He’s in the UK like me where 36 and the block above it is the only non DFS channel on 5GHz therefore keeping your equipment on any other 5GHz channel is quite a challenge, almost everything will imagine it can hear nearby 5GHz radar signals within a few days and moves itself down to 36. Sometimes I manage to keep an access point on channel 100 for a couple of weeks, but it’s pretty rare.
Idk how is in the UK, but in the EU there are some channels after the DFS range, but not all APs allow you to do. Sometimes is worth to try a smaller channel BW in the DFS range to avoid radars.
Did you tried it on windows, just to take any doubt that you may have a defective or very old pre release batch?
Before 6.16 it was really rough for me, but since then it have been rock solid for me the Qualcomm card, there is just an bug with the reg domain, where when you connect to a WiFi6 AP, there is no 160MHz channels available, but when connected to a WiFi7 one it is available.
Personally I have the Qualcomm QCNCM865 and if I connect to a 6ghz network then wavemon shows a deluge of dropped packets (~50 per second which adds up to ~60% of the packets). If I connect to a 5ghz network then I’ll get a dropped packet maybe once every 10 seconds. I assume this is why my laptop would stay on 5ghz when connected to a combined 5+6 ghz network. The problem happens both on Arch Linux with kernel 6.16.10 and NixOS with kernel 6.16.11. Access point is a U7 Pro. This happens even when I am 6 feet from the access point with a clear line of sight.
Screenshot showing the dropped packets (in the 3rd section on the right)
If it wasn’t for wavemon, I might not have noticed, because I still get hundreds of megabits per second when connected to a 6ghz network, but the laptop’s refusal to upgrade to 6ghz in a combined 5+6 ghz network made me launch an investigation which eventually led me to wavemon where I noticed the dropped packets.
Yay. My NCM865 is finally working in Fedora after updating this morning.
2.4G, 5G, 6G and MLO SSIDs work.
However, 2.4G SSIDs on 2 different APs perform poorly. There’s not a lot of packet drops (aprox 5%) recorded, but it’s so much slower than any of the other bands (0.3Mbps vs aprox 80Mbps). I also had to toggle airplane mode off and on when switching SSIDs before they would connect.
Performance is also not what I would expect. The MLO SSIDs do not perform as well as just using the singular 5G or 6G SSIDs. But take all that with a grain of salt as Each of my APs are located in different rooms.
Kernel: 6.17.4-200.fc42
Firmware: atheros-firmware-20251011-1.fc42
Access Points Tested:
1 - Ubiquiti U7 Pro XG running firmware 8.0.49
2 - TP Link Omada EAP773 running beta firmware 1.1.1
Output of iw reg get:
```
global
country CA: DFS-FCC
(2402 - 2472 @ 40), (N/A, 30), (N/A)
(5150 - 5250 @ 80), (N/A, 23), (N/A), NO-OUTDOOR, AUTO-BW
(5250 - 5350 @ 80), (N/A, 24), (0 ms), DFS, AUTO-BW
(5470 - 5600 @ 80), (N/A, 24), (0 ms), DFS
(5650 - 5730 @ 80), (N/A, 24), (0 ms), DFS
(5735 - 5835 @ 80), (N/A, 30), (N/A)
(5925 - 7125 @ 320), (N/A, 12), (N/A), NO-OUTDOOR
phy#0 (self-managed)
country 00: DFS-UNSET
(2402 - 2472 @ 40), (N/A, 20), (N/A), NO-80MHZ, NO-160MHZ, NO-320MHZ
(2457 - 2482 @ 20), (N/A, 20), (N/A), NO-80MHZ, NO-160MHZ, NO-320MHZ, PASSIVE-SCAN
(5170 - 5330 @ 160), (N/A, 20), (N/A), AUTO-BW, NO-320MHZ, PASSIVE-SCAN
(5490 - 5730 @ 160), (N/A, 20), (N/A), AUTO-BW, NO-320MHZ, PASSIVE-SCAN
(5735 - 5895 @ 160), (N/A, 20), (N/A), AUTO-BW, NO-320MHZ, PASSIVE-SCAN
(5925 - 7125 @ 320), (N/A, 24), (N/A), NO-OUTDOOR, AUTO-BW
```
Anyone using NCM865 with Linux able to have working s2idle? For me the card is causing the kernel panic during wake up (blinking caps lock light on wake up). While it works pretty well with recent kernels (I am using Arch), the fact it kills the sleep mode makes it unusable on daily basis.
On windows it works for sleep, but for hibernate it wakes the PC right away after hibernation.
Really hoping for Intel WiFi 7 card working with AMD one day - MediaTek card suck, Qualcomm probably will take years to make it work smoothly. If Panther Lake ends up being comparable with Zen 5 in performance and Framework release a board with it, I will seriously consider switching back to blue camp - AMD should remember that there are actual users of their platforms out there, not just datacenters and AI.
I would say that Intel is no longer a good choice on Linux. It increasingly appears that they moving away from open source and the Linux ecosystem.
This shift is discussed here: Intel’s Open-Source Strategy Is Changing At Odds With The Ethos Of Open-Source
https://www.phoronix.com/news/Intel-IWD-Development-Halts
Regarding the NCM865, I can run tests on my side. Could you let me know what information or tests would be most useful for you?
Here is the best place to report issue with ath12k and NCM865 : ath12k mailing list
Thanks for the links - will read them!
I personally was in team blue for a while and never had big issues with Linux (I am big fan of upscale nuc series - had skull canyon, hades canyon and now running serpent canyon), and Framework with Ryzen AI 300 is my first device from team red in years. Intel was exemplary in Linux support. Sad it is changing now. I have high hopes for RISC-V, but is it pretty distant future still.
As of issue - just try if you framework can sleep and wake up with Qualcomm’s card (s2idle). Mine failes to wake up with kernel panic (blinking caps lock key light). Works great with AX210. Appreciate if you can check and share how it behaves for your device. Btw, what Linux distro are you running?
There are almost 300 posts in this thread and I don’t think any are having sleep issues with this card, including me. Seems like there is something uniquely wrong with the card you have.
archlinux on linux 6.19-rc6. I have no s2idle issue
1 Like
Thanks! Might be a hardware issue - I am getting another card tomorrow, will check if it works better. I am on stable kernel right now, will probably try rc as well, if no luck with stock.
P.S. No luck - same behavior with different card on 6.18 stable kernel.
P.P.S. Works on LTS kernel, looks like a bug in 6.18. I will stay on LTS kernel till 6.19 is out where fix is probably in, but otherwise WiFi 7 is cool - here is the iperf3 results with my Zyxel BE22000 access point:
My card does a similar thing on Windows11, when I use on Fedora it works fine, I can put the machine in sleep without issues, but, when I put the machine on sleep on W11, it wakes up and the WiFi card completely vanishes, only a cold boot solves.
I also put the same model WiFi card on Wife’s Asus laptop (only W11) and there is no issues at all.
So maybe there are some hardware bugs on some units
Framework 13 AMD (Ryzen 5 7640U) with Qualcomm WCN7850 on kernel 6.12.73-1 Debian 13.3. Had to grab the ath12k-firmware from git CodeLinaro / ath-firmware / ath12k-firmware · GitLab
AP is TP-Link Omada (EAP series).
5 GHz (WPA3) is stable.
6 GHz (WPA3) repeatedly disconnects and shows very asymmetric throughput (~1 Mbps down / 90+ Mbps up).
If I combine 2.4 + 5 GHz under one SSID it also starts bouncing between BSSIDs and dropping. Splitting 5 GHz out alone is stable.
Very temperamental and weird behaviour.
dmesg log occasionally shows things like:
ath12k_pci … failed to enqueue rx buf: -28
Looks like a 6 GHz / ath12k issue rather than AP config. Has anyone confirmed this is resolved on newer kernels (6.18/6.19+)?
I have never experienced those issues with mine. You may have bad firmware.
My card does not behave good with any firmware after 2025.06, but in my case I have 5GHz almost unusable.
but did not try using the latest from the CodeLinaro, only the ones available at Fedora repos.
My guess is that I may have an pre-prod unit, as when testing on windows, I have different issues (no wifi after wake from sleep)
Firmware is:
fw_version 0x110cffff fw_build_timestamp 2025-06-25 09:26 fw_build_id QC_IMAGE_VERSION_STRING=WLAN.HMT.1.1.c5-00302-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.115823.3
It loads cleanly. 2.4/5 GHz is stable. 6 GHz repeatedly disconnects and re-authenticates in a loop.
During the drops, dmesg does not show driver resets or firmware crashes at the moment of testing.
Do you guys know of any alternative wi-fi 7 card that works out of the box (I have the strix halo motherboard) without sketchy drivers or behaviours like disconnects and so on? A stable wi-fi 7 card or the next best thing.
If you on Intel, BE200 is a way to go. If you are on AMD, there are none. You can choose between Qualcomm and Mediatek. I have both: Qualcomm is fast, but buggy; Mediatek is slow and buggy. Choose one 
There is hope for new WiFi7 from Intel, but they have no interest to make it working with AMD processors.
What about a WIFI router that can be placed like close to the board –> which connects to the main wifi (bridge) and just use normal ethernet? that would basically replace the need of the small wifi card?
No need for this - just grab Intel AX210. It is WiFi 6E card and works perfectly with AMD boards. Cost is fine as well. Issue only is there if you are looking for WiFi7 solution.
1 Like