Netgear R7800 and LEDE/OpenWRT

[avtella]

If bufferbloat is the main issue and if you don’t really need that high speeds in 5Ghz I think you will be fine. I suppose you could try replacing the WiFi firmware in DD-WRT using latest ones from kvalo’s site. Used to be pretty easy to do in OpenWRT/LEDE, I would assume the same for DD-WRT.
https://github.com/kvalo/ath10k-firmware/tree/master/QCA9984/hw1.0/3.6.0.1


Edit: just tried DD-WRT again with a friend who has similar issues as I did with bufferbloat on Centurylink, still a no contest SQM Piece of Cake consistently still requires slightly less bandwidth reservation and still gives an A/A+ vs B average and on accession A with FQ_Codel on DD-WRT.

 

[drawz]

OpenWrt now supports software flow offloading, which isn't quite as fast as hardware NAT acceleration, but it's pretty close. With a fast CPU, like the one in the R7800, it's a non-issue with or without SW flow offload. That is until you enable QoS, which isn't compatible with SW flow off load or HW NAT. OpenWrt is also starting to support HW NAT for some chipsets - I think Mediatek only right now, but likely more to come in the future.

OpenWrt is the clear leader in adding support for new bufferbloat mitigation techniques. The guys seeking to solve bufferbloat issues do all their work in OpenWrt, so it's only natural. SQM with cake (or piece of cake) is probably the best option out there.

For what it's worth, I think DD-WRT now uses OpenWrt kernels, although it is usually several versions behind. I personally don't like DD-WRT's rolling release schedule with no defined and supported releases. OpenWrt has defined and supported releases that are becoming more regular, but they also offer nightly snapshots if you wanted to be on the bleeding edge. I also find that DD-WRT is more likely to have regressions than OpenWrt. DD-WRT's 15 year old interface doesn't give it any points either!