TBB Speed test. Has it changed?

Recently, I have only managed a test result of just over 100Mb using the TBB speed tester. Until a month or so ago I was getting >200Mb. Other testers, like speedtest.net, Uswitch and testmy.net give me 210Mb any time of day. Even file download tests give me >200Mb, so I know my connection is fine.

I am concluding that there is congestion between me an TBB servers or they have reconfigured the test so my set up can no longer use it reliably?

Any ideas?

It would be a good idea to post the link to that result so we can see the shape of the graphs.

What product are you on? It sounds like FTTP.

There are multiple versions of our test, so posting a link to the results would help to identify which version you are using or used.

If this is Virgin Media and its 200 Mbps service then no sign of widespread issues as people are getting the expected speeds from us http://tbb.st/1464689002817010855 and http://tbb.st/146469359612764970838

Our test is more sensitive to congestion than some others, especially as some may opt to use servers internal to the ISP.

Mt connection is 200Mb/12 with Virgin Media, FTTN


Here is a speedtest I just did via TBB.

TBB test


Here is speedtest.net (Preferred)

Speedtest.net


Speedtest.net (Non-Virgin)

Speedtest.net.


Here is a trace to TBB.

Microsoft Windows [Version 10.0.10586]
(c) 2015 Microsoft Corporation. All rights reserved.

Tracing route to thinkbroadband.com [80.249.99.130]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms router.asus.com [192.168.1.1]
2 * * * Request timed out.
3 10 ms 8 ms 8 ms brhm-core-2b-xe-822-0.network.virginmedia.net [80.0.2.201]
4 * * * Request timed out.
5 * * * Request timed out.
6 13 ms 13 ms 17 ms brnt-bb-1a-ae0-0.network.virginmedia.net [62.254.42.98]
7 13 ms 13 ms 13 ms tclo-ic-2-ae0-0.network.virginmedia.net [212.250.14.202]
8 14 ms 14 ms 13 ms linx-gw2.thdo.ncuk.net [195.66.236.240]
9 13 ms 14 ms 14 ms te1-1-1-31.core-rs2.thdo.ncuk.net [80.249.97.85]
10 14 ms 13 ms 13 ms www.thinkbroadband.com [80.249.99.130]

Trace complete.

Okay well can confirm no changes have happened to the flash speed test,
http://www.thinkbroadband.com/speedtest/button/14646... shows the graphs that are actually more important than the numbers and suggests maybe a PC/Software issue

What does http://labs.thinkbroadband.com/speedtest give

Reason for thinking its not us is if we look at other users on the exact same version of test that you used we see
http://tbb.st/146469359612764970838 i.e. what you expect

Your labs.thinkbroadband link gave me 210Mb down 11.9Mb up. Latency was high at 31ms Bufferbloat C and E. It is worth noting that the test fluctuated between 190Mb and 225MB as it progressed. (The single-thread test struggled to get above 100Mb as it progressed) I do understand what you are saying about such fluctuations. Perhaps the test operates near the limits of my processor. I checked processor usage during the test. It never went above 88%.

My PC running Windows 10, although fairly old, has never had problems with the processor managing greedy tests.

Intel(R) Core(TM)2 Quad CPU Q8300 @ 2.50GHz, 2500 Mhz, 4 Core(s), 4 Logical Processor(s)
Memory 8Gb Nvidia Geforce 220

Edited by deleted (Tue 31-May-16 13:57:40)

Okay so suggestion is that something in the flash library and how our flash test operates is misbehaving on your particularly combination of software and hardware. We are seeing plenty of others with the same provider and the flash test still showing good results so sure its not anything we've done.

Guess we will have to leave it then.

I will just have to figure why, all of a sudden, my TBB average has dropped from >200Mb to just over 100Mb. The only change here has been a change of Windows 10 version, to the most recent.


Better result using Microsoft Edge. So, could well be a Flash problem in Firefox and Chrome.

TBB Speed test

Edited by deleted (Tue 31-May-16 15:47:34)

In reply to a post by roughbeast:

Better result using Microsoft Edge. So, could well be a Flash problem in Firefox and Chrome.

TBB Speed test

Yes that looks right now, the single thread speed is to be expected with Virgin unfortunately, points to some potential congestion issues within the local neighbourhood network.

In the past as versions ebb and flow for the flash in different browsers oddities show up for some people and then they vanish, ie. oddities are introduced as security fixes are rushed out. This is why we are leaving flash behind.

How does that work?

If I can download files at >200Mb; complete multi-thread tests from non-Virgin servers at 210Mb any time of day, yet a single thread test can trip over local congestion.

It works like this (not very technical and skips over a tonne of technical aspects, but it demonstrates the issue at hand).

If you imagine 2 users, call them Bob and Mike.

There is 100Mbps split between the two of them at the local exchange (virgin media use nodes but anyway, this is just an example).

Bob downloads a file (1 threaded), he gets 100Mbps.
Now Mike runs a speedtest, single threaded. Bobs download slows to 50Mbps, Mikes speedtest gets 50Mbps (sharing available backhaul).

Say Mike runs a speedtest with 9 threads, Bobs 1 threaded download falls to 10Mbps, Mikes speedtest shows 90Mbps.

With multi threaded speedtests, congestion is hidden a lot more.

Hence ISPs send you to www.speedtest.net as it's multi threaded and very good at covering up congestion issues.

Edited by ukhardy07 (Tue 31-May-16 18:27:20)

As is the BT Wholesale one. When it used to display it, it was 15 or 16 threads.

If you use TBB download test files (under 'Tools'), this may help pinpoint things.

OK. I get that. So multi-threading is a bit like downloading using torrents. With a torrent I can download off 25 seeders, from 25 different global locations. Indeed I have downloaded downloaded a 2Gb film in 80 seconds, thus using up most of my bandwidth. (And everybody else's by the sound of it!)

Tell me: When I download files from TBB, in order to test my speed, I clock up >200Mb only by downloading 2files at once. Each file downloads at just over 12MB a second. Is this because TBB servers only dish out at about 12MB or is this because I am using 2 threads, thus getting around local congestion?

Edited by deleted (Tue 31-May-16 19:01:54)

What is replacing Flash?

I did that. To achieve 200Mb I had to download two files simultaneously. They both downloaded at just over 12MB a second = 25MB X8 = 200Mb. I achieve this every time, whatever the time of day. As I say. There is nothing wrong with my connection. It seems that Flash problems might be the issue; that and possible problems with the single thread test.

To compare the single-thread TBB speed test result, you need to use a single download. Do they match? You can then compare them with testmy.net single-threaded and single downloads from elsewhere.

How fast is the TBB test file server? Can it download a file at 25MB a second? (200Mbs) For me one file maxs out at 12.8MBs.

Edited by deleted (Tue 31-May-16 19:51:53)

You are probably getting round local congestion. The TBB servers can feed well above your 200Mbps - given their location it is probably up in the Gbps range.

Yes, and the rest ...

Just now got 202.5Mb on a single thread at Testmy.net. The other day I got 207Mb on the single thread.

Edited by deleted (Tue 31-May-16 20:02:58)

That one has just given me a peak of 729.44 Mbps !

Testmy.net? I'm looking for your result on the shared results graph. I can't see it.

Try this one ... with a 645Mbps peak: http://testmy.net/db/XtuUQJ0a2&framed=1&framed2=1


or: http://testmy.net/db/vFRd4Kaks&framed=1&framed2=1

Edited by MHC (Tue 31-May-16 21:01:05)

Tell me: When I download files from TBB, in order to test my speed, I clock up >200Mb only by downloading 2files at once. Each file downloads at just over 12MB a second. Is this because TBB servers only dish out at about 12MB or is this because I am using 2 threads, thus getting around local congestion?

Congestion by virginmedia yep. Yes you are getting around it. The server can easily put out those speeds.

Virginmedia offer these headline speeds, higher than other ISPs but due to the nature of the network, it only takes less than a handful of users on your local node and congestion hits, often a lot worse than you experienced. We are talking just 4 users torrenting could have a major impact. I used to drop down to 10Mbps on a 200Mbps package, which believe me is not that uncommon, just most users do not run speedtests ever & if netflix streams & asos loads pictures they are happy.

FTTC providers are generally far less prone to congestion... It's why I avoid VM.

Edited by ukhardy07 (Tue 31-May-16 20:56:16)

So, you have access to a Gigabit connection?

No, I have an 80Mbps FTTC which runs at around 75 ish!

Just shows how inaccurate some testers can be,

So you see your test result here:
http://www.thinkbroadband.com/speedtest/results.html...

A line without congestion, you would see the green and yellow lines almost mirror one another.
Yours do not, you have the green line less than half the yellow line.

And, for roughbeast, this is what a non-congested result should look like:
http://www.thinkbroadband.com/speedtest/results.html...

The download file server can exceed 200 Mbps by very large margin

The http://labs.thinkbroadband.com/speedtest which can run over SSL too if you goto https://labs.thinkbroadband.com/speedtest if you want make things work harder.

Out of curiosity what does fast.com give you? (quick download test only)

Have you tried turning off any anti virus protection? See how the free version of AVG affects my upload graph.

With AVG enabled

http://www.thinkbroadband.com/speedtest/results.html...

With AVG disabled

http://www.thinkbroadband.com/speedtest/results.html...

I don't see the point of Fast.com unless you want to know how your connection with Netflix is. My results range between 82Mb and 212Mb.

I use F-Secure as provided by Virgin. It's processor usage is pretty low. At no point during a Flash-style speedtest has my processor been more than 88%.

Your HTTP tester gave me 205Mb on its second down speed run with processor usage 25%

The HTTPs tester gave me only 122Mb on its second down speed run with processor usage 50%

OK! You made your point. As MrSaffron has pointed out, Most of the time on a shared connection with a potential download speed of 212Mb (My config) users don't experience a diminished service. Only speedtests reveal issues.

The only time I cane my connection is when using multiple torrents or a single torrent with many good seeders. Because this is effectively multi-thread I often get the full 25MB. I can only guess the effect on other users on the same coax. For this reason I time my torrent use for only between the hours of 12:00 midnight and 6:00 am.

Edited by deleted (Thu 02-Jun-16 10:35:58)

Higher CPU to be expected as all the transfers are end to end encrypted on SSL. We know our end can go faster but it can be useful to show up oddities in how providers handle traffic.

I'm with VM coax. Since I opened this discussion I have upgraded to 300Mb/20Mb.

Typically on Speedtest.net I get 317Mb/20.4Mb. Using TBB file download method of speed testing I usually achieve between 36MB and 39Mb per second. This is between 300Mb and 317Mb. Invariably on TBB's main single thread speed test I barely get 120Mb whether it is on or off peak, 8 in the evening or 3 in the morning. Congestion seems to have nothing much to do with it, unless people on my quiet uncluttered suburban street are caning their connections with torrent downloads all night!

Frankly, I'm not sure what use the TBB speed test is. It does not relate to the real world. Most internet experience is multi thread in nature and / or we are restricted by server limitations. For example, my son on his new gaming computer was downloading his game configure files from Origin last week. They downloaded at 311Mb during peak time. If the TBB speed test varied from congested times to uncongested times, I would have more faith in it.

I remain convinced that the TBB test has changed. For years I always got results close to my paid-for speed. Suddenly , without any changes at my end, I never get above 120Mb even on a 300Mb connection.

I would say the opposite.

The biggest chew of bandwidth today is streaming, every streaming service I know of is single threaded, youtube, netflix, lovefilm, iplayer, 4od, itvplayer etc.

If you grab a file of github, chrome update, http download, again its single threaded.

Web browsing is multi threaded usually but that on its own doesnt really need bandwidth it just needs a clean connection, and ideally low latency.

Where the VM use case scenario is optimal for (only good with high # of threads) is perhaps torrents, newsgroup binary downloads, steam, battlenet and origin.

FTP is single threaded by default although some ftp clients can make it multi threaded providing what you downloading is split into different files, to download a single file on FTP via multi threading is possible but a bit of a hassle.

Email is also by nature single threaded per email account.

Play store downloads, single threaded.

Anything else?

If you look on the VM forum, around the time the new packages got introduced, some weird QoS implementation was introduced as well, presumably to allow VM to market these new speeds without paying for the upgrades required in their infrastructure, sure there will be upgrade work been done, but this fiddling will have made the upgrade work not so expensive. Bear in mind this comment is not backed up by proof, as VM have been silent when asked about it on their forums so no confirmation.

Edited by Chrysalis (Mon 29-Aug-16 16:27:03)

Care to post a link to one of our speed test results?

TBB speed test

TBB Speed test

Edited by deleted (Mon 29-Aug-16 23:00:32)

Streaming doesn't need much bandwidth so on a 200Mb or 300Mb connection it doesn't matter, from the point of view of performance, whether it is single or multi-thread. Similarly with FTP streaming. I use FTP from Livedrive cloud service to stream video through Kodi. My band width allows me so stream a 50Gb 3D video file seamlessly after a few seconds delay. (Naturally I have maximised cache in Kodi.) I don't know if Livedrive actively split files over FTP or SFTP or not. I wouldn't have thought so because Kodi would struggle to coordinate a video split multiple ways. (uTorrent seems to manage that though, I often stream a video file as it downloads from multiple seeds.) I certainly use my VM connection for torrents, the ultimate in multi-threading, using a good VPN server offshore for security. Download speeds of over 15MBs even via VPN server in Canada are quite common.

Anyway, I take your point about single thread versus multi-threaded applications. My twin teen boys use Steam, Origin and Battle net heavily, so I guess in our bubble multi-threading is perhaps the norm.

What surprises me is that my 300Mb connection, apparently, is effectively 120Mb or less over single threaded test regardless of time of day if the TBB test is anything to go by. What really puzzles me is that there is no noticeable difference between peak and off-peak times. This is a quiet suburb with large semis and detached properties with retired folk or average-sized families in them, not a contended area. I would expect a variation in speed test results over a 24 hour period. Perhaps there is something I am not understanding.

Edited by deleted (Mon 29-Aug-16 21:44:50)

congestion is not the only thing that can affect single threaded speeds, tcp buffer tuning and congestion control will also affect it.

What OS are you running the speedtester on?

You appear to have missed that the the speed test shows single and multiple thread at similar speeds, but as that is the older flash based test and flash versions are morphing with all the security threats all the time there can be the odd issue, so can you run http://labs.thinkbroadband.com/speedtest and also https://labs.thinkbroadband.com/speedtest and post the results from both.

The last time Ofcom released its SamKnows results there was a note about a complaint from Virgin Media that the download test of that was not showing their top speeds. The SamKnows boxes at that time used a three thread download test, guessing they'll being doing to 6 or more to give higher speed but hopefully retain the lower more congestion sensitive version.

The difference between 1 thread and six can be seen in these recent tests
http://tbb.st/1472499159482015655
http://tbb.st/1472481446250852455
http://tbb.st/1472481350556651055

A sample test from flash test user
http://tbb.st/147248579818632990092

None of this visible on most other speed testers since they don't show the throughput graphs with a proper scale and do not do single versus multiple thread testing.

TBB Speed test (Labs HTTPS)

TBB speedtest (Labs HTTP)

TBB Speed test (Old Flash)

TBB speed test LABS HTTPS 8.55 am

My tests on the old flash test don't show that much difference between single thread and multi-thread? Is that so?

Not a massive difference between the two Labs tests either? 30Mb difference isn't great in the context of a possible 300Mb.

Neither test shows anything like my actual speed.

Edited by deleted (Tue 30-Aug-16 08:56:02)

Windows 10

In reply to a post by Chrysalis:

every streaming service I know of is single threaded, youtube, netflix, lovefilm, iplayer, 4od, itvplayer etc.

Not true for Netflix and Amazon, can you post your test results?

I know you asked Chrysalis but I'm not sure if the Netflix speed tests will show much up. I can confirm that the Amazon speed test is multi-thread, not sure about their video output. I don't know where their test servers are, but I only got a princely 13.5Mb off them. Netflix say they download files from a variety of servers for their test. I got 120Mb off them. Again I don't know where their test servers are.

Anyway, outfits like Amazon and Netflix set up their speed tests to make their connection to you look real good.

Hi, my post is referring to streaming using the real service, not speed testers.

Netflix test will depend on where the CDN is and for a large provider like Virgin Media there is likely to be one on their network

And what test/app is showing your actual speed?

Nothing much in terms of overall difference between all those tests

Post deleted by roughbeast

In reply to a post by MrSaffron:

And what test/app is showing your actual speed?

Nothing much in terms of overall difference between all those tests

Not sure what you mean.

http://www.dslreports.com/speedtest/4851010.png[/img]"]DSLreports

Okay and at 24 download streams (which are from very varied locations usually) you are showing your connection can hit the sold speed, but the fact you need to jump from six to 24 threads shows something is not right.

Some of those files may even be on the Virgin Media network itself, rather than outside the ISP.

So yes that test shows the connection speed has been sold to you, but would highlight that 24 downloads is not a common scenario. Nothing has changed on our tester, and no plans to jump to 24 threads as this masks so many provider issues it is unfair.

So regardless of time of day, regardless or the variable congestion over a 24 hour period, even in the small hours my connection is way off what I have been sold. The amount of congestion is making no difference? I know that VM always default to speedtest.net, and we know why, but do you think I have grounds to complain? Would a complaint stick?

Edited by deleted (Tue 30-Aug-16 13:41:06)

If you can show that standard type services are impacted rather than just 'speed tests' then yes I'd personally pursue a complaint, resolution may simply be to let you leave, which if no other superfast options may not be good.

All that said there is no reason a 10 Mbps connection cannot stream two HD videos and do some web browsing at the same time.

In reply to a post by MrSaffron:

All that said there is no reason a 10 Mbps connection cannot stream two HD videos and do some web browsing at the same time.

To stream iPlayer in HD you need >5 Mb/s nowadays.

I am talking about streaming not their speedtests.

I not missed anything, not sure what you think I got wrong. I was actually defending your speedtester.

Although perhaps what isnt helping matters is why there is different speedtesters on here, I would think by now the flash version should be replaced with a equally functional html5 version that works in the same way, both multi threaded and single threaded testing. Also I think the guest speedtester should be the same way also.

Me too.

then why are you asking for test results?

streaming of netflix is single threaded. Easy enough to know just by watching live connections during the streaming.

Also when I deliberately did some QoS testing on my network to cap individual threads (like VM style QoS), netflix and every other streaming provider I tested was unable to exceed that cap, yet if they were multi threaded they would have been able to as it only applied per tcp connection.

To the OP of this thread I suggest reading this thread on the VM forums, it may give you some answers.

https://community.virginmedia.com/t5/Speed/Very-poor...

Edited by Chrysalis (Tue 30-Aug-16 22:51:03)

In reply to a post by Chrysalis:

then why are you asking for test results?

streaming of netflix is single threaded. Easy enough to know just by watching live connections during the streaming.

As you have just confirmed, you tested the services to determine the number of threads. These are obviously the test results I am referring to. I am asking because I have seen it is not true that Netflix is single-threaded so it would be interesting to see what's going on or where the difference lies.

Running both alongside each other is ideal for comparing changes in one version versus one that has had no changes.

Flipping to the html5 for all with just a link to the flash for those that still prefer it will be the norm shortly and the html5 will then also link back into the profile.

You will note that I did not say two full HD streams and did not specify source. So two 720p streams should be feasible.

BBC has a list of its bit rates at http://www.bbc.co.uk/rd/blog/2015-07-the-development...

Perhaps I should rate limit a connection to 2,4,6,8 and 10 Mbps again and see what does and does not work i.e. did this two years ago to see what did/did not work.

In reply to a post by MrSaffron:

You will note that I did not say two full HD streams and did not specify source. So two 720p streams should be feasible.

BBC has a list of its bit rates at http://www.bbc.co.uk/rd/blog/2015-07-the-development...

You said there is no reason it won't work, using one of the most popular services cannot suddenly become a reason.

The link you give confirms that the video bitrate alone is >5 Mb/s for 720p. More than 10 into 10 does not fit I'm afraid.

@Andrew
I'm looking at the results from both of your Flash and HTML5 tester ran in sequence and although speeds are within 0.2Mbps, the latency figures are markedly different.
The Flash version states <1ms while the HTML5 says 18ms
Flash
HTML5
Is there an answer as to why the values are so different?

In reply to a post by vimto_girl:

In reply to a post by MrSaffron:

You will note that I did not say two full HD streams and did not specify source. So two 720p streams should be feasible.

BBC has a list of its bit rates at http://www.bbc.co.uk/rd/blog/2015-07-the-development...

You said there is no reason it won't work, using one of the most popular services cannot suddenly become a reason.

The link you give confirms that the video bitrate alone is >5 Mb/s for 720p. More than 10 into 10 does not fit I'm afraid.

In reference to iplayerhelp.external.bbc.co.uk/tv/hd_general:

What do you actually mean by �High Definition�

To be classed as "true" high definition, we encode in at least 1280x720 resolution or 720p. We use h.264 with a bit rate of 3.2Mbps and 192kbps audio. This means that streaming programmes in HD quality will put demands on your network connection.

Thanks for the outdated info.

In reply to a post by vimto_girl:

Thanks for the outdated info.

If you feel it's outdated, I suggest you take that up with the BBC

Why don't you do it yourself?

In reply to a post by vimto_girl:

Why don't you do it yourself?

Because I don't think it's outdated.

In reply to a post by vimto_girl:

The link you give confirms that the video bitrate alone is >5 Mb/s for 720p. More than 10 into 10 does not fit I'm afraid.

There's nothing like empirical data...

I just started playing "Ripper Street" in HD on our YouView box, and watched the traffic statistics accumulate on DslStats - which is reading the throughput data from the modem/router.

Before playing: throughput was around 2,000 kB/min
While playing: throughput at minimum was 30,000 kB/min, peak 45,000kB/min, average around 40,000kB/min.

It looks like the iPlayer stream ends up averaging around 38,000 kB/min ... or a hair over 5Mbps.

But it seems to be impossible to get the STB to tell me what kind of streams it has chosen for me.

Not a surprise that the empirical data matches the info provided by the developers themselves. You can see the bitrate and resolution on a browser, which again confirms.

You have been giving people an attitude in several threads of late, nothing wrong with disagreeing, but your attitude is very confrontational and likely to scare people away from daring to ask anything.

So when I get chance it should be impossible for me to stream 4 HD streams on a 20 Mbps connection, will try it out when I have a chance.

Don't trust that flash latency figure at all, in your case. <1ms on FTTC is very odd.

It's FTTP

Still odd. But your BQM is impressive.

What do you get from a ping or tracert to bbc.co.uk?

In reply to a post by georgelnx:

It's FTTP

Lol at that.

I'm nearly done moaning here.

Having reconciled myself to the fact that a VM connection needs at least 5 threads to demonstrate its potential for transferring data. 24 threads is optimum it seems. The question remains- Why doesn't the TBB speed test show at least some difference on a VM connection at different times of day? Is the connection quality so bad that changes on congestion make no difference?

Indeed. Like this:

5 Mb/s, not HD
http://www.thinkbroadband.com/speedtest/button/14727...
https://s22.postimg.io/3q37b8whd/5_Mb.jpg

6 Mb/s, HD
http://www.thinkbroadband.com/speedtest/button/14727...
https://s22.postimg.io/92s1pdkdt/6_Mb.jpg

ping bbc.co.uk averages 4ms at 21:00hr for packet size 32b, size 1464 gives 6ms
This on Plusnet's old network and fixed IP.

I'm not really too bothered by the absolute TCP Latency on the speed tests as I'm more concerned with the relative variations between tests. And the trends in the results.

Edit: to add some details

Edited by deleted (Thu 01-Sep-16 21:08:52)

Can you show me where? Can only think of one where a regular got some cheeky attitude with winks and smiles, trace it back and you'll see why, no respect or politeness given in the first instance, then he wonders why he isn't treated like his majesty in return. I accept it's just his nature though, and I deal with it tongue-in-cheek actually, bless his socks.

In reply to a post by vimto_girl:

... bless his socks.

Case proven?

Okay have seen your BQM and still don't trust the flash answer. When flash was written there was very little consumer low latency connections, so was never a priority.

Remember BQM should be the same as what everyone calls a ping since it is using ICMP packets.

That answer would mean taking a known set of hardware to various places to test it and confirm everything.

Plenty of other people are seeing time based variations with Virgin Media, but it may be that your area does not have this as its so busy, or there may just be something wrong with the cable connection

To be frank your post is a perfect example, and given its a reply to a moderator could be construed as trolling.

As I said will look at in practice on a variety of devices and also since video streaming is not purely restricted to BBC iPlayer will look at what they are offering these days. Some streaming services use more variety of bit rates too, i.e. just because the BBC use one set of encoding rates does not mean everyone uses the same.

In reply to a post by MrSaffron:

Some streaming services use more variety of bit rates too, i.e. just because the BBC use one set of encoding rates does not mean everyone uses the same.

Of course, nobody ever suggested that. You can try all you like with iPlayer, how can >5 ever fit into 5?

I was just pointing out a fact you didn't seem aware of given your advice, it's up to you if you want to revise it.

If a mod is going to make allegations, I am entitled to ask for some indication where it happened and to defend myself. It's not trolling and it's not confrontation. Strange that the genuine confrontation "AS IF!" to simple informative post and the genuine personal abuse "you are snooty" get a free pass, but the tongue-in-cheek coping with it gets scrutiny.

I just re-read that BBC Video Factory article. The speed table seems in this thread to be being taken somewhat out of context. In particular things like (my underlines):-

The new profiles were designed to improve the picture quality available on devices without significantly increasing the associated bit-rate and therefore distribution cost. They were also designed to exploit the benefits of Adaptive Bit-Rate Streaming, with just noticeable changes in video resolution within an Adaptive Bit-Rate adaptation set.
...
The 50fps, 1280x720 profile, however, will be available to those with 5Mbit/s broadband connections.

Additional lower bit-rate profiles will be made available to computers on wifi. This enables video playback to continue, even when an individual's available bit-rate is reduced by users sharing a connection. These profiles will also cater for similar bit-rate restrictions on public wifi connections.

It is clear that if four users on a single 20Mbps connection were all streaming in "HD", then their picture quality would be fractionally and almost imperceptibly reduced to fit on large screen TVs, and perhaps more drastically reduced on smaller devices.

Post deleted by RobertoS 01:36

Edited by RobertoS (Fri 02-Sep-16 01:36:17)

Good point, though I am not so sure about the just noticeable bit for 50fps stuff as only a couple of modes seem to do that. Accepting of course that a lot of stuff is only 25fps to start with anyway.

https://github.com/get-iplayer/get_iplayer/wiki/mode...

There always were many profiles to choose from and if you make get_iplayer verbose you can see bitrates listed as below.

I guess these are estimates of network rates, downloaded files are a bit less, but dividing file size by duration can still be > 5mbit. By default on hls streams. get_iplayer also rips the codecs out of the ts to make mp4, so that also reduces the size slightly compared to the sizes you can see listed with get_iplayer -i [url/pid]

INFO: Found mode hlshd1: (gip_hls_iplayer_2341) hls h264 1280x720 2341kbps stream (CDN: akamai_hls_open/10)
INFO: Found mode hlshigh1: (gip_hls_iplayer_778) hls h264 640x360 778kbps stream (CDN: akamai_hls_open/10)
INFO: Found mode hlslow1: (gip_hls_iplayer_387) hls h264 512x288 387kbps stream (CDN: akamai_hls_open/10)
INFO: Found mode hlsstd1: (gip_hls_iplayer_497) hls h264 640x360 497kbps stream (CDN: akamai_hls_open/10)
INFO: Found mode hlsvhigh1: (gip_hls_iplayer_1460) hls h264 832x468 1460kbps stream (CDN: akamai_hls_open/10)
INFO: Found mode hvfhd1: (gip_hvf_iplayer_5380) hls h264 1280x720 5380kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvfhd2: (gip_hvf_iplayer_5380) hls h264 1280x720 5380kbps stream (CDN: mf_akamai_uk_hls/1)
INFO: Found mode hvfhigh1: (gip_hvf_iplayer_956) hls h264 704x396 956kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvfhigh2: (gip_hvf_iplayer_956) hls h264 704x396 956kbps stream (CDN: mf_akamai_uk_hls/1)
INFO: Found mode hvflow1: (gip_hvf_iplayer_340) hls h264 384x216 340kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvflow2: (gip_hvf_iplayer_340) hls h264 384x216 340kbps stream (CDN: mf_akamai_uk_hls/1)
INFO: Found mode hvfsd1: (gip_hvf_iplayer_3043) hls h264 960x540 3043kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvfsd2: (gip_hvf_iplayer_3043) hls h264 960x540 3043kbps stream (CDN: mf_akamai_uk_hls/1)
INFO: Found mode hvfstd1: (gip_hvf_iplayer_551) hls h264 512x288 551kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvfstd2: (gip_hvf_iplayer_551) hls h264 512x288 551kbps stream (CDN: mf_akamai_uk_hls/1)
INFO: Found mode hvfvhigh1: (gip_hvf_iplayer_1759) hls h264 960x540 1759kbps stream (CDN: mf_limelight_uk_hls/2)
INFO: Found mode hvfvhigh2: (gip_hvf_iplayer_1759) hls h264 960x540 1759kbps stream (CDN: mf_akamai_uk_hls/1)

Edit: link to point at a specific table

Edited by deleted (Fri 02-Sep-16 17:15:55)

In a black or white world your situation would apply, but as pointed out adaptive bit rates may mean people can manage to stream without issues, hence why I was wanting to do some actual experimentation with TV connected devices

Tongue in cheek replies in a technical based topic, when not flagged explicitly as such are like a speaker making a slightly inappropriate joke, i.e. it might be Ok at a dinner where everyone knows them personally, but when broadcast to a much wider audience it becomes inappropriate.

i.e. what works when talking face to face does not always work in text based medium.

In reply to a post by MrSaffron:

That answer would mean taking a known set of hardware to various places to test it and confirm everything.

Plenty of other people are seeing time based variations with Virgin Media, but it may be that your area does not have this as its so busy, or there may just be something wrong with the cable connection

My area has never shown signs of being busy, not at least until the sudden change in TBB results, way back in May. Up until that point I consistently got my VM advertised speed on TBB. That is how this thread started, me wondering what caused the change. Since nothing changed on my premises, in terms of kit, and nothing changed at the TBB end, I can only assume there has been a VM network change. I know that re-segmenting has occurred and new Cisco or Asus cards have been installed ready for speed upgrades. Maybe as part of that my street ended up as part of a busier area, although I didn't think that VM architecture was flexible enough for a node change.

Obviously still a puzzle. Nothing is affects my quality of service, but we all hate unanswered questions don't we?

In reply to a post by RobertoS:

I just re-read that BBC Video Factory article.

It is clear that if four users on a single 20Mbps connection were all streaming in "HD", then their picture quality would be fractionally and almost imperceptibly reduced to fit on large screen TVs, and perhaps more drastically reduced on smaller devices.

The way I read that was slightly different.

What you've written suggests "adaptive bitrate streaming" means that micro-adjustments are made to encoding on the fly, to make small adjustments to the bitrate.

I believe that it means the iPlayer app can adapt the bitrate dynamically, but can only do so within the confines of a set of pre-determined encodings (so encoding isn't done on the fly). This article seems to say that.

In this way, the iPlayer app (on the STB) would stream the HD variant (1280x720 @ 5.2Mbps), but adjust down to an SD variant (eg 960x540 @ 2.9Mbps) in mid-stream if throughput (or CPU, or memory) became challenging.

I've certainly seen the iPlayer app start on a terrible quality at first, and adjust upwards to a decent quality after 10-15 seconds, so I assume it can do that downwards, and at any time.

If I'm right with that ... then I imagine 4 people trying to stream HD on 20Mbps would likely (after adjustments) find that 3 get the stream in HD while 1 gets demoted to SD.

However, my experiments showed that the bitrate is also variable, averaging out at just over 5 Mbps. That averaging might allow 4 users to keep HD, just, (good outcome) or might mean that one stream keeps switching between HD and SD (not so good outcome).

It'd be interesting to see the outcome of a real-life experiment.

I thought about that before choosing what I quoted, which are the bits your are talking about.

I think the second part is where a different profile is chosen due to the receiving kit specification, but the first is about lowering the bit rate within the chosen profile. Otherwise the mention of adaptive bit rates is meaningless, as it would simply be a question of the bitrate fixed by the profile.

I accept I could be wrong, but that was my logic.

did you read the thread I linked to?