How relaible is the Thinkbroadband Speedtest.

Hi folks,

For a good number of years now I have used the Speedtest on Thinkbroadband site, but was always a bit skeptical of the results.

I am on full fibre 500 down with BT on a wired connection from ONT to Desktop, just ran a Thinkbroadband speed test and also one from Ookle and got the following results, which one do I trust?

Thanks.
[img]https://i.postimg.cc/r0Spdbqd/st1.jpg[/img]"]speedtest


[img]https://i.postimg.cc/xNn1LRd6/st2.jpg[/img]"]Speedtest 2

Edited by joemardo1 (Tue 14-May-24 12:49:14)

I've never had a reason to question TBB as they give me similar results to other sites.

Depends on what you are testing.

Most sites will show you a high end maximum speed achieved during the test. TBB will generally give you more of an average. So, with TBB if it runs slow but has some bursts at higher speeds then you will see a lower overall speed than a provider that shows you the best speed achieved.

In addition TBB has one data centre I believe. You may have congestion on other Internet links from your provider to that data centre. Other speed testers might have a number and they will target you at the one that gives the best result potentially avoiding any routes where there are slow downs (and therefore will show you the ideal conditions rather than what could be real world).

Some speed testers also have servers within the ISP network. If that is the case you are only testing the ISP network and not their onward connections to the Internet.

The other thing to say is that TBB sometimes overly taxes the processor in your device - it could be your device is not able to process the test fast enough and using Linux or a more powerful device would give you better results - Windows has a habit at times of slowing down the TBB test on some machines.

For raw speed of connection TBB is probably not the best. To get an idea of real world performance then TBB is probably more realistic.

Thanks,

I do have an old Desktop admittedly 14 and a half years old Amd Quad core Phenom 945 CPU, with 12 GB of Ram.

I don't have any problems running apps, the computers is fast enough.

Still think there is something up with the TBB speed tester.

Here is another speed test from another site.

https://ibb.co/9H9R8mc

Edited by joemardo1 (Tue 14-May-24 17:14:36)

If you downloaded a free Linux distribution and put it on a memory stick and rerun the test using that you might find it works better - there have been a number of people here that find the TBB client is too much for their machine running Windows. Other thing to try is open the task manager and watch the CPU whilst the test is running - see if it is maxing out.

Thanks,

The highest the Cpu has got to is 86% though a lot of the time a lot lower than that, the RAM highest is 40%

What does this https://www.fireprobe.net/gb speed test result for you?

I have never had a problem with TBB speed test as you can see.

Seems to work ok,

speedtest 4

I find its often pessimistic about my single-thread speed.



The download test files usually go faster as you have no browser overhead, especially on Linux where you can download to null or /tmp.
5GB.zip 30%[=========================================> ] 1.54G 112MB/s eta 36s

Its also only one server over one route via your ISP. The bigger speed tests have many different servers in the UK which may take completely different routes. So the odds of hitting congestion is much lower.

Edited by alexatkin (Thu 16-May-24 23:39:52)

In reply to a post by alexatkin:

The bigger speed tests have many different servers in the UK which may take completely different routes. So the odds of hitting congestion is much lower.

However, you're only testing to one endpoint for a given test.

In general: you should remember that speedtest.net (Ookla) is paid for by ISPs, and is designed to make your ISP look good. That's why they default to a multi-threaded test with high concurrency; it papers over the cracks of low-level packet loss. And that's why they default to an endpoint which is very close to your ISP, or even inside it.

TBB's speedtest doesn't have that bias, so it gives you a more realistic view of what you might observe in real life when doing a single download.

Multi-threaded is more realistic for an office environment where you have many people using the same link at the same time.

Generally speaking, I find the TBB tests to be very low compared to any other test and frankly real world experiences. especially with faster connections. It's been the case for as long as I can recall.

In reply to a post by adslmax:

What does this https://www.fireprobe.net/gb speed test result for you?

FWIW, I found this to be closer to reality in a few tests I did, but it does usually under-report upload speed.

It's always seemed accurate to me having used it for many years. It has been instrumental several times over the years in uncovering issues with my ISP principally through single threaded results being lower than multi-threaded.

Thankfully since switching to IDNet such things are extremely rare.



If the results are not what you expect my many years of experience tell me that you should be talking to your ISP's support department.

I've found it to be very good lately, previous I would get random results! I'm in N.Ireland so distance can be issue, but last few tests are all great.

Been a while since I tried one.. most definitely the best single thread I've ever had on it..

Seems to work here ok, though I don't know how accurate the single thread is as no one else to my knowledge does a single thread test

fast.com can do it if you manually tell it to.

So, I just tried that as min 1, max 1 and I get 1.7Gbs
Trying min 1 and max 16 gives me same result
Even more confused now

Mine ...

In reply to a post by joemardo1:

Hi folks,

For a good number of years now I have used the Speedtest on Thinkbroadband site, but was always a bit skeptical of the results.

I am on full fibre 500 down with BT on a wired connection from ONT to Desktop, just ran a Thinkbroadband speed test and also one from Ookle and got the following results, which one do I trust?

Thanks.
[img]https://i.postimg.cc/r0Spdbqd/st1.jpg[/img]"]speedtest


[img]https://i.postimg.cc/xNn1LRd6/st2.jpg[/img]"]Speedtest 2

I receive very similar results, using the Ookla app. https://www.speedtest.net/my-result/d/9284e0e0-74ec-...

However, I have discovered that the choice of browser is also significant. This is the result of a TBB speedtest with Microsoft Edge on a Windows 11 computer, with a wired connection to the router.
https://www.thinkbroadband.com/speedtest/17247756714... .

But using Firefox on the same computer, the results of the TBB speedtest are comparable to Ookla.
https://www.thinkbroadband.com/speedtest/17247759695...

In reply to a post by joemardo1:

I am on full fibre 500 down with BT on a wired connection from ONT to Desktop, just ran a Thinkbroadband speed test and also one from Ookle and got the following results, which one do I trust?

Ookla (speedtest.net) is paid for by ISPs, and is designed to make your ISP look good. In particular it has a high level of concurrency (multiple TCP streams in parallel) which papers over low-level packet loss.

I always use samknows speedtest!
http://www.broadbandperformance.co.uk/

Edited by adslmax (Tue 27-Aug-24 22:00:44)

It's worth noting this thread died in May, until presumably being woken up by a spam post that has been removed.

I've been using it for many, many years and it has proven itself to be very accurate. That should however always be taken under the advisement that it only measures the single route between you and their servers. It's like measuring the journey time between Plymouth and London. Interesting up to a point but if the journey you really want is Plymouth to Manchester it might not mean as much.

It is best used as a benchmark over time but where it really excels is comparing single to multi-threaded results. If those two are significantly at variance then something is wrong somewhere.

In reply to a post by Colinh58:

So, I just tried that as min 1, max 1 and I get 1.7Gbs
Trying min 1 and max 16 gives me same result
Even more confused now

Suggests you are with a very good ISP. On an ideal network there should be no difference between single- and multi- threaded tests. In practice single-threaded will show more variability but the average of the two should be very similar.

If multi-threaded is significantly faster than single-threaded it suggests congestion or traffic shaping.