Tried on all kernels where ath11k is working, including Kale Valo pending-master git. When wifi is connected (just connected, no transfer is happening) - CPU never goes below package state C6(pc6) (measured by powertop). Sometimes even gets stuck on C3(pc3). Power consumption with nothing running and screen closed - 1.9 watts. When wifi is disconnected - CPU stays 80-85% in C10 (PC10) and consumes 0.8-0.9 watts of power under the same condition. for the comparison iwlwifi based XPS 9310 (same but 16gb ram) stays 80-90% in C10 (PC10) and 0.8-0.9 watts power even with wifi connected.
Tested on my 9310 32GB 5.14.11 Simply disabling wifi drops 0.8-1 watts from power consumption. Comparing almost same model (16 gb RAM vs 32) but based on Intel card shows the same difference - Intel based laptop with WiFi connected consumes about same as ath11k with Wifi discnnected. Powertop newer shows PC states below PC6. Disabling WiFi allows PC8, disabling ath11k completely (bios) - PC10. Wifi enabled Pkg(HW) C2 (pc2) 18,5% | C3 (pc3) 14,2% | C6 (pc6) 47,5% | C7 (pc7) 0,0% | C8 (pc8) 0,0% | C9 (pc9) 0,0% | C10 (pc10) 0,0% | Wfi disabled (by networkmanager/nm-applet) Pkg(HW) C2 (pc2) 10,8% | C3 (pc3) 18,5% | C6 (pc6) 1,1% | C7 (pc7) 0,3% | C8 (pc8) 28,3% | C9 (pc9) 0,0% | C10 (pc10) 0,0% | Seems that even on no network activity ath11k forces CPU to stay awake and eat battery. There are some background transfer happening as I have Viber, thunderbird and firefox opened so I bet they do sent some but there is no file transfers, media streams etc. Under the same conditions (taking this disk from one 9310 to other 9310 - same software, same kernel - just wifi's different) - iwlwifi manages to reach PC8-PC10 being connected. With ath10k on dell XPS-13 9350 intel also was saving battery much better than ath10k, bit there was like 88% of PC10 with intel and 65% PC10 and 20% PC8 with ath10k. But it was not such terribly bad as with ath11.
There's now a proposed patch but unfortunately it needs some work: https://patchwork.kernel.org/project/linux-wireless/patch/20211011190625.4463-1-cjhuang@codeaurora.org/
The commit which should fix this: https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/commit/?h=ath-next&id=b2beffa7d9a67b59b085616a27f1d10b1e80784f If all goes well the commit will be in v5.17.