and just 12 minutes for the same file, from the same source, through FiOS.
I'm also astounded at the amount of free equipment: a suitcase-size fiber termination box outside my house, a UPS (uninterruptible power supply), and an 802.11g wireless router inside came with the installation (if I quit FiOS, I pay). The UPS, by the way, is for my phone line, which now comes in through the home-powered fiber box. Without the UPS, if I lose power, I lose the phone, too.
I'm tempted to go to the top of the Empire State Building and declare my good luck to the world, or at least Manhattan. Of course that's bound to make those without access to fiber even angrier.
http://www.pcmag.com/article2/0,1759,1994566,00.asp
By John Brandon
Broadband, shmroadband. Surfing the Web is still sometimes more like wading against the current. Images crawl up the screen, chunks of text appear . . . slllowwwly . . . at your ÂÂfavorite portal. What if Web browsing were much faster? What if the entire page popped up instantly? In this sceÂÂnario, you'd probably get your browsing done faster, visit more ÂÂpages, find time for friends and family, even leave your house on occasion.
Faster browsing is certainly possible, but there's a wide gap between theoretical broadband speedâ€â€the one your Internet service provider lists on your bill, such as 3 megabits per second (Mbps)â€â€and actual browsing speed, which might be more like 200 kilobits per second (Kbps). Two critical factors cause these dramatic slowdowns while browsing the Web. One is the fact that Web pages are made up of lots of little elementsâ€â€includÂÂing the underlying HTML code and text, plus all the images, embedded Flash, and other contentâ€â€and downloading in this way is just less efficient than grabbing a single object. But another big factor is that, despite claims that often sound alike, not all ISPs are equal: Some simply run faster than others.
The question is, which is the fastest? There's never been a bulletproof, scientific method to test actual surfing throughput. Sites such as BroadbandReports.com, PC Pitstop, and Testmy.net send one large file to and from your PC to test upload and download bandwidth. These can be good tests if you often transfer large files, but they don't do a good job of measuring the browsing experience, which is the dominant Internet activity for most of us.
To find out whether our suspicions about slow browsing speeds are true and to report which ISP provides the best surfing bandwidth, PC Magazine created SurfSpeed (go.pcmag.com/surfspeed), a free utility that tests actual page download speeds for ten of the most popular sites on the Web. From mid-May to mid-June, more than 10,000 readers downloaded the utility from PCMag.com. SurfSpeed uploads performance data to a central database, so users can easily compare their lines' performance to that of others in their area and to their ISP's performance in other areas. (The utility spells out what data is collected.)
A few ISPs are beginning to roll out fiber-to-the-premises services, promising much faster speeds than cable or DSL. Fiber optics transmit data using hair-thin glass tubes. Verizon's FiOS (which stands for Fiber Optic Services) is by far the largest fiber-based broadband service, but it's available only in select areas of 16 states. Still, we had more than 200 SurfSpeed users who report that they're already using fiber. At an average of 271 kilobits per second, fiber was 35 percent faster than cable and more than 60 percent faster than DSL. (By default, SurfSpeed reports results in kilobytes, which is equal to 8 kilobits, but the industry standard is to report in kilobits.)
Verizon was the only fiber service with sufficient responses for us to analyze its results, which, at 293 Kbps, were even faster than the ÂÂaverage fiber number. If you can get FiOS, you'd be crazy not to go for it. Verizon currently charges $44.95 a month for an advertised 15 Mbps download rateâ€â€about the same as you're probably paying for cable and perhaps slightly more than DSL. Quadruple that price to jump up to a 30 Mbps plan.
Even if you can't get fiber to your house, the closer the fiber comes to your residence, the faster your connection is likely to be, because you're connecting to the Internet backbone with fewer hand-offs. The New York/New Jersey/Connecticut area has a heavy concentration of fiber optic lines. Optimum Online, the broadband service from Cablevision, which serves many parts of the tristate region, averaged 235 Kbps on the SurfSpeed test, giving it a 12 percent speed advantage over Cox Cable, the second-place cable-modem finisher, and is 40 Kbps faster than AT&T Yahoo! (formerly SBC Yahoo!), the fastest DSL provider. These speeds have helped Optimum Online win PC Magazine's Readers' Choice award for ISP satisfaction in our annual survey for the past two years. (For more on our survey, see go.pcmag.com/sr.) On the other end of the scale, Qwest was, by far, the slowest ISP; its DSL service averaged only 109 Kbps on the SurfSpeed test.
Cable tends to be faster than DSL because cable operators optimize channel bandwidth frequencies over the coaxial cable, and because DSL has to live within the limitations of legacy telephone lines. It's worth noting, however, that DSL is typically cheaper than cable access, and according to a recent JupiterResearch report ÂÂreleased in June, it will become more popular than cable by 2010, thanks to lower prices.
Other Findings
SurfSpeed gathers data about the user's CPU, RAM (both total and available), and operating system. Although we suspected that faster CPUs and more RAM might affect download speed, they turned out to be nonfactors. In fact, the results are all over the map: SurfSpeed ÂÂreported faster access with slower processors in some cases, and higher speeds with less RAM. But don't take this to mean that you won't benefit from a faster PC. SurfSpeed focuses on how long it takes to grab bits off the Internet, not how long it takes to display those bits. A faster computer will render Web pages more quickly once the images and text arrive in your browser.
What's the best time to go a-browsin'? See the results chart. SurfSpeed time-stamps all its tests, so we used that data to see when you should set your alarm clock to get up and go online. Not surprisingly, performance does peak in the middle of the night, when fewer people in the U.S. are using their PCs. For cable customers, the best time was 1 a.m. eastern. DSL's best time was 5 a.m. eastern, but there was very little variability between 1 a.m. and 7 a.m. In fact, overall, there was only a 7 percent difference between DSL's slowest hourâ€â€3 p.m. eastern (lunchtime for the California crowd)â€â€and its fastest. Interestingly, cable bottomed out at 8 p.m. in New Yorkâ€â€but again, the difference wasn't great. The difference with fiber was much greater, but we'd like to see more data before drawing conclusions.
The broadband ISP business is still fairly regional. See how the ISPs are spread across the country. When we divided the country into four geographic regions, Comcastâ€â€by far the best-represented provider in our SurfSpeed dataâ€â€was the only ISP to show up in all four. It speeds slightly ahead of Adelphia and Cox in the West and runs slowest in the Northeast.
The Best ISP
So which ISP should you choose? Would that you really had options! The types of broadband you can get still depend on where you live. You probably have only one cable company to choose from and maybe a few DSL providers. Which technology should you throw your lot in with? If you can get it, the answer is clear: fiber. And in general, cable will run faster than DSL, though that's not always the caseâ€â€AT&T Yahoo! outpaced four large cable providers. How should you choose? Grab a copy of SurfSpeed at go.pcmag.com/surfspeed and get up-to-the-minute data about all the ISPs in your area.
Urban Stats
Fastest: San Francisco/Oakland/Fremont, averaging 227 Kbpsâ€â€11 percent faster than the New York/Northern New Jersey/Long Island metro area.
Slowest: Denver/Aurora, moseying along at only 150 Kbpsâ€â€the Bay Area is 51 percent faster.
Happiest: Minneapolis/St. Paul/Bloomington, despite barely passing Denver's poky connections, reported the highest satisfaction level. (It's fast enough to stream A Prairie Home Companion. What's to complain about?)
Most unreasonably grumpy: Despite near chart-topping speeds (203 Kbps), residents of Silicon Valley's own San Jose/Santa Clara/Sunnyvale area had more gripes than anyone except those contending with the St. Louis area's slow connections (164 Kbps).
How Surfspeed Tests
When you download large files, high bandwidth generally translates into faster file download speeds. But remember, a file download also relies on the site that's serving the file to you. If the site has a small pipe (limited bandwidth) or it's overloaded with users, you're unlikely to see the speed that your provider touts on its Web site.
With that in mind, Bruno Sonnino, the author of many PCMag utilities, developed SurfSpeed for us as a test of real surfing speed. It's a free download (registration required) from go.pcmag.com/utilities, and we urge you to give it a try. We hope to revisit this analysis in the future.
How SurfSpeed Works
Bandwidth testing sites download a single large file, initiating a single connection. Instead of downloading one large file over a single server (with lots of bandwidth), SurfSpeed grabs pages (and page elements) from multiple sites with varying amounts of bandwidth. In addition, just like a real Internet browser, SurfSpeed initiates multiple connections to get all of the page elements. Each connection takes time, and that becomes a part of the SurfSpeed equation.
SurfSpeed
SurfSpeed behaves more like a real browser than like a bandwidth test, and in doing so gives a unique measurement that more accurately reflects how you spend 90 percent of your time on the Web, whether you're surfing or sitting and waiting for a page to load. By default, the utility runs once per hour, "browsing" to top sites AOL, Apple, eBay, Go.com, Google, MapQuest, Microsoft, MSN, MySpace, and Yahoo!. You can run it on the sites of your choice, but we considered results only from these preset sites.
Once you run SurfSpeed, you can begin comparing your results with others. Each time you run a test, the results are sent to the PC Magazine servers. This data is then used to compare your speed with that of others in your ZIP code, state, country, and the world. You can also see how your ISP performs compared both with other ISPs and with itself from state to state.
SurfSpeed
How we got the results
For this story, we considered all the tests run from May 12, 2006â€â€the day after we unveiled SurfSpeedâ€â€to June 8, 2006. In all, we had data from more than 12,000 users in 126 countries, and 128,000 test runs. Since each test hits ten Web sites, we had over one million data points to consider. We restricted our analysis to users in the United States.
Anyone who has spent any time online knows that, no matter how fast your connection, occasionally sites won't be available; pages won't load properly; or downloads will take an inordinately long time to ÂÂarrive. So we worked with Datatechture, a New Jersey-based research company, to help us identify outliers and clean the data so we could do our analysis.
We considered only those ISPs and connection types for which we had at least 50 users. We further filtered out data from users who didn't produce enough test runs over the period that we analyzed.
Fiber Lust
Envy It's a rare day when people regard me with such, but ever since I got fiber-to-the premises (FTTP) installed, it's been a daily experience.
I've won the broadband lottery. No one else at PCMag's offices has FTTP (Verizon calls it FiOS), and scarcely a day passes that someone doesn't ask, "So, how is it?" The answer is always the same: It's great!
Fiber's most obvious benefit is speed: You basically get as much as you're willing to pay for. I bought the second-tier service, for roughly $45 a month, to get 15 Mbps. Interestingly, my cable company offers the same speed for the same price, but my actual cable broadband speeds were far slowerâ€â€45 minutes for a 1.7GB file download on cable and just 12 minutes for the same file, from the same source, through FiOS.
I'm also astounded at the amount of free equipment: a suitcase-size fiber termination box outside my house, a UPS (uninterruptible power supply), and an 802.11g wireless router inside came with the installation (if I quit FiOS, I pay). The UPS, by the way, is for my phone line, which now comes in through the home-powered fiber box. Without the UPS, if I lose power, I lose the phone, too.
I'm tempted to go to the top of the Empire State Building and declare my good luck to the world, or at least Manhattan. Of course that's bound to make those without access to fiber even angrier.
http://www.pcmag.com/article2/0,1759,1994566,00.asp