Fibre 900 speed

I recently upgraded to BT full fibre 900. Since install, speed only reaches 600/650. BT tests from their end show no problems. I have a Realtek 2.5GBe card and have tried a few optimum settings that I found online but speed still down. I have used thinkbroadband, BT wholesale, Fast, Ookla testers, all with similar results.

Often or not these problems are the machine running the test rather than the internet connection.

Have you tried testing with another device - ideally try with a different OS/platform - like Linux/Mac if testing using Windows?

In reply to a post by tmarsh12345:

I recently upgraded to BT full fibre 900. Since install, speed only reaches 600/650. BT tests from their end show no problems.

Assuming you are on windows, which third party security/AV tool have you installed, and have you tried removing this and using Windows Defender alone? You should get to the 900 Mbps easily.

Is this using the supplied BT Smart Hub 2 ?
Upload Speed ?
Machine spec ? - As to use broadband this fast, you need a NIC with hardware offload - have you tried using the onboard ethernet card as opposed to the 3rd party one ? (the internet will not max a 1Gbit ethernet connection).

In reply to a post by teeerex:

As to use broadband this fast, you need a NIC with hardware offload - have you tried using the onboard ethernet card as opposed to the 3rd party one ?

Wot? I've used the maximum 942 Mbps on gigabit ethernet between my PCs and other devices (e.g. NAS) for years, when broadband was upto 16 Mbps on ADSL2+

Connect the BT Smart Hub 2 they provided.
Login to your BT account online and run a speed test.

It runs a speed test direct to a tester built in to the BT hub that rules out all of your devices, wired and wireless.
It's 100% the best test you can run.

If that shows low speed then the problem is the line/BT.
If it shows full speed then the issue lies in something beyond the Hub, i.e your kit.

After that people could advise further on diagnosing your kit but let's work out if your hub is receiving the full speed 1st.

I've recently tried getting a Speed Test via logging in to my BT Smart Hub 2, only to get the response that the facility is no longer available.
BT went on to say that Speed Test is available through the BT App, so I downloaded that to my iPhone and was able to run the full test and see the results. I don't know if the equivalent App is available through the Android App Store.

Edited by dtinstaffs (Sat 21-Jan-23 17:45:19)

Just tried through the router and it said the same, web based speed test no longer available.

However, tried the test through the My BT app. It ran the test and gave results both for the speed to the hub and the speed to the device. Speed to the hub showed 947/119 so pretty much there (other devices are on the network and in use so not surprised it isn't higher). Device speed was much lower as connected through wireless but it does suggest the test does show the raw speed of the connection from the hub.

the OP has specified they are using a 3rd party 2.5G card - has it been setup correctly, what is their cpu speed and has the hardware offload been enabled on the card, otherwise it WILL struggle to reach gigabit speeds if it has been disabled. Just because your setup works out the box - doesn't mean the OP's will - thus the questions and asking the OP to try the standard 1Gig port - as I agree 100%, it 'should' work without any issues.

In reply to a post by teeerex:

the OP has specified they are using a 3rd party 2.5G card - has it been setup correctly, what is their cpu speed and has the hardware offload been enabled on the card, otherwise it WILL struggle to reach gigabit speeds if it has been disabled.

20 years ago with a thinkpad T41 using a Pentium M CPU, I could easily manage 940 Mbps over a GigE port, with no such thing as hardware offload. A misconfigured 2.5 GbE card should dump load on the CPU, but any CPU in the last 10 years should be able to handle GigE surely?

We aren't talking 10 or 40 Gig here.

It’s typically all the security bloat and third party AV rubbish that folks think they need that messes it up under Windows.

Boot same box under Linux and it magically passes full rate.

Heaven help them if they wanted to fill a 2.5, 5 or 10 GbE pipe!

Linus tries (and fails) to break an *old* M1 Mac Mini with multiple 10 GbE 🤣

In reply to a post by jchamier:

20 years ago with a thinkpad T41 using a Pentium M CPU, I could easily manage 940 Mbps over a GigE port, with no such thing as hardware offload. A misconfigured 2.5 GbE card should dump load on the CPU, but any CPU in the last 10 years should be able to handle GigE surely?

I have nine-year-old Celeron N2820s that can also fill a gigabit without problems (using iperf3). But they are not running Windows.

Sorry for only coming back now, thanks for all the replies. I set up dual-boot Windows and Linux mint and have just tested at 940 in Mint so it looks like it's a Windows thing. I've googled optimim settings for the 2.5GB card so I think that's OK. Windows averages 600 which is well below the (BT) guaranteed speed and sometimes it's as low as 400. Upload speed is around 110 which appears to be very good. So I'm not sure if the speeds I'm getting are actual speeds or a false reading, is that possible? I'm happy enough with the speeds but I'm noit keen on paying for 900 if I'm not getting it.

Should have said I'm using the latest smart hub and have run tests direct from BT.

In reply to a post by tmarsh12345:

Sorry for only coming back now, thanks for all the replies. I set up dual-boot Windows and Linux mint and have just tested at 940 in Mint so it looks like it's a Windows thing. I've googled optimim settings for the 2.5GB card so I think that's OK. Windows averages 600 which is well below the (BT) guaranteed speed and sometimes it's as low as 400. Upload speed is around 110 which appears to be very good. So I'm not sure if the speeds I'm getting are actual speeds or a false reading, is that possible? I'm happy enough with the speeds but I'm noit keen on paying for 900 if I'm not getting it.

The speeds you are seeing are actual speeds.

The Mint test proves that BT are delivering what you are paying for.

If your Windows machine isn't sufficiently powerful to handle this speed of traffic, that is a problem with your machine and you can't blame BT for that. (You can blame Microsoft).

You could downgrade the Internet link to 300M or 500M and be perfectly content with that - but if you're already in a contract for 1G that might not be possible. (Personally I think that hardly anyone can notice the difference between 300M and 1G in normal use, i.e. without using a Speedtest site, especially on a wifi-connected device)

In reply to a post by tmarsh12345:

I'm happy enough with the speeds but I'm noit keen on paying for 900 if I'm not getting it.

If you're getting 900+ in Mint then you have no issue with BT. The problem is with your installation of Windows.

If you use third party security / antivirus / anti-malware software you may need to fully remove this, and locate any "completely remove" tool on the vendor website. Then try some of the tips at https://www.speedguide.net/ which attempt to repair some of the MS settings that third party AV tools mess up. (as even some new computers come pre-loaded).

Some of this security garbage is run by companies that are finding it almost impossible to make money, so can't invest in updating for new versions of Windows, as most corporates have abandoned third party security software.

Symantec was a leader in this, they sold to Broadcom and now Broadcom have sold that part on.... they realised they'd bought something crazy.

You need defence in depth, which means a) checking email (free if you use GMail, Yahoo, Outlook/hotmail or similar) and b) adblocker or similar in your browser, and then let the operating system's built in tools protect the operating system.

In reply to a post by candlerb:

If your Windows machine isn't sufficiently powerful to handle this speed of traffic, that is a problem with your machine and you can't blame BT for that. (You can blame Microsoft).

Ryzen i5 3600 with 32gb ddr4 ram and an nvme driove so probably OK. I'm using avast free. How does av limit speed?

In reply to a post by tmarsh12345:

How does av limit speed?

It modifies the Windows operating system in a way that MS never intended, and the software written by the small (compared to MS) number of staff company is not of the same standard, or tested to as many variables. Seen across all the consumer products, Kaspersky, Norton, AVG, Avast etc, and known for years by those whom run their own networks (copying large files / photos between machines) and now more obvious as internet speeds increase over 200 Mbps.

In reply to a post by jchamier:

In reply to a post by tmarsh12345:

How does av limit speed?

It modifies the Windows operating system in a way that MS never intended, and the software written by the small (compared to MS) number of staff company is not of the same standard, or tested to as many variables. Seen across all the consumer products, Kaspersky, Norton, AVG, Avast etc, and known for years by those whom run their own networks (copying large files / photos between machines) and now more obvious as internet speeds increase over 200 Mbps.

Interesting. Didn't realise that. I do a re-install of windows every now and then so will try without avast. So is it windows defender that I would use instead?

If you can do a fresh / clean slate Windows build all the better.

Don't install Avat. Leave it otherwise vanilla and re-run all your speed tests...

In reply to a post by Pheasant:

If you can do a fresh / clean slate Windows build all the better.

Don't install Avat. Leave it otherwise vanilla and re-run all your speed tests...

Too late!!! I'm now getting around 700 so looks like av is the culprit. I'll keep an eye on it to see if it improves. Definitely a fresh install soon.

Personally I think that hardly anyone can notice the difference between 300M and 1G in normal use, i.e. without using a Speedtest site, especially on a wifi-connected device

This is I think generally true. However, the difference in upstream between the packages is significant and I am considering staying on 1Gb when my contract is up in March because of the faster upstream even though I generally don't get more than about 300-400Mb in the downstream due to using wireless on the majority of devices.

Avast has probably still got its hooks in, even if it’s disabled. A fresh install will remove all the changes Avast has made in the system and you ought to get the whole nine yards, sorry 900 Mbps, throughput.

I was experiencing similar drops in speed and removed my AV (Avira) leaving just the default win 10 / win 11 defender. Now back to 900~mbps so yes AV does effect dl speed cheers

Chuck in Malwarebytes, it's brilliant and has barely any impact on PC and internet performance

In reply to a post by Grimers:

Chuck in Malwarebytes, it's brilliant and has barely any impact on PC and internet performance

On macOS it runs the cpu heavily when at 900Mbps (recent cpu using multiple cores). So I suspect on an older cpu it may bottleneck. When downloading on my connection MacBook ends up running with fans audible quite loud which is disappointing due to malware bytes load on cpu .

Interesting, I have not found that on any Windows PCs and Macs I have come across, I think it may just be your particular MacBook...

More likely how our Malwarebytes is configured, I do have it set to use High CPU for example. There's a balance of security vs usability and rarely is 900Mbps being downloaded for continuous periods. There's likely ways to mitigate. The deployment is via profile inc config, and I can't be bothered adjusting it significantly as it would impact the estate. Largely it is configured out of the box, with exception, the scan frequency is reduced as its operating with the agent active, loaded onto a machine that is a fresh image. Hence, the agent is monitoring anything added/removed, which again is bounded by security mechanisms set by privilege. Generally my machine is outside the scope anyway, I have tools deployed *should* I use it for system access, but that would mean loading on more profiles for monitoring of device etc. I leverage Citrix VDI, so this is the middle ground and allows me to see how sw performs on a BYOD device (which is approved), where VDI is unworkable (rarely).

EDIT: Config was ratified by their approved consultancy, leveraging Jamf Pro. I did raise the CPU issue a while back, and was informed it is set to scan each file as it is "saved." What this means in practice is if I download a single large file, it doesn't scan until the end. If I load something such as a one drive and sync locally, with say 10k files, it will spin up the cpu as each file is saved, this is when it can use lots of CPU when lots of files are dropping all at once, at high speed.

One Drive is configured to adapt number of simultaneous connections to balance the end user experience. At faster speeds there's a greater number (provided there is not retransmission / packet loss).

Edited by ukhardy07 (Mon 06-Feb-23 19:18:10)

Well, your previous post made it sound like it behaves like that on every instance of MacOS

Makes my old laptop very slow on start up with high CPU load.

Well, that is probably because it's an old laptop. I've installed Malwarebytes on so many PCs and never had one complaint...

A previous posting about this showed many people having issues with it.

Would be interested to see this "post".

Safari: 932/108
https://www.speedtest.net/result/14463862014.png

Chrome: 710/108
https://www.speedtest.net/result/14463881881.png

Same server, same time, reproducible.

Been like this consistently for a few years. OP I just put this down to the browser / hardware maybe.

In reply to a post by oldskool:

I just put this down to the browser / hardware maybe.

Seems Chrome can have issues... I suspect on all the Operating Systems it supports.
https://support.google.com/chrome/thread/3946425/slo...

In reply to a post by oldskool:

Safari: 932/108
https://www.speedtest.net/result/14463862014.png

Chrome: 710/108
https://www.speedtest.net/result/14463881881.png

Same server, same time, reproducible.

Been like this consistently for a few years. OP I just put this down to the browser / hardware maybe.

Browser-based speed tests take a variable (chunky) amount of system resources to execute.

Therefore the results can vary, depending on how much grunt the base system has, and the optimisations (or not) of the browser and the particular speed test.

Interesting observation. I use an older iMac (still 3.4Ghz quad core i5 with SSD etc - so fine).

In Chrome
Speedtest with my Google profile logged in an syncing: 700Mbps
Speedtest as a guest (barebones) 930Mbps

So I thought, must be an extension, disabled them all, still 700Mbps
Tried a brand new profile from a different account, still 700Mbps

So what on earth are Google doing that would affect the transfer speed.

Hope its not sniffing / traffic analysis.

Edited by oldskool (Wed 15-Mar-23 17:58:30)

It might be worth trying the Speedtest app available from the Microsoft Store (free). This eliminates any browser profile issue.
There is also a command line (CLI) version available.
https://www.speedtest.net/apps/cli

In reply to a post by chriswillsher:

It might be worth trying the Speedtest app available from the Microsoft Store (free).

Also in the Mac App Store for the poster with a Mac
https://apps.apple.com/us/app/speedtest-by-ookla/id1...

For Macs: also have Activity Monitor open in the dock, while you're running the speedtest. Seeing the CPU jump to 100% proves that it's the laptop/browser/OS combination which is the bottleneck, not the service from the ISP.

On my dual-core i7 Macbook Pro (2015 model):

* In the browser, my CPU load goes to 100%. This is despite me only having a 300/50 FTTP connection.

* But if I use the Speedtest standalone app, CPU load is much lower and it shows the full speed available.

In reply to a post by candlerb:

For Macs: also have Activity Monitor open in the dock, while you're running the speedtest. Seeing the CPU jump to 100% proves that it's the laptop/browser/OS combination which is the bottleneck, not the service from the ISP.

On Windows the same for Task Manager on the Performance tab. The CPU should not spike either unless some third party software has broken the path, usually third party misguided security software.

In reply to a post by Grimers:

Interesting, I have not found that on any Windows PCs and Macs I have come across, I think it may just be your particular MacBook...

As a test, I have MacOS Ventura running on an HP Zbook 14U which has an 8th Gen i5 and 16Gb ram so quite old.

When downloading via HTTP it did ramp up the CPU load as seen using Intel power gadget and temps increased from 45'c at idle to 65'c which wasn't enough to hear the fans at this point before the ISO had completed.

This was using Brave so CPU usage could be browser dependent.

On an M1 or M2 based system I don't think it would be noticeable at all and impact is much less using Safari e.g. temps do not go above 60'c.

Edited by smouty (Mon 20-Mar-23 13:44:59)

In reply to a post by smouty:

In reply to a post by Grimers:

Interesting, I have not found that on any Windows PCs and Macs I have come across, I think it may just be your particular MacBook...

As a test, I have MacOS Ventura running on an HP Zbook 14U which has an 8th Gen i5 and 16Gb ram so quite old.

When downloading via HTTP it did ramp up the CPU load as seen using Intel power gadget and temps increased from 45'c at idle to 65'c which wasn't enough to hear the fans at this point before the ISO had completed.

This was using Brave so CPU usage could be browser dependent.

On an M1 or M2 based system I don't think it would be noticeable at all and impact is much less using Safari e.g. temps do not go above 60'c.

Rarely go much above 43'c on an M1 MBP - whether its browser or app. speed tests. So not even breaking a sweat on efficiency cores. Theres barely a ripple on Activity Monitor on any Ookla based tests.

OpenSpeedTest via browser using a slightly more perceptible amount of CPU around 10% (sytem+user)

It is crazy just how good the M1/M2 Macs are for the price.
Here is my result now I have 900/900 at home.
This was done using Brave which is now native Arm based.

Swish Speedtest

I have a M2 Mac. However, I could get 900 down on a 2013 Trashcan MacPro. When Zen was not messing things up. 100+ up on an "standard" 900/100 line.

Nice to know that it's symmetrically good.

900 Mbps symmetric is just a 1 Gbps connection, as used in corporate buildings for over a decade. A Pentium M laptop (ThinkPad T40) could easily saturate a 1 Gbps connection. I would have thought a Raspberry Pi could do the same. Nothing special there about your M1/M2 processor.

I wasn't specifically talking about using Gbit which I've been using in my house for at least a decade

I'm sure an old gen laptop could do it at the time but I think it would struggle to do it on Win10/11 in a browser test with all the bloat.

A lightweight linux install maybe but you're are right that there is nothing special about it and any modern OS on a fairly recent CPU should handle this easily but it is not the case.

In reply to a post by smouty:

I'm sure an old gen laptop could do it at the time but I think it would struggle to do it on Win10/11 in a browser test with all the bloat.

I have a core2duo laptop with 4GB RAM and Win10, and it can handle it?? Most of the "bloat" is third party security software, usually erasing that makes a big difference. Or, if corporate machine, third party poorly written disk encryption. Yes if 2GB or less of RAM, Win10 is going to struggle, page to disk and not be able to run an up to date browser.

I sadly don't have an internet connection fast enough to try.