@mkostersitz 

Hi,
You can refer to the following instructions to learn why the wireless HB610 has slower speeds than the HB810 or the wired HB610. (This article applies to any kind of mesh or extender system.)
Why is my satellite Deco unit not getting the same speed as the main Deco?
 

For the reported phenomenon, please provide the following information for further analysis.
  1. What are the download and upload speeds from your ISP? What are the download and upload speeds from the HB810, wired HB610, and wireless HB610?
  2. Does the HB610 have a good Wi-Fi signal from the HB810? Or is it connected to the wired HB610?
  3. Have you connected a PC to the HB610/HB810 via an Ethernet cable to test the download and upload speed? Have you tested the speed on www.speedtest.net?

Best Regards

  @Solla-topee 

I am clear that the speed will be different from the HB810 but I would expect the speeds to be symetric and not highly asymetric. All speeds are measured using the SpeedTest UX on the device. I also compared with a laptop plugged into each device and the difference is neglibable.

HB810 Ethernet Internet connection: 

Down: 5613.26 Mbps

Up: 5344.47 Mbps

HB610 Wireless connected to HB810 (2.4, 5 and 6 GHz): 
Down: 814.26 Mbps

Up:1635.21 Mbps

HB610 Wired plugged into a GigaBit Switch port:

Down: 940.61 Mbps

Up: 936.24 Mbps

Mike

  @mkostersitz 

Hi,
If you swap the locations of the two HB610 nodes, does the same HB610 experience the same phenomenon? Do the HB810/HB610 work in AP mode?
Best Regards

  @Solla-topee Both are set as Mesh nodes, swapping them does not make a difference. I also went and hooked into the router directly and I am seeing the same asymetric behavior on the 1Gbit ports but not on the 2.5Gbit port.
When I turn on QoS and move devices into the CPU bound high priority path they get the maximum possible speed setting for High which is 60 %. 
It seems there is a non CPU bound path without QoS on this specific SOC that results in this asymetric behavior, the mesh configuration makes it more pronounced I guess.

I have done a bunch more troubleshooting and 

- Confirmed wired and wireless clients show the same asymmetric ratio -> not an AP or driver issue, has to be something common to all traffic through the router
- Enabling QoS (Client Acceleration) and moving the test device into the High Priority group improved download from a ~40-60% penalty to ~80% of upload speed
- Working theory: the hardware NAT/flow-offload fast path on this platform has an asymmetric limitation on ingress (download) traffic. Enabling QoS forces per-flow classification decisions, which pushes traffic off the fast path and onto full CPU-based forwarding - this looks like it corrects the asymmetry as a side effect of the path change, not because of QoS shaping itself
- The priority-tier bandwidth sliders seem to have hidden floors: pushing High down to its minimum (5%) still caps Medium at ~40%, which suggests Low has a hardcoded reserved floor somewhere around 55-60%, presumably to protect unclassified/default devices from starving
- The High Priority group is capped at 16 devices. Still confirming whether Medium tier alone (or simply having QoS enabled at all, independent of tier) produces the same download fix - that matters a lot for whether this scales to every device in a household
- This unit is provisioned through Ziply's own ACS/Aginet firmware channel, separate from TP-Link's standard consumer update path. I've seen references elsewhere to a beta firmware for the retail Deco BE85 that may address a related issue, but nothing has reached the Aginet/Ziply SKU yet

Questions for the community:
1. Has anyone else seen this specific wired-LAN, 2.5G-port, download-only degradation on a BE85 or an ISP-rebranded variant (Aginet or otherwise)?
2. Does the BE85 beta firmware actually address an ingress/NAT asymmetry issue, or is it targeting something else? I've seen separate threads about WiFi throughput and NAS file-corruption bugs that may be unrelated.
3. Does anyone know the hardware NAT/offload architecture on this platform well enough to confirm or refute the "QoS forces traffic onto the CPU path" theory above?
4. Has anyone gotten an ISP-branded/rebranded unit onto TP-Link's beta update channel, or is going through the ISP's own support the only path?

Happy to share more detail on methodology (speed test setup, port-swap results, etc.) if useful for anyone chasing something similar.

  @mkostersitz 

Hi,
You can view the Deco BE85's pre-release firmware release notes to see the changes in the firmware, and the download-only degradation is not a known issue on the Deco.

Since your HB810 & HB610 are ISP-specialized models, the parameters may be different from the generic router. It's suggested that you contact your ISP for further assistance.

Best Regards