Model: Archer AX21

Hardware Version: V5 (US/5.46)

Firmware Version: 1.1.2 Build 20250814 rel.14122(5553)

If you found this thread because your Windows PC gets almost no download speed on this router while upload is normal, phones and TVs are fine, and nothing you've tried on the PC has helped: the router is the cause, and there is a one-line workaround.

Workaround (Windows, per machine)

Open Command Prompt or PowerShell as administrator and run:

netsh int tcp set global timestamps=enabled

No reboot needed. Reconnect to Wi-Fi and test again. On my machine this went from 0.04 Mbps down to 531 Mbps down instantly, with no other change. The setting persists across reboots. To revert: netsh int tcp set global timestamps=allowed (the Windows default).

Symptoms, so you can recognize it

Download 0.04 to 0.34 Mbps, upload 100 to 800+ Mbps, on every speed test.

Same result wired to a LAN port and over Wi-Fi, 2.4 GHz and 5 GHz, any distance from the router.

Same result with different Wi-Fi adapters, a USB adapter, and after a clean Windows reinstall.

Ping is fine, including large packets. Web pages mostly load. Anything that has to download in volume dies.

Phones, tablets, TVs, and Macs are unaffected against ordinary websites.

The same PC gets full speed on a phone hotspot, and full speed plugged directly into the ISP modem/ONT with the AX21 removed.

If any of your household is on a corporate VPN (Zscaler, etc.), that machine will look fine, because the VPN tunnel hides the problem.

What is actually wrong

The router mishandles any TCP connection that is set up without the TCP timestamps option. Windows does not send that option by default; Android, iOS, macOS, and Linux do. That is why only Windows machines are hit across the board. It is not Windows-specific in principle: on a website whose server declines timestamps (in my case a hospital patient portal), my Android phone and the Windows PC both crawl through the AX21 and both load normally on cellular. There is no client-side fix for that case; only firmware can fix it.

Things that do not help, so you can skip them

Factory reset; firmware update (the updater reports 1.1.2 as current); SPI firewall off; QoS on or off; guest network; 802.11ac-only mode; channel or band changes; on the PC, disabling IPv6, ECN, RSC, wireless power saving, or setting a static IP; replacing the Wi-Fi card; the "Enable NAT" checkbox (turning that off simply kills internet). NAT Boost is not exposed on this firmware.

Verification I did, for anyone who wants to check

Three Windows 11 machines (Lenovo IdeaPad Slim 5, a Dell business laptop, a Surface Pro), wired and wireless, three different Wi-Fi chipsets: all affected identically.

The Dell wired straight to the fiber ONT with the AX21 out of the path: 900 down / 597 up.

The Dell wired into an AX21 LAN port: 0.34 down / 877 up.

Adapter statistics during a download show almost no inbound payload arriving at the client, only ACK-sized packets, so the data is being withheld upstream of the PC.

Enabling timestamps on the client: 531 down / 492 up over Wi-Fi.

I have filed this with TP-Link US support with full details and am happy to test a beta firmware. If you have this problem on another hardware version or firmware build, please add your details below so TP-Link can see the scope.

There's a related unresolved thread from January 2024 on a V2 unit with wired gigabit Ethernet dropping to 0.18 Mbps while phone Wi-Fi tested fine, which was locked without an answer; this is very likely the same defect.