How to find a reliable network speed test

Your ISP may be delivering slower download and upload speeds than it promised when you signed up, but wide variations in test results could leave you scratching your head.

Dennis O'Reilly Former CNET contributor
Dennis O'Reilly began writing about workplace technology as an editor for Ziff-Davis' Computer Select, back when CDs were new-fangled, and IBM's PC XT was wowing the crowds at Comdex. He spent more than seven years running PC World's award-winning Here's How section, beginning in 2000. O'Reilly has written about everything from web search to PC security to Microsoft Excel customizations. Along with designing, building, and managing several different web sites, Dennis created the Travel Reference Library, a database of travel guidebook reviews that was converted to the web in 1996 and operated through 2000.
Dennis O'Reilly
3 min read

Fastest download speed recorded by SpeedOf.me
SpeedOf.me's HTML5-based network speed test recorded the fastest download rate of dozens of tests run at different testing sites. Screenshot by Dennis O'Reilly/CNET

Are you getting all the network bandwidth you're paying for? Good luck trying to find out.

As Rani Molla reported recently in the Wall Street Journal, some ISPs are delivering download speeds up to 41 percent slower than they advertise. The figures were compiled by speed-test service Ookla, which owns Speedtest.net.

According to Ookla's figures, the folks in Idaho Falls, Idaho, realize only half the download speed their ISPs claim to provide. Internet users in London, Ky.; Huntington, W. Va.; and Odessa, Texas, don't fare much better: all receive information over their network at speeds far below what their ISPs promise.

DSLReports.com's slowest speed-test result
The slowest download speed test result from DSLReports.com was a fraction of the rate promised by my ISP. Screenshot by Dennis O'Reilly/CNET

DSLReports.com's fastest speed-test result
Rerunning the DSLReports.com download speed test generated a result more than 10 times faster than earlier tests. Screenshot by Dennis O'Reilly/CNET

When I tested more than a half-dozen network speed calculators, the results varied by a factor greater than 10: the lowest reported download speed was a snail's-pace 783Kbps using DSLReport.com's Flash-based test. Another test using the same service resulted in a download speed reading of 11.237Mbps.

The highest download speed test result I recorded was 13.06Mbps using the HTML5-based test at SpeedOf.me (shown at the top of this post). However, the same test generated a download speed of 4.87Mbps on the connection two days later.

(For the record, my ISP promises download speeds up to 12Mbps. I ran the tests in both Firefox and Google Chrome on a Windows 8.1 laptop; many of the services also test phone network speeds, but I didn't run any of them.)

Slower SpeedOf.me network speed test result
A subsequent download speed test at SpeedOf.me recorded a much slower rate than a test conducted days earlier. Screenshot by Dennis O'Reilly/CNET

Conversely, the results of the upload tests at the various services were consistently at or just under 2Mbps. The exceptions were upload-test results at DSLReports.com, whose testing was so inconsistent I ended up discarding all of the service's results.

I ran the tests at DSLReports.com about a dozen times: three times the download results were under 1Mbps, five times they were between 1Mbps and 4Mbps, twice they were around 8Mbps, and three times the download speed the test reported topped 10Mbps.

Does the type of speed test make a difference?

Many experts claim HTML5-based speed tests are more accurate than tests that use Java and Adobe Flash. Others point out that multithread tests such as those used by Ookla (Speedtest.net and branded by many ISPs) don't represent real-world network traffic as well as single-thread tests.

The most consistent test results were recorded at Speakeasy's Flash-based Speed Test and at TestMy.net's HTML5-based tester. Of course, the services' tests may be consistently wrong. After running several tests over a span of days, all of Speed Test's download results were within a few kilobits of 11.5Mbps. TestMy.net's download scores in both its single- and multithread tests exhibited a bit more range than those of Speakeasy's Speed Test, but they averaged about 11.2Mbps.

The results of the HTML5-based speed tests conducted at Bandwidth Place ranged from 5Mbps to 11Mbps, those at Toast.net exhibited a similar range, and the Flash-based tests at ZDNet's Broadband Speed Test recorded speeds from 5.8Mbps to 11.4Mbps.

Not surprisingly, the highest consistent speeds were reported when I ran the tests offered by my ISP, AT&T. The company's speed tests are provided by Ookla, as are the tests at many other network providers. (Note that the Java-based network tester at the FCC's Broadband.gov runs on the Measurement Labs platform, which doesn't support the Safari, Google Chrome, or Opera browsers. The FCC's test also requires that you supply your street address.)

With only one exception, all the download tests I ran at the AT&T Internet Speed Test and at Ookla's Speedtest.net indicated speeds of 11.5Mbps or greater. One of the dozen-or-so tests recorded a download speed of 10.4Mbps, and several of Ookla's Flash-based test results exceeded 12.5Mbps for downloads.

After conducting more than 100 network speed tests from many different providers over the course of several days, I'm confident my ISP is delivering speeds approximating -- and perhaps exceeding -- those it promised when I signed up for the service. Whether any of the speed tests I tried truly represent real-world network traffic is debatable.

HTML5-based speed tests such as those offered by SpeedOf.me and TestMy.net seem to have an advantage in that they require no additional software. If you suspect you're paying for more bandwidth than you're actually getting, you needn't trust your ISP's test results to make your case -- especially if you happen to live in one of your service's dead zones. Hello, Pocatello!