Solution:
Root Cause: TCP receive window auto-tuning is restricted or disabled Windows uses Receive Window Auto-Tuning to grow the TCP receive window dynamically, which lets a single connection fill the available bandwidth-delay product on fast or high-latency links. Network 'optimizer' scripts, old tweak guides and some security tools set the level to disabled, highlyrestricted or restricted. The window then stays small, so a single stream cannot reach the plan speed even though the link itself is healthy. Multi-connection tests can still look fine, which makes this fault confusing. The default, normal, is correct for almost all networks. Disabling auto-tuning is only a diagnostic step for rare cases where an old router or firewall mishandles TCP window scaling.
Diagnostic Verification: Run netsh interface tcp show global in an elevated terminal and read the line 'Receive Window Auto-Tuning Level'. Anything other than normal on a fast connection is suspect. Run Get-NetTCPSetting -SettingName Internet | Select-Object SettingName, AutoTuningLevelEffective, CongestionProvider to confirm the effective level and congestion provider. Compare a single-stream download with a multi-stream speed test. A big gap between them suggests a limited TCP window. Step-by-Step Fix: 1 Open an elevated terminal:
Right-click Start and choose Terminal (Admin). 2 Restore the default auto-tuning level and related settings:
Run the commands below.
Copy
powershell
netsh interface tcp set global autotuninglevel=normal
netsh interface tcp set global rss=enabled
netsh interface tcp set global ecncapability=disabled
netsh interface tcp set heuristics disabled
3 Check the congestion provider:
Run netsh interface tcp show supplemental. On a current Windows 11 build the Internet template normally uses CUBIC. If a tweak changed it, restore it with the command below.
Copy
powershell
netsh interface tcp set supplemental template=internet congestionprovider=cubic
4 Retest:
Restart the computer, run a speed test, and compare the result with another device on the same network. 5 Use the diagnostic fallback only if normal still performs poorly:
Run netsh interface tcp set global autotuninglevel=disabled, retest, and if there is no improvement set it back to normal. If disabling helps, update the router firmware, since the router is likely mishandling TCP window scaling. Prevention & Long-Term Monitoring: Do not apply registry or netsh 'speed boost' tweaks without a measured baseline. Record netsh interface tcp show global output after a clean install so you can compare later. Update router firmware regularly to avoid TCP window scaling problems.