This is what ChatGPPT said btw:

This is not the NIC. It’s Windows struggling with a single TCP flow.

Why Windows can be much worse on x1 than x6 (same hardware)

A speed test like TBB does two different things:
• x1 = one TCP connection trying to fill the pipe
• x6 = six TCP connections in parallel sharing the pipe

Windows in your run can’t keep one connection “open wide” (congestion window / receive window behaviour, pacing, and/or occasional loss/retransmit), so the single stream keeps backing off. With six streams, the test “cheats” around that: even if one stream is wobbling, the others keep the total high.

Linux, in your run, keeps a single flow stable, so x1 ≈ x6.

What typically causes exactly this pattern on Windows

These are the usual culprits, in practical order:
1. TCP autotuning or receive window limits not behaving well
• If autotuning is disabled/restricted, single-flow throughput collapses.
2. A filter driver messing with traffic
• VPN client, antivirus “web protection”, traffic shaper, “Killer”/OEM network suites, some firewall products.
• These can murder single-stream performance while multi-stream looks fine.
3. Offload / checksum / LSO settings interacting badly with the driver
• Less common than #1/#2, but real (especially Realtek 2.5GbE + some Windows drivers).
4. Browser-path effects
• TBB uses HTTP; the Windows browser stack + security hooks can differ vs Linux CLI tools.
• This usually affects shape and a bit of speed, but your x1 gap is big enough that I’d look at #1/#2 first.

Quick way to prove it (2 minutes, no guesswork)

On Windows, run in an elevated Command Prompt:

netsh interface tcp show global

Look for Receive Window Auto-Tuning Level.
• If it’s disabled or highlyrestricted, that can explain your x1 result immediately.

Also check if you’ve got any of these installed/enabled:
• VPN (even “off”)
• third-party AV web shield
• “network optimizer” / OEM suite

Those are classic single-stream killers.

If you paste the output of netsh interface tcp show global (just that), I can tell you straight away whether Windows is artificially hobbling the single TCP flow.

In reply to a post by adslmax:

In reply to a post by adslmax:

seb

I am not 100% sure if it was thinkbroadband speedtest server issues with 2.5Gbps link using Cat5e ethernet from TP Link Archer BE230 (BE3600) as you can see it got bottleneck with Full Fibre 1000/115 below?

https://www.thinkbroadband.com/speedtest/17682385484...

There is no bottleneck using 1Gbps link using Cat5e ethernet from TP Link Archer vx230v below:

https://www.thinkbroadband.com/speedtest/17682077117...

I just found out why......

Windows 11 throttling network issues.

Running on Linux using 2.5Gbps link using Cat5e ethernet from TP Link Archer BE230 (BE3600) with no bottleneck!

https://www.thinkbroadband.com/speedtest/17682971055...

TBB Speedtest Linux v Windows 11
https://i.ibb.co/4RKB0xCL/Screenshot.png

Have you tried temporarily changing the link speed on the Windows 11 device from 2.5Gbps/Auto down to 1Gbps when connected to the TP Link Archer BE230 (BE3600) to see if the same issue happens.

All tests running on TP Link 2.5Gbps ethernet (anti-virus disabled by Windows Defender)

Running clean install of Windows 10 no third party apps on Windows Edge

https://www.thinkbroadband.com/speedtest/17683155055...

Running clean install of Windows 10 no third party apps on Google Chrome

https://www.thinkbroadband.com/speedtest/17683157078...

Running clean install of Windows 10 no third party apps on Firefox

https://www.thinkbroadband.com/speedtest/17683158891...

In reply to a post by PCJM40:

Have you tried temporarily changing the link speed on the Windows 11 device from 2.5Gbps/Auto down to 1Gbps when connected to the TP Link Archer BE230 (BE3600) to see if the same issue happens.

I try that later tonight

In reply to a post by adslmax:

Running clean install of Windows 10 no third party apps on Windows Edge
Running clean install of Windows 10 no third party apps on Google Chrome
Running clean install of Windows 10 no third party apps on Firefox

So it seems firefox is great (it's Gecko based) whilst Edge and Chrome (Chromium based) are slightly struggling on single thread.,.
All looking better - what may have been the original issue?

Yeah Linux also using Firefox.

Last one

Running clean install of Windows 10 no third party apps on Brave

https://www.thinkbroadband.com/speedtest/17683184703...

Final test back to firefox with no other device running (told my daughter to stop using xbox for couple of minutes)

https://www.thinkbroadband.com/speedtest/17683189882...

Anyway closed this thread now. Seem my full fibre are fine. All seem ok. Case closed now.

Edited by adslmax (Tue 13-Jan-26 16:01:09)

just for fun.

linux, with 10gbit nic, to 10gbit switch then to 10gbit port on router, Router to ont is 2.5gb (since the ont is 2.5gb) both via ipv6

https://www.thinkbroadband.com/speedtest/17683199384...

windows with 10gbit nic to 10gbit and then same as the above
https://www.thinkbroadband.com/speedtest/17683202245...

interestingly the server has a lower spec'd cpu 😁 but that shouldn't matter

In reply to a post by Taras:

just for fun.

linux, with 10gbit nic, to 10gbit switch then to 10gbit port on router, Router to ont is 2.5gb (since the ont is 2.5gb) both via ipv6

https://www.thinkbroadband.com/speedtest/17683199384...

windows with 10gbit nic to 10gbit and then same as the above
https://www.thinkbroadband.com/speedtest/17683202245...

interestingly the server has a lower spec'd cpu 😁 but that shouldn't matter

Look great.

Only one problem with Aquiss don't selling openreach 1600/115 but doesn't matter as speed is not everything we need mostly. 😀

In reply to a post by adslmax:

Only one problem with Aquiss don't selling openreach 1600/115 but doesn't matter as speed is not everything we need mostly. 😀

As long as you're happy with what you have now that is the main thing, Aquiss may make 1600/115 available in the future