tbb speed test

I'm in the middle of a line investigation by PN on my line. I am puzzled by the results I get on the tbb speedtest.

First run, single thread download around 4-5Mb
Multi download around 26Mb

Second run, single thread slowly climbs to around 20Mb
Multi to around 32Mb

Third run, single thread at 36Mb, multi at 36Mb

??? Why does the test improve and how many more should I do?

Different forms of congestion make for different patterns in the graphs.

Peak hours tend to be centred on 8-9pm; if congestion is encountered, it tends to be around those hours now.

The answer to the latter is ... as often as necessary to show when the problem occurs, and when it doesn't.

Still not understanding why repetitive testing is increasing my 'speeds' - it is almost as if the server is 'getting used' to my calls.

"to show when the problem occurs, and when it doesn't" - this phenomonen occurs on every speed test at tbb, but not on other sites.

No idea - there is no getting used to a user algorithm used on the server

The best idea would be to post links to all three results, so we can see both the graph shapes and the times. Times that may be related to streaming by others of something popular perhaps.

It won't be the Paralympics opening ceremony, that's just starting on CH 4. Maybe a combo of Bake Off and Murray tennis? AIUI Bake Off has finished. Murray is in the third set.

Edited by RobertoS (Wed 07-Sep-16 21:07:48)

RS - I could do, but it happens every time I run a test and has done for the last few months, so it ain't any 'event'. The other puzzle is why the speeds increase during the duration of each test. Other testers don't do this

Is there a guide to how I save and post the graphs?

Run the tests from your account. Go to profile and you will see a list of tests. Click on te test and you will see links, copy and paste those for three consecutive tests.

It is also worth having Task Manager running at the same time, do a screen shot, host it and then provide a link to that.

Where is 'profile'?

Click on the MAIN SITE link ... then in the top right corner, MY PROFILE.

The question is, do you have an account there too? you need to register there, same User-name and password if you wish.

Edited by MHC (Wed 07-Sep-16 21:37:19)

Test1:
http://www.thinkbroadband.com/speedtest/results.html...

Test2:
http://www.thinkbroadband.com/speedtest/results.html...

Test3:
http://www.thinkbroadband.com/speedtest/results.html...

Is that what you want?

Much simpler - there is a share link given in the test app - just copy and paste that URL as it is unique to each test.

You are running the flash version, so worth checking you get same results via http://labs.thinkbroadband.com/speedtest that does not use flash. If the new tester give better results in terms of the 'slope' and 'improving each time' then it points to an issue with your flash install (given the pace at which flash is getting security holes fixed, there being weird side effects is not unlikely)

Interesting. Wireless but only a couple of feet from the router. (No ethernet on the machine).

Lab 1.
Lab 2.

Flash test.

Download speed achieved during the test was - 54.59 Mbps
For your connection, the acceptable range of speeds is 39.14 Mbps-55.91 Mbps .
Additional Information:
IP Profile for your line is - 55.91 Mbps

Upload speed achieved during the test was - 11.09Mbps
Additional Information:
Upstream Rate IP profile on your line is - 20 Mbps

Lab 3.

Intriguing that the labs tests are so jagged compared to the Flash one.

I do not understand your system, Andrew. My initial post here was based on the results of the non-flash test which was the one that 'came up' from your link. I then established a 'profile' as requested, and straightaway the next speedtest call went to the flash test. Nothing to do with me!

NB Both tests give the same result, Same 'improvement' over three tests and the same figures more or less.

It appears to be your coding Andrew - selecting speedtest NOT LOGGED IN takes me to non-flash, LOGGED IN to flash.

If you look at the tests in my post, I used the labs link Andrew gave you. I discovered that maintains its own history at the moment, which doesn't appear in the main profile history. Does non-Flash history appear if you get to it your way, logged out if the Main Site? It seems you can only get to the labs history by running it again, where a link to it appears at the end.

The logged in and non-logged in swapping has been known about for a long time.

Does non-Flash history appear if you get to it your way, logged out if the Main Site? It seems you can only get to the labs history by running it again, where a link to it appears at the end.

Correct - non-flash history available (logged out) by running a speedtest.

Not too sure what you are asking, though - 'my way'?.

The logged in and non-logged in swapping has been known about for a long time.

- a link to allow choice of speedtest for 'logged in' would be a good idea, then. Most modern sites are going away from Flash for good reason.

"Your way" involves logging out of the main site, which I very rarely if ever do. My way was just to click the link Andrew gave, which I haven't bookmarked but is easily done .

Ah! Confusion arose as I do not (normally) visit the site as a 'logged in' user so I do not 'log out'!

Test1:
http://www.thinkbroadband.com/speedtest/results.html...

Test2:
http://www.thinkbroadband.com/speedtest/results.html...

Test3:
http://www.thinkbroadband.com/speedtest/results.html...

Not displaying the same increase in single thread download today. Interesting, since I have OR 'working' on my line at the request of PN. I will test again tomorrow.

EDIT: Even more confusion! Just ''logged in' and cannot get the flash test to load at all...............

Edited by mikejp (Thu 08-Sep-16 14:26:10)

The flash will be going away, but running both alongside each other for a long period of time is prudent when so many people rely on the results.

Was not aware the original test was on the non-flash, so was just wanting to double check behaviour was the same, as there is a small difference in how tbbx1 works in the two tests.

Just to add, on a browser where flash is not installed then we always present the non-flash version and this is an increasing number of people as they uninstall flash.

Edited by MrSaffron (Thu 08-Sep-16 15:28:01)

Minimal smoothing on the labs to make any 'software' interactions more visible.

Copied. The only way now I can see any 'history' is to use your link when I am logged in as Flash still does not load.

Other problem I have is that the same behaviour is exhibited again now, using the non-flash test, where the single thread starts 'slow' and on repeated tests climbs nearer to the multi-thread figure. Problem is I cannot paste more than one test from a sequence as each one tells me I am viewing 'someone else's test' (although logged in) and only lists the last test done.

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

Here is the next test
http://www.thinkbroadband.com/speedtest/results.html...

Out of interest after OR 'work' today my sync has gone from 43 to 49, although speed has not yet changed.

The non-flash will always say 'someone else' as they are not currently linked to your profile, once a site makeover happens this will all change and the moans and groans go away

On the single thread, the traditional response would be you have the RWIN badly configured, but if on the new Plusnet network it may be they have something configured wrong in their part of the network.

Your IP Profile should have changed, visible in the Further Diagnostics of the BT Wholesale Performance Test, (ignore the red instructions, just say yes), as per my post earlier, but Plusnet hold a slightly lower copy of that in your Current line speed, which they use to limit your speed. That only updates three or four times a day.

Check it now, and if it hasn't changed by tomorrow get them to update it. The auto update does occasionally fail, but the notification of a new IP Profile from BTW only arrives once. So if it has failed, it will stay that way until the next IP Profile change.

RWIN badly configured

- how do I check that?

but if on the new Plusnet network it may be they have something configured wrong in their part of the network.

- no, this phenomonen has been happening for months, well before any PN change.

RS - nae wurry! I am well onto that...

OR work now complete and so far no noise on the line and speeds improved. I have also, after much procrastination, moved my home plug circuit onto the same ring main with good improvement.

Speedtest still showing dramatic uplift of single thread speed on second test.

Still hoping someone will tell me about RWin?

Those speedtests do not look great. Congestion of some kind at play here?

I always always get basically an identical speedtest on this forum. Certainly do not get these improvements etc like yourself. I think it is just by fluke.

Fair enough, it's your test and whatever software interactions means may be a valid reason, but it does make normal connections look jagged and sometimes leads to crazy peaks way above what the line is physically capable of. Even "normal" tests burst over line speed yet the text says

The burst speed figure shows the speed your computer and connection sustained for a short period during the test.

Well that's not true - my computer can not send faster than my line - it's just that with low smoothing plus measuring using tcp you see bursts as it is expected that a normally functioning set up will drop packets - as it should, unless you have some really extreme buffer bloat/atypical tcp settings.

In reply to a post by mikejp:

OR work now complete and so far no noise on the line and speeds improved. I have also, after much procrastination, moved my home plug circuit onto the same ring main with good improvement.

Speedtest still showing dramatic uplift of single thread speed on second test.

Still hoping someone will tell me about RWin?

Though you've had the issue for longer than the recent Plusnet problems - any testing now may be misleading due to them.

On the rwin - if that were too small/window scaling was off it should really just cap you at a constant speed, but your results are stranger than that.

If possible testing using a different PC/device connected with eth to see if it's better would be the first thing I would try. If you haven't rebooted your router for ages try that. Maybe just trying with a different browser could be a start.

To see your tcp settings you could post the output of -

http://www.speedguide.net/analyzer.php

Though especially on Linux the rwin results (as is noted) are a bit meaningless.

Ta legume - I was under the impression that RWIN tweaking was a thing oif the past in Windows10?

Certainly since my line improvement I have a significantly faster 'first pass' on the single thread (it staerts at around 20 and works up to the mid 40's, then on the second pass - wham straight up to the mid 40's. Multi-thread consistently high 40's.

(Not going to talk about PN's problems of yesterday as I have sent a red-hot snotogram to Andy about the CS aspects of it....)

Test on laptop via Wireless to same router is much less varied, starting at 20'ish with a slight pick up on test 3 to the multi speed of mid 20's. i wonder if there is something adrift with my NIC on the mainframe?

Re TCP settings - Billion 8800NL - where do I find what you seek please?

In reply to a post by mikejp:

Ta legume - I was under the impression that RWIN tweaking was a thing oif the past in Windows10?

Test on laptop via Wireless to same router is much less varied, starting at 20'ish with a slight pick up on test 3 to the multi speed of mid 20's. i wonder if there is something adrift with my NIC on the mainframe?

Re TCP settings - Billion 8800NL - where do I find what you seek please?

I wouldn't try tweaking RWIN on anything in 2016.

On the NIC - unlikely, but testing with a real cable to the router may eliminate any homeplug wierdness causing it.
Windows 10 is an unknown for me, it or it in combination with any antivirus software could be involved.

The link I gave would confirm all was OK with settings, routers are not involved with TCP settings, it's between your computer and the server., though saying that routers will likely clamp the mss setting for you, but this is normal.

C:\WIN10\system32>netsh interface tcp show global
Querying active state...

Text

1 23 45 67 89 1011 1213 1415

TCP Global Parameters ----------------------------------------------Receive-Side Scaling State          : enabled Chimney Offload State               : disabledNetDMA State                        : disabled Direct Cache Access (DCA)           : disabledReceive Window Auto-Tuning Level    : normal Add-On Congestion Control Provider  : noneECN Capability                      : disabled RFC 1323 Timestamps                 : disabledInitial RTO                         : 3000 Receive Segment Coalescing State    : disabledNon Sack Rtt Resiliency             : disabled Max SYN Retransmissions             : 2TCP Fast Open                       : enabled

« SpeedGuide.net TCP Analyzer Results »
Tested on: 2016.09.10 07:42
IP address: 51.7.xx.xxx
Client OS/browser: Windows 10 (Firefox 48.0)

TCP options string: 020405ac0103030801010402
MSS: 1452
MTU: 1492
TCP Window: 66560 (NOT multiple of MSS)
RWIN Scaling: 8 bits (2^8=256)
Unscaled RWIN : 260
Recommended RWINs: 63888, 127776, 255552, 511104, 1022208
BDP limit (200ms): 2662kbps (333KBytes/s)
BDP limit (500ms): 1065kbps (133KBytes/s)
MTU Discovery: ON
TTL: 112
Timestamps: OFF
SACKs: ON
IP ToS: 00000000 (0)

In reply to a post by mikejp:

« SpeedGuide.net TCP Analyzer Results »
Tested on: 2016.09.10 07:42
IP address: 51.7.xx.xxx
Client OS/browser: Windows 10 (Firefox 48.0)

TCP options string: 020405ac0103030801010402
MSS: 1452
MTU: 1492
TCP Window: 66560 (NOT multiple of MSS)
RWIN Scaling: 8 bits (2^8=256)
Unscaled RWIN : 260
Recommended RWINs: 63888, 127776, 255552, 511104, 1022208
BDP limit (200ms): 2662kbps (333KBytes/s)
BDP limit (500ms): 1065kbps (133KBytes/s)
MTU Discovery: ON
TTL: 112
Timestamps: OFF
SACKs: ON
IP ToS: 00000000 (0)

That's OK, I wouldn't change anything despite what it says.

It seems that like Linux has for ~15 years Windows 10 does dynamic right sizing for buffers. so that's why the BDP is low - the Recommended RWINs just doesn't really matter in 2016 with window scaling and messing with anything will likely just have bad side effects for some other test.

Edited by deleted (Sat 10-Sep-16 13:06:35)

There is one setting that you could try in the advanced settings of your network adapter.
Make sure that:
Large Send Offload is set to disabled in both IPv4 & IPv6.
The setting is known to slow down modern systems due to the faster throughput.
You could also try "DNSBench" to make sure that you are not using unresponsive Name Servers.

Edited by deleted (Sat 10-Sep-16 17:35:31)

Just had another uplift in sync for 48+ to 54+ which for a 55/10 contract is not too shabby.

iron - tried that, and after a reboot there did not seem to be any sig change. Will it take time to work as in DLM-style?

In reply to a post by mikejp:

Just had another uplift in sync for 48+ to 54+ which for a 55/10 contract is not too shabby.

iron - tried that, and after a reboot there did not seem to be any sig change. Will it take time to work as in DLM-style?

If it was disabled already then np, if you have just disabled it and then rebooted then it might take a short time to sort the system but should just work.
Did you try DNSBench.?
No big deal if you did not but it can and does sort out the best Name Servers near your location.
You could also try setting the Jumbo Frame setting to 7KB MTU if you have it.
You can just put them back to what they were if you see no improvement.

"Just had another uplift in sync for 48+ to 54+ which for a 55/10 contract is not too shabby."
That is good because 54+ on a 55/10 line is as good as you can get as about 2mbps is deducted for overheads etc.

Edited by deleted (Sun 11-Sep-16 21:00:35)

Just run DNS Bench (good old Steve!) and swapped to 'better servers'. Yes, speeds improved, although it is possible the up has taken a small loss..

Thanks for the attention. I'll leave things static for a while and review after a week.

In reply to a post by mikejp:

Just run DNS Bench (good old Steve!) and swapped to 'better servers'. Yes, speeds improved, although it is possible the up has taken a small loss..

Thanks for the attention. I'll leave things static for a while and review after a week.

Cool! hope you don`t mind me giving you the tip but don`t use OPEN DNS or other name servers that refer to advert pages before doing the job.

Below is a link to the results I am getting from the TB tester, it seems ok.
http://www.thinkbroadband.com/speedtest/results.html...

Edited by deleted (Mon 12-Sep-16 00:12:53)