The modern HTML5 way to test internet speeds, dslreports.com/speedtest

Posted by Paul Braren on Aug 16 2015 (updated on Jan 6 2016) in
  • Network
  • Productivity

    • A DSLReports engineer recently contacted me to let me know about their new HTML5 speed tests. He had apparently read my informal speed tests at:

    I replied to him that I had planned to retest using his site more carefully soon anyway, and that I've used DSLReports quite a bit, all the way back to their start in the early 2000s.

    I had already been using their new HTML5-based speed test these past few months, which of course works fine for Fiber, Cable, DSL, 4G, 3G, Satellite, and many more.

    The contenders.

    DSLReports - HTML5-based speed test:

    SPEEDTEST - Flash-based speed test:

    At DSLReports, the claim:
    Why is this the best speed test?

    This is the worlds first multi-stream HTML5 speed test. Multi-stream testing overcomes the need for the speed test server to be physically very close to your location. It also represents the modern usage pattern: someone is streaming netflix, a console is downloading a game patch, a VOIP or Skype call is taking place. Of course if you want single stream testing you can set that option in the handy preferences panel.

    Test time!

    Today, I had that brief chance to kill my household's internet when nobody was actively using it, unplugging the WiFi router, connecting directly to my cable modem. Also good timing overall with the recent Windows 10 Pro release, with a SSD install on a ThinkPad T61p laptop.

    Before we get started here, worth noting that my speed tests are usually rather consistent 24x7, about 170 Mbps down, and about 20 Mbps up, with a latency around 20ms, partially dependent upon the chosen test server's distance.

    SPEEDTEST-Results-History
    speedtest.net results history sample. http://www.dslreports.com/speedtest

    Just in case there was local internet congestion, I chose Sunday morning to do the tests, less likely a lot of Netflix streams in the neighborhood, here in Wethersfield, CT.

    So I took that ThinkPad's gigabit port and Cat5E direct-attached to the Motorola DOCIS 3.0 SB6183 Cable Modem, rebooted everything, and made sure no Windows Updates were going on in the background. That aging ThinkPad's older CPU means a perfect chance to see if Adobe Flash ads and test tools seem to skew test results.

    DSL Reports Tracking Over Time results
    DSLReports results history

    Ookla had some issues today.

    Well, as is often the case with recording live video (seen below, with no cuts), things did not go quite as planned. There must have been some server-side speed issues today, since both:

    were slowed today. They both leverage Ookla, both underperformed, and both gave inconsistent results.

    Did I check my cable modem? Sure, it was fine, with 12 of the 16 channels bonded, normal for my particular Cox High Speed Internet Ultimate tier of service.

    That's Ok, I had already tested both before.

    The video below focuses more on CPU and UI comparisons anyway. So with this particular test run, I really only lost this one chance to know if the top-end results were significantly different. I had already done enough testing mostly off-camera earlier this month, through my Linksys EA6900 (with almost no other traffic). So I'm still able to say that the test results for both SPEEDTEST and DSLReports tend to be similar to one another, and consistent. Your results may vary, of course.

    Aug 04 2015 SPEEDTEST and DSLReports results

    176.85 Mbps down / 22.97 Mbps up www.speedtest.net/my-result/4556559358
    177.62 Mbps down / 22.36 Mbps up www.speedtest.net/my-result/4556562101
    178.68 Mbps down / 22.97 Mbps up www.speedtest.net/my-result/4556564986

    165.30 Mbps down / 21.36 Mbps up www.dslreports.com/speedtest/1015212
    165.10 Mbps down / 21.70 Mbps up www.dslreports.com/speedtest/1015220
    170.10 Mbps down / 21.61 Mbps up www.dslreports.com/speedtest/1015240

    Observations.

    • The many ads on Flash-based SPEEDTEST are really annoying, and if you're not even using your own machine, you don't necessarily want to install and maintain an ad blocker just to side-step this.

    • The one ad on HTML5-based DSLReports isn't as annoying, and overall CPU utilization is a bit lower during speed test runs.

    • SPEEDTEST allows me the ability to login and track the results on somebody else's system, DSLReports merely shows all users results mixed together. This is unfortunate.
      Aug 18 2015 correction - Note the screenshot of "Test History" above also says "This browser or IP address" which may not be an accurate depiction of just the data you want osee. Merely creating an account at DSLReports, then logging in, fixes this. I'm sorry about the above mis-statement. DSLReports history is conveniently saved for you.

    • DSL Reports allows me to annotate the reported results during the speed test.


    What could be improved?

    (nitpicking because I care)

    • Why did only one of my upload tests complain about port 8080 being blocked?

    • Something that looks a little more like a tachometer or bar graph would be helpful, but not if it chews up significantly more CPU cycles.

    • The font for the city names for the latency tests can be blurry in some browsers in some circumstances, and perhaps the "winner" with the lowest latency result could be highlighted.

    • Just to make things slightly easier to type and remember, thank you for that redirect from speedtest.dslreports.com to www.dslreports.com/speedtest. But wouldn't it be better the other way around, landing visitors on speedtest.dslreports.com, dropping the www? While we're at it, what about going https? I realize that's easier said than done, having gone through.

    • DSLReports may need a native mobile app, since I found the HTML interface in Safari and Chrome on iOS to be a bit problematic.

    • Both sites seem to grab just the moment in time the download test finishes as the final speed result, rather than any kind of running average. That behavior in SPEEDTEST can be seen with this animated image.

    • Results + Share page needs a re-test button, rather than having to back-arrow to re-test. C'mon, I admitted I'm nitpicking here.

    Closing thoughts.

    For modest systems, the overall time you need to get some simple tests done is likely a bit less with DSLReports. Not just because of fewer ads, but because you don't have to worry about Flash or fiddling with trying out different test servers to see which gives the best results. The algorithms that SPEEDTEST uses to find the nearest server may pick the geography right, but for me, running it without tinkering with the test server chosen doesn't tends to get me significantly slower results.

    I've also found Comcast and Cox's own customer-accessible speed tests to be lacking, and I'd rather have one site I can rely on. Once in a while, only Cox's tool would give me the fastest upload speeds, but of course, your results may vary.

    I realize the underlying designs of these test sites might not be to give you the absolutely fastest results. When you think about it, something more like the results you'll feel as you "surf" and download and upload to and from sites across the globe is also be of value.

    For the kind of quick tests I tend to do for folks right after a new cable modem or WiFi router or PC is installed, it seems DSLReports is better. They're ready to leave Adobe Flash behind, quite the trend lately, with more focus on battery life and mobile. It's about time.

    Please drop your comments or experiences with internet speed test tools below, with no login required!

    Speed Test Comparison - Close look at SPEEDTEST Flash] and DSLReports [HTML5].

    Aug 18 2015 Update

    I have now noticed that DSLReports seems to have small but consistently lower upload speed tests than SPEEDTEST. If absolute top speeds are your goal, this could be an issue for you. If a test tool that works consistently from any browser at any household with a minimum of fuss, DSLReports may be a better choice. Was helping a family member moved to a new place yesterday, here's some RCN test results from that home. Perhaps noteworthy that WiFi was the only option there, and the difference in upload speeds ware higher than my home's wired Cox connection.

    60.90 Mbps down / 12.56 Mbps up dslreports.com/speedtest/1120619
    58.10 Mbps down / 10.04 Mbps up dslreports.com/speedtest/1120660

    59.69 Mbps down / 18.53 Mbps up speedtest.net/my-result/4590435886
    62.59 Mbps down / 18.30 Mbps up speedtest.net/my-result/4590446522

    Aug 26 2015 Update

    I have some more tests to report, from August 22nd, while briefly visiting Cornell University, using Chrome on a Windows 10 Core i7 system with a wired internet connection, leaving all tests at defaults. DSLReports did very well,and I think those are the fastest upload speeds I've ever seen first-hand. Way faster than at the Microsoft NERD Center. See also Bufferbloat post, faq, and Measure Your Bufferbloat! New Browser-based Tool from DSLReports.


    135.50 Mbps down / 244.00 Mbps up dslreports.com/speedtest/1157965
    130.90 Mbps down / 259.70 Mbps up dslreports.com/speedtest/1157997
    134.30 Mbps down / 265.00 Mbps up dslreports.com/speedtest/1158002

    92.05 Mbps down / 39.68 Mbps up speedtest.net/my-result/4603333577
    91.63 Mbps down / 38.78 Mbps up speedtest.net/my-result/4603335130
    91.12 Mbps down / 39.75 Mbps up speedtest.net/my-result/4603335925

    I also received some interesting email from Justin at DSLReports, and I have permission to share it. I think you'll appreciate his detailed response to my August 18th test results:

    On Aug 20, 2015, at 10:02 PM, Justin wrote:

    Cool I'll get to the other part of your email soon but some comments on the test of the speed tests:

    The reason for the lower upload speed you noticed was because the test opted to use web sockets for the upload portion. You can change that in preferences and if you do, you'll find it matches flash again. web sockets are cool, and I wanted to use them, but some browsers don't handle them well (a lot depends on CPU and browser implementation). Unfortunately the test you did on RCN was right at the border where web sockets are no longer used because of these issues. Any faster on the download and it would have done upload measuring the "normal" way and you would not have read low. I'll adjust it to be even more cautious about using web sockets.

    Second comment, the speeds are not the last speed - as you correctly point out is not ideal - but are actually a rolling median of instant speed. The graph that unfolds during the test is actually the median, so you're not seeing instant speeds and wherever that ends up, but you're seeing the median speed over a trailing period..

    The idea is to try to capture the steady state speed but without being dragged down too much by the ramp up period at the start. Generally it works well but sometimes when someone has an iffy connection, the median can be less repeatable from run to run because it sometimes is "full speed" and sometimes has enough slower numbers to drop down a chunk. One of the drawbacks of median speed. Anyway I guess the takeway I wanted to point out was the speed result is not the ending instant speed.

    thanks
    -Justin

    Motorola SB6183 showing 12 of 16 channels in use
    12 of 16 channels in use is normal for my Cox Communications connection.

    Disclosure: I have no affiliation with SPEEDTEST or DSLReports, and there are no affiliate links here. No promises were requested or offered.

    Nov 30 2015 Update

    Managed to hit a new high today, thank yous to:

    Feb 12 2016 Update

    2016-02-12_8-35-42

    Good numbers, clean consistent performance for months now, still using the same Linksys EA6900 WiFi router and Arris/Motorola SB6183 cablemodem for my home network. Figured I'd say something positive about an ISP, for once:

    Now I'm just waiting for Cox Gigablast to become available in my 06109 area ;-).

    2016-02-12_9-03-50

    Feb 13 2016 Update

    Notice the shape of the curve, appears like traffic shaping, with a pretty sweet 320 MB/sec peak. Perhaps this is what Cox Communications calls PowerBoost, but this boost is of a rather short duration, so I don't really know.

    2016-02-13_13-24-59

    Jan 06 2016 Update

    It would appear that Cox is no longer throttling the download speed a few seconds into the test. Note the dramatically different histogram shape. By far the fastest test I've ever seen in my home, and this is through a wired connection to my Linksys EA6900, to my Motorola SB6183 DOCIS 3.0 cablemodem. Yes, 343.7Mbps down and 31.7Mbps up ain't too shaby!

    See also at TinkerTry

    See also

    TinkerTry - PCs, EVs, home tech, efficiency and more, including virtualization. My opinions here, not my employer's.

    Paul Braren, MCSE in 1998, VCP #2681 in 2005, vExpert 2014 to 2024, Veeam Vanguard Emeritus, 2015 to 2019.

    Support TinkerTry

    PayPal | Patreon | VMware Store

    After 6 successful years testing then shipping well over 1,000 Xeon D Bundles, Wiredzone had to stop selling them in mid-2021 due to cost, supply, and logistics challenges. So far, Xeon D-1700/2700 (Ice Lake D) was a minor refresh for 2023, with Xeon D-1800/2800 (Granite Rapids D) refresh looking a little better for 2024. More cores, but still just PCIe Gen4 and DDR4. Looking ahead, I'm glad it's Pat Gelsinger at Intel's helm. I'm also grateful to have had the honor of working at VMware when Pat was the CIO there. I'll leave it at that, given the whole Broadcom thing.

    Easily update VMware ESXi at TinkerTry.com/easy-update-to-latest-esxi