Sean

Good explanation. I'm sure very few (if any) connections required a test block size approaching 100MB at that time, whereas now some some of the bigger towns are getting 1Gbps fibre. I'm only just learning MySQL and PHP. I see you fixed it. The result I posted on another forum now also shows '96kB'. I remember a few years ago I tried reporting a bug to Ookla about its Speedtest app reporting faster uplink speeds 2 to 3 times faster than my DSL's physical link speed and they came back asking me to run more tests with other servers to confirm its not a problem with my Internet connection. Obviously if the test result is reporting faster than the physical link speed, it's not an ISP issue.

Do you have another device such as a phone or tablet to try the download test? This way it'll help rule out the ISP and router. For example, if this problem also occurs running the download test on your phone, then you know it's not an issue with your Mac. I've a friend with a Mac also and so far he has no problem running the individual tests as well as the scheduled test. Another suggestion would be to try it in a private window (Incognito mode in Chrome.) This will rule out cookies, add-ons (unless specifically enabled for this mode) and some other non-cache related items.

Before I headed off to work, I started a combined scheduled test in 30 minute intervals and had the download test configured as 'Global Multithread' with the Google CDN chosen from what I recall. While at work, I did a few manual speed tests on my mobile and on the workplace connection and switched from multithreaded to the 'Classic Linear' test. I was logged in on both devices. When I returned home, the scheduled test still had a few more intervals to go, but one thing I noticed was that the last few test results shown were linear tests even though I thought I had it set to multithreaded. So I opened up my results in Firefox (active schedule still running in Chrome) and noticed that after I switched to 'Classic Linear' on my workplace PC, the scheduled test started using that mode as shown by the UK flag on the two following test intervals above the 'Eircom' tests below: So for curiosity, I switched the test mode back to Multithreaded in the Firefox window and waited for the next test interval and sure enough the scheduled test ran a multithreaded test: I'm not sure whether this is a bug or intentional, so I thought I'd report it as it caught me by surprise. Otherwise an idea would be to add an extra option to the 'Auto Test' page as to whether to run a linear or multithreaded test along with the various server choices, so that switching between the linear and multithreaded modes while logged in on another PC or mobile device will not affect the current running scheduled test. Hopefully this is the right place for this report as my other bug report thread seemed to have disappeared.

I admit I'm running quite a lot of tests also using the automated tool, testing a recent fixed wireless ISP I subscribed to. Its an ISP that requires multiple tests as its speeds are completely all over the place. My existing DSL connection is consistent, but slow (actual link is 5Mb down / 0.3Mb up) and the new one claims to offer 50Mb down / 10Mb up, but intermittently struggles to offer 1/10th that. While checking over the results, I came across the following one which I actually tested over the Three mobile network while I was at my workplace: On the test results and share result page, the capital letter 'i' appears as a the lower case 'l': Although I regularly use VPN connections either to connect to work (from home) or get around geographic restrictions, I was not using a VPN to cloak my connection at the time.

I live in the north west of Ireland which I'm fairly sure has the worst mobile coverage in the country. In my town, I get 2G-only data coverage and even in Donegal town, the Three network are having pretty bad network issues lately, especially on its uplink. As a result, the starting 96KB test TestMy does in its automatic test is enough to test the connection as it takes 15 seconds or so to complete and thus does not need to proceed to larger blocks. When I went to share the result, I just happen to notice the '96MB' test size for the upload. I am fairly certain it did not upload 96MB of data as that would have taken a few hours... The following are the last two tests I ran, one in Donegal town and the other near home where I get 2G (Edge) data coverage showing the '96MB' upload size: The upload size is shown correctly for larger blocks under 1MB, such as the following test result on my home DSL connection: On the other hand, the speed test itself appears to be the most accurate of any mobile best test I've come across to date , unlike Ookla's app that gives rather odd results such as half the downlink speed and 2x to 3x the upload speed my DSL connection is synced at.

My current fixed wireless ISP Bluebox (Ireland) experiences this issue also, particularly when downloading from UK servers. If I download a large file, I typically get 1MB/s, but if I use a multi-threaded download manager, I can get up to 2MB/s and sometimes higher. From what I heard, Speedtest does a multithread (without saying) and sure enough it reports 15.1Mb/s for me even when I rarely get this on a single transfer. With this site TestMy, I'm getting 8.1Mb/s in the standard download test and 13.2Mb/s in the multithread test using the UK server, which more closely matches my download experience. In the evening, the variation is much greater for me where a multithreaded transfer can be 4 times faster than a single transfer. While this is fine for downloading, it sometimes affects HD playback on streaming services such as YouTube which stream with a single connection. On my workplace DSL connection, I get 11.9Mbps with both the single and multithreaded tests.