Do not trust built in Speed Test


  • , speed test from the device shows 150mbs connect with a gig port device shows actual speed. Lesson learned from calling internet provider out to complain about not getting the speeds I paid for only to be shown that the device does not represent factual speed test 500 mb. This was confirmed from tech support. “Speed test are inaccurate” Seriously disappointed with the purchase and the quality of the user interface and features.


  • I've tested the built in speed test against speedtest.net and testmy.net. All have come back with similar results.

    That said I am on an Amplifi Alien. I had the HD mesh router 6 months ago and it was garbage so I returned it.


  • @John-A-0 It is well documented that this feature is under construction. We are constantly working to improve the user experience and hope to soon get the speed tests to an accurate place.


  • @UI-JT well documented? Wouldn’t it make sense to mark the feature as beta or work in progress within the app?

    It’s been useless for quite some time. Perhaps a couple of years.

    Bad information is worse than no information.

    I get ~200mbps from Amplifi. From a Mac connected via Ethernet ~900mbps to fast.com or Speedtest.net. This isn’t just a little bit wrong, it is completely wrong.

    Just remove it from the app if it doesn’t work.


  • @Robinlmp Thanks for the positive feedback. There are users who have no issue with it, we are working to create a accurate experience for everyone.


  • @Robinlmp The feature works, however the server selection based on the users location can sometimes select a test server from a further location than desired, causing these false speeds. This however, does not effect the majority of our users. Because most can use the built in tests successfully, we decided to leave the test in the app for those users, and let all users know that depending on your environment, you may get false readings until we have a wider server selection available.


  • @UI-AmpliFi is this sarcasm?


  • @UI-AmpliFi it quite patently doesn’t work. If you know it doesn’t for a large number of people then it isn’t fit for purpose. It has never worked for me.

    Would you ship a physical product that you know doesn’t work for a large number of people?

    Does this really work for anyone? I’m not exactly in a remote island in the middle of the ocean. I’m in London and used this in two different parts of London and it was similarly useless in both.


  • @Robinlmp I'm going to throw this out there, because it might be helpful. Speedtests in general cannot be relied upon for rock solid reports. There are a lot of variables which are beyond the control of the device. I have done extensive speed testing on PC, via WiFi AX Alien as well as direct through the Dream Machine Pro.

    I also had the opportunity while the cable guys where in, to see their diagnostics machine running through my switch. I did observe, that their speed testing and the read speed on the panel of the DMP were accurate to each other, meaning the speeds are calibrated accurately. It would appear as well, some aspects are slightly more "optimistic" in the app, but in my case may be due to the fact I'm running it behind a router. The units test consistently, so at the speeds you are talking about, I would be looking for something else. It is not that far off from the panel on the Dream Machine, if anything, speeds are under reported as compared to panel tests.

    What I will mention is the concept of speed tests to a near or distant server will not deliver very consistent results. This could be not only due to your own hardware/software/internet connection, but also due to every point in between you and the test point. Making the issue more complicated, is to consider the load, traffic and internet connection at the test facility.

    We do put ALL our faith in these strange "services" which are essentially "free" I find that I can run 10 tests, on Ethernet and get 10 different results at times, and using different a different service at time will give me weird results, or better or worse than expected. That being said, don't beat on it too bad.

    I hope this info is helpful.

    0_1595466800996_twoweeks-01.png

    0_1595467206092_Screenshot_20200722-201423.png

    The only way to improve this, is to do it router to router, and check/calibrate, and I have to imagine they do that.

    Best Gear I have seen so far, in 30+ years. I would like to see a technical discussion regarding ways to optimize for various types of performance, but all in all, for 95% of users, it's simple, reliable, and it works. The rest is up to you, or us collectively as users to sort out.


  • @unseenone So is the Alien accurate like the DMP?


  • @unseenone if what you’re saying is true, the feature is of no use and it should be dropped. I don’t believe that speedtests are so useless in general though.

    For example, BT use one to determine whether the speed you get from their lines is living up to the speed sold.

    And having dealt with numerous issues over the last month with Amplifi support they are quite happy to discuss issues with speeds that have been obtained from online speed test services.

    Since my post above regarding the speedtest on my AmplifiHD, I’ve since got a Unifi USG (so many issues with routing from Amplifi). The speedtest on the USG is similarly completely wrong. It reports slower speeds than I can get from WiFi devices (WTAF).

    If that isn’t worse than useless I don’t know what is.

    If a feature doesn’t work reliably and consistently it shouldn’t be included. That applies to Amplifi and to Unifi.


  • @Robinlmp I still think it is a good baseline. In fact, the Unify Dream Machine Pro has at least one of it's own machines in the pool of speed test machines, in Dallas and it seems the most consistent on performance I have seem. Unfortunately they are using a random pool you cannot control. That makes sense, thinking logically to run it across a broad spectrum of machines and locations for a better sense of true performance. For your general consumer, it is simply another tool, which I use but I now recognize it is not perfect. You have enough data I provided to demonstrate satisfactory results which I provided. I hope you have your issues resolved by now, if not I would be curious what is wrong.


  • @Gofast I think it is consistent. My upload speeds seem consistently overstated, but i know the DMP was dead on the Spectrum Analyzer - noted, my ms seem really low, I think that's because I am using it Bridge Mode. They also seem over stated, as compared to the DMP. The speeds seem fairly consistent, but nothing near actual speeds the DMP say. This may be due to the fact they use different pools. All in all, I can't complain about either device, and it's the best on the market I have ever seen.

    I will mention, I believe the hub at my ISP is flaky or saturated at times, so our speeds are not consistent, and they are in the process of tracking down the problems because they won't take my word about which of their machines has the problem.


  • @unseenone by consistently, I meant from one user to the next. I agree, for me it is consistently completely wrong. Not just a bit wrong, completely wrong.

    No one would manufacture a car on the basis that the internal lights worked for some users but fully in the knowledge that they don't work for many other customers.

    Knowingly producing a product that is known to wrong is negligent. This is supposed to be higher end kit. You expect stuff to work. And it doesn't.


  • @Robinlmp Your analogy does not work, because the manufacture can control all elements of the vehicle, in the vehicle and it does not rely on external forces.

    I would love to play with your unit and do my own tests.

    I don't consider it negligent, every company out there offers some sort of test, all based around the same pool of test services. The only way any of them could control the speed test, would be to develop a local only test, which measure wifi to client speed, and Ethernet to client speed. It is impossible to consistently measure internet speed, there are too many variables. In order for my PC to get to this forum, it rides over 10 different networks. If there is a problem on any one of them, I might have a problem, speed or otherwise.

    Anyway good luck in your quest


  • @unseenone It is a good analogy because Ubiquiti have essentially said they know it doesn't work! If you are worried about external factors, then how about tyres that don't grip the road for some users? No car company would sell a car that they know the tyres don't grip for a sizeable percentage of users. You may then blame the road or something but it doesn't get away from the fact that:

    a) they know they are shipping a product that doesn't work
    and
    b) others manage it OK

    And no, wifi is not involved. That is the attraction of a router speed test because it helps identify wired network speed.

    I get better speeds via wifi than the router gets with its own speed test. It is completely pointless.

    I'd rather they got rid of it or clearly labelled it as Beta or something because it is just not reliable enough to actually be useful.


  • @Robinlmp This situation is not comparable to your analogy, as building a car like that could lead to physical harm. An inaccurate speed test certainly can be frustrating for users, but is not as dangerous as the car analogy. I hope to answer this question for others as well. It takes a very large number of test servers to give accurate results, you need a server located in almost every city. Building a platform for speed tests that is 100% accurate from the start would be extremely difficult. We are working to add to our speed test network and improve it's accuracy, it will take time to do this. If your speed test is inaccurate there are other resources available to use until we complete ours like speedtest.net or fast.com and generally these are accurate.


  • @UI-AmpliFi firstly, I think that is really a pretty daft response. Please point me in the direction of an Amplifi feature that is life critical? Based on your response you’d never bother making anything work because it isn’t a safety issue, just an annoyance. It seems you are content with annoying your customers. Great.

    The point was about attention to detail and overall quality of a product. There is something seriously wrong when a product is released and it is known to be broken. That indicates a real abdication of duty and poor quality control.

    What would you like as an alternative? A car radio that for large numbers of people plays the wrong stations? It’s not important, the point is that the product doesn’t function correctly.

    And as for not having servers in every city. I understand there is one in London, UK. And guess what? That’s where I am. So cross off that lame excuse.

    If it doesn’t work, remove the feature or at the very least stick a big banner on top saying “this feature doesn’t work”. But you won’t do that because it is embarrassing. How about getting embarrassed about the feature not working rather than embarrassed about being honest about it being useless?


  • @Robinlmp I am sorry you were not satisfied by my answer. I agree if the feature is in the app then it should be a working feature. Maybe you can supply us with support files and I will do my best to improve the issue your having with the speed test feature. You can PM me the supp file. The frustrating part about the speed test is that for many users it does work properly, putting a banner over it saying it does not work would be troubling for our users that do not have this problem. Let's fix the issue your having, drop me a support file. Thanks!


  • @UI-AmpliFi I'll PM you a support file.

    To illustrate the problem.

    1. MacBook Pro connected via wifi amplifi instant access point connected via ethernet backhaul to network switch and then Unifi USG:

    0_1595578486614_Screenshot 2020-07-24 at 09.12.41.png

    1. (another) macbook pro connected via ethernet to network switch:

    0_1595578692087_Screenshot 2020-07-24 at 09.16.55.png

    1. Amplifi speedtest:

    0_1595578727370_IMG_0011.png

    1. Unifi USG speedtest:

    0_1595578844566_Screenshot 2020-07-24 at 09.20.34.png

    Both the amplifi and unifi speed tests are wrong. Completely useless.

    It makes no sense whatsoever to get a faster speedtest via a wifi connected laptop than a hardwired device, especially the router which has a single direct cable to the fibre ONT.


Log in to reply