Widely differing speed test results? Why?

baudwalk
baudwalk Member Posts: 7
First Answer First Anniversary Name Dropper First Comment
In southern NJ, my ad-free Fing on a Comcast xFi router-modem returns significantly different speed results connected to New York City. Both devices are current, cleared of junk and rebooted each evening c. 6:30 pm I have tested. A Samsung Galaxy Tab S3 LTE tablet reports ~10 to ~20 Mbps download and ~8 up. A Samsung Galaxy S9 Plus 64 Kb phone reports ~120 to ~150 Mbps and ~12 up. Both are on the 5 GHz WiFi connection. I'm on my sofa about 10 feet from the xFi unit. These tests are run one consecutively. Obviously the phone has a newer and faster processor, but I would expect Fing speed tests to report the virtually the same results (regardless of the processor) when the time between running tests is <30 seconds.

Now, I understand a speed test should be done with a device wired to the modem -- not happening as I haven't needed a PC since retiring a decade ago -- and Comcast's distribution has many homes sharing the neighborhood loop (versus Fios). 

But the disparity above has been noted for months. I decided to run the test daily, as described above, making conditions in the home as consistent as possible without any other devices on the network, over the past two weeks.

Any ideas as to why the speed test results' disparity? It just doesn't seem logical, Mr. Spock. Thanks.

Tom
Tagged:

Answers

  • Marc
    Marc Moderator, Beta Tester Posts: 3,028
    250 Answers 1,000 Likes 2500 Comments 250 Awesomes
    ✭✭✭✭✭✭✭
    Hi @baudwalk , there are a few threads in here on this subject you might want to search for.  One of the Fing staff referenced this URL for how MLabs performance tests work in one of them (MLabs are who Fing uses for results).  https://www.measurementlab.net/faq/#why-are-my-m-lab-results-different-from-other-speed-tests
    Thats Daphnee, she's a good dog...
  • baudwalk
    baudwalk Member Posts: 7
    First Answer First Anniversary Name Dropper First Comment
    Hi @Marc. I appreciate the link but it doesn't answer my question. The link focuses on different speed tesf apps going to different servers. Obviously the popular Speedtest app and Fing will always differ. If you note, my test was just the Fing app on two devices in house tested immediately in turn over ~2 weeks. If you have an idea on my "test" and question, I would be very much interested. Thanks.
  • Marc
    Marc Moderator, Beta Tester Posts: 3,028
    250 Answers 1,000 Likes 2500 Comments 250 Awesomes
    ✭✭✭✭✭✭✭
    Interesting... Do you get similar results when you only run it from a single device?  I think you could still be going to different servers on your tests, even consecutively plus network paths and congestion can differ minute by minute but I can't answer definitely........  maybe we can get some smart network people to chime in on why your seeing what your seeing?  @Pixelpopper , @vulcansheart , @Hronos , @Ciaran , could you folks shed some light?
    Thats Daphnee, she's a good dog...
  • baudwalk
    baudwalk Member Posts: 7
    First Answer First Anniversary Name Dropper First Comment
    No, two devices as described (S3 and S9+). One test, I agree differences in even closely spaced routing a possibility, but the disparity pattern over ~2 weeks? I've been on line since 1982 with my telephone BBS, on the net since my first web site in late 1991. This stuff not new to me. My guess -- my question -- is the Fing Android software senses the different microprocessors that in turn affects results displayed, but I can't believe the software engineers wouldn't have caught the problem.

    And the test results are similar for each device when the speed test is rerun, one after the other. Obviously the speed test results should be at least statistically similar between the two devices over a population of days. 

    I will look forward to comments of @Pixelpopper , @vulcansheart , @Hronos and @Ciaran as I have no other ideas to answer what I see. Thanks.
  • vulcansheart
    vulcansheart Member, Beta Tester Posts: 117
    100 Comments 25 Awesomes 5 Answers 25 Likes
    ✭✭✭
    edited December 6, 2019 #6
    Simple answer is there's likely a bug with the wireless chipset on the tablet. I can run speedtest benchmarks from a handful of devices on my network using the fing app and get results within 10% of each other. As I'm writing this post, I just compared a Razer Phone 2 to a Galaxy S7 to an iPhone 6S (very, very different hardware) and got 218.6mbps and 233.4mbps and 222.1mbps respectively.
    @Marc is right about how mlabs performs their tests. They have a much different algorithm than the typical ISP provided benchmarks (speedtest.net), which better mimics real-world results that are subject to dynamic routing, bottlenecks, and congestion. Thus, your results can and will vary drastically over time. The goal is to create a trend with an upper and lower normal, and anything that deviates beyond that could be considered a problem.
    41 4c 4c 20 59 4f 55 52 20 42 41 53 45 20 41 52 45 20 42 45 4c 4f 4e 47 20 54 4f 20 55 53
    [Deleted User]