Hello everyone,

I recently replaced my old ASUS RT-AC86U with a TP-Link Archer GE550, expecting a noticeable improvement in performance and stability under heavy traffic. My main reason for upgrading was recurring bufferbloat issues on the ASUS, especially when running Stremio on a TV while gaming, which caused latency spikes during gameplay.

After installing the GE550 and placing my ISP modem in bridge mode (no double NAT), I started testing the connection using the Waveform bufferbloat test. To my surprise, with QoS enabled, the GE550 consistently scores a bufferbloat grade F, caused almost entirely by extreme latency increases during upload (hundreds of milliseconds).

I have already tried the following troubleshooting steps:

• Enabling bandwidth-based QoS

• Manually limiting upload and download bandwidth well below my actual line speed

• Prioritizing my gaming PC using Game Accelerator

• Disabling traffic analysis and monitoring features

Despite these adjustments, the issue persists whenever QoS is enabled. Upload saturation causes massive latency spikes, making gaming and other real-time applications unusable under load.

However, when I disable QoS completely, the behavior changes drastically:
the bufferbloat disappears, latency remains stable, and I consistently achieve an A+ rating on the Waveform test.

This suggests that the problem is not related to my ISP, modem, or double NAT, but rather to how QoS / queue management is implemented in the GE550 firmware. This is disappointing, as the GE550 is expected to be significantly more powerful and modern than the RT-AC86U, yet in my environment it only performs well with QoS turned off.

At this point, if this behavior cannot be improved via configuration or firmware updates, I am seriously considering returning the GE550 and reverting to my ASUS, or moving to a different solution altogether.

My questions are:

• Have other users experienced similar upload bufferbloat issues with QoS enabled on the GE550?

• Is there any plan to improve queue management or implement proper SQM (fq_codel, CAKE, or similar) in future firmware updates?

Any feedback from TP-Link staff or other users would be greatly appreciated.