To test this I disabled HFO and tested SQM on MT6000. I have a 1 Gbit / 25 Mbit rated cable service (yes ISP is very asymmetrical).
Test with SQM from a couple months ago: https://www.waveform.com/tools/bufferbloat?test-id=fe841d79-96bf-4de2-bf27-17a34f5fff03
rc5 test with SQM: it's barely working the CPU cores (watching them with htop): https://www.waveform.com/tools/bufferbloat?test-id=fc6df2a2-a99f-4424-b0dc-c566bf24ed4e
EDIT: rc5 test with SQM + Packet Steering (all CPUs)/ RPS 128 enabled: https://www.waveform.com/tools/bufferbloat?test-id=abfaccee-47d0-4c45-a1cd-60f4593f91a3
EDIT2: ran it again on rc5 with my internet more quiet, SQM + PS seems fine:
https://www.waveform.com/tools/bufferbloat?test-id=905160ee-92c5-49df-a2f0-90ee2da3834c
I'd say SQM is ok, not seeing much regression with SQM + packet steering. Maybe just my congested internet I'm not the only one active at my home right now.