Running 3.213,Adgaurd Home v0.106.3(DNS servers are Nordvpn), All services use NordVPN UDP with vpn policies and killswitch engaged on a Cable modem with 800 Mbits down and 25 Mbits up
So I used dslreports to speed test with vpn client running and got between 75-99 Mbits download and 2-25 Mbits upload. Bufferbloat was a C it said. I tweaked the SQM basic settings and to 70000 Kbits(70Mbits) download and 10000 Kbits(10Mbits). Queue Discipline is CODEl with piece_of_cake.qos. Speed was about the same but I got A+ for all the test scores. (This the best I can do?)
So my question is should basic settings be based on speed using VPN or without? If I switch to wireguard will I need to change the basic setting every time to accommodate the added bandwidth.
I would suggest it depends about your situation, the most important step is the overhead, but if you use vpn policies aswell a too big overhead might also not something what you want.
I have been testing this for the fun, but I have seen some crazy latency issues when a link suddenly gets alot higher bandwith I guess best is to recommend, is to create a SQM policy for each individual connection or device group with in mind that you use a vpn more as a static connection and not to use it dynamicaly to enable/disable, it seems sqm is sensitive on this.
as your example, I think you should be fine, I always make a small calculation and substract 8% from each device group… for example my TV I give the highest priority, my tv is isolated on its own network, and then with the next group my lan I substract it again from my tv data, this way I always create some overhead and also some kind of priority tree.