Assume turning PSK encryption off entirely (just for testing) would further reduce CPU overhead?
Thanks for that. I always assumed but didn’t know for sure that HW offloading only really played a part with wired to wired traffic. Plus that for your average office environment, where most traffic is Wi-Fi, HW offloading could be pretty much ignored. The block diagram confirms that all Wi-Fi traffic goes via the CPU?
That’s a sound deduction, though I wouldn’t be able to test it on my unit. This is servicing a live production environment.
Raw tests using external speedtest (ISP = 500/30) using Samsung S25 as client
5Ghz, 5300, no channel width limit
wpa3 PSK: Around 300/29, one CPU towards 75% (3 runs)
wpa2 PSK: a tiny bit more, around 320/29, sometimes a bit more, cpu about the same (3 runs)
No security: inching towards 330/29, cpu about the same (3 runs)
But speedtest done from PC using ethernet, about the same results as above.
So now I need to redo everything using internal iperf3 server...
iperf3 -P 2, tested both with and without -R
PC (baseline, connected to ethernet, ether1 of hAP AX S towards router): 892/862 - 1 core inching to 40%
wpa3: 350/254 (cpu 58/62)
wpa2: 381/252 (cpu 63/52)
No security: 389/244 (cpu: towards 70/around 55)
Esp. on download it looks like it has an impact but upload is not so sure without more and better testing (logically it should).
I assume you need a client with a Wi-Fi adapter cable of using 3 chains? But even then, those speeds seem rather slow?