Yes should have mentioned I ran all tests 3x. It's likely SQM is a unique test and not 1:1 with how iperf3 would scale, just did it to see how it would turn out. Mostly I think the results are fantastic and shows this is a 900Mbps SQM target with the right settings.
The WireGuards test I ran was based off this thread: [A Wireguard comparison DB]. It was noted in the thread and github that it's a ballpark test and not fully indicitive of realworld tests so take that how you will, but still getting a little lower on average that the results posted for our target on that thread.
The results are great, although I do have concerns with the new script as it seems to require you to adjust it at any given time depending on what you're doing e.g. using WireGuard, SQM or speedtest-cli. But previously you could toggle the packet steering setting and it seemingly worked well with everything.
Yep. I gave my results in that thread too, with the best one being 972Mbps after clearing all nft rules.
I'll retest things later tonight with the latest firmware.
Right? I was surprised to see a 166Mbps improvement on my device by removing the rules. And of course you shouldn't permanently remove the rules, but I never would of guessed that they're slowing things down that much.
What would be the easiest way to incorporate your mt76 repo into my own build? I've been pulling in package/kernel/mt76/Makefile from your next branch into my repo before each build via wget but there must be a better way I think?!
Around a 50-100Mbps drop in WG is what I observed too (compared to my test a few months back). This regression is hard to pin down since with 6.6 I have the best SQM and Ksmbd performance test results yet with good memory usage too. It's unclear what's happening with WG is there a kmod missing or something not sure. Going to try 6.6.30 with the PCIe bug fix tonight just to make sure that didn't affect anything but don't think it would.
It shouldn't be that more difficult, you've got a couple of possibly options. e.g. you could just clone pesa1234's repo and then build from that, or if you've already got master, add pesa1234s repo as an a remote repo and either rebase or cherry-pick the commits you want...
I'm sure there are other and possibly better ways to do things, but the above methods worked for me when I was building NSS enabled firmware for my soon to be replaced Netgear device
Great build, im still unable to use any DFS channel above 100 @ 160mhz if I set the country code to GB though, im guessing thats an OpenWRT bug though, working fine if I set the country code to US.
Just for note if anyone else is unsure, to install FF, I downloaded the ipk file from directnupe link above, downloaded WinSCP and installed it, logged into my router and copied FF.ipk to the tmp directory on the router, then using putty execute the command opkg install /tmp/fantasic-feeds_2.2-2024-02-11_all.ipk
Finally continue with the rest of the commands as above.
You can also - just simply upload ipk from Luci > System > Software > Upload Package and then just install - this method works and is a bit simpler IMHO
Naturally - you will have to browse to directory where you have downloaded and stored
fantastic-feeds_2.2-2024-02-11_all.ipk
But to each their own - but it is good to have options
Thanks for your latest version. Was it intentional to have the WED WO firmware version stay on the 2022 version. The MTK release showed a 2024 update as well.
