Try setting up another network that has more conservative wireless settings, just as a test case, like I outlined in a comment above. That should eliminate the dropouts, it has in my case. It’s obviously not ideal as it will still be comparatively slow, but it at least makes the connection usable for most use cases until the driver situation for this chipset under Linux gets fixed.
Hey, thanks for the advice. However, I’d rather like to see if there is any chance in getting the MT7925 to work properly under Linux, or if I need to switch out my wifi card.
I took a look at the linux commit history of the MT7925 driver in the kernel and I do see quite some movement between v6.17 and the current v.6.19-rc2 (as of writing right now).
In particular I see
commit 066f417be5fd8c7fe581c5550206364735dad7a3
Author: Quan Zhou quan.zhou@mediatek.com
Date: Tue Nov 18 19:54:54 2025 +0800wifi: mt76: mt792x: fix wifi init fail by setting MCU_RUNNING after CLC load
commit 3bc62aa4484dcb9529fea70e756743769b1145e6
Date: Fri Oct 31 17:03:51 2025 +0800wifi: mt76: mt7925: add auto regdomain switch support
commit 2a035ae2062f38dc5183002d1c5a64ca682170c1
Date: Wed Oct 1 09:25:06 2025 +0800wifi: mt76: mt7925: cqm rssi low/high event notify
Potentially:
commit f5160304d57c5543c63404a15cc8737ef714b669
Date: Mon Sep 1 15:02:33 2025 +0200wifi: mt76: mt7996: Enable MLO support for client interfaces Report MT7996 MLO STA capabilities to mac80211 stack. Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> Link: https://patch.msgid.link/20250901-mt7996-enable-mlo-client-v1-1-50c46317325d@kernel.org
alltough this should just be a refact for 7925, but maybe they also “sneaked” in a fix here.
and maybe
commit d54424fbc53b4d6be00f90a8b529cd368f20d357
Date: Mon Sep 1 15:32:00 2025 +0800wifi: mt76: mt7925: add pci restore for hibernate
and
commit c7c682100cec97b699fe24b26d89278fd459cc84
Author: Rosen Penev rosenp@gmail.com
Date: Thu Sep 11 15:16:19 2025 -0700wifi: mt76: mt76_eeprom_override to int
There are more commits regarding the mt7925, so in general it seems to remain a region of active development in the Linux kernel, even tough its initial commits are by now two years old.
As I’m on bazzite changing the kernel is quite a pain for me and to see if v6.19-rc2 does any good to my wifi issues. Anybody here who can run some iperf performance tests on the v6.19 kernel and give me some feedback?
Kernel v.6.18 seems to only contain the
wifi: mt76: mt76_eeprom_override to int
commit. I rather not expect much improvement between v6.17 and v6.18, but you never know. 
1 Like
Its good to see more changes are getting merged.
However, the reality is that there still many issues with this card on Linux, regardless what tricks to try.
As stupid as it sounds, for now, swapping the card for an Ax210 really was the best solution for me. If in the future the drivers, firmware etc of the 7925e improve more and the card become more stable, I can always swap back.
I did some more experiments.
I exchanged the firmware blobs for the MT7925 with the most recent one on: linux-firmware
This is not ideal, as it keeps the linux diver (part of the kernel) unchanged but changes the mediathek provided firmware blob used by the kernel. (On Bazzite this can be done with OverlayFS.)
I didn’t do poper testing with the original and then changed firmware blob, but I compared it to another laptop with an old Intel 8265 / 8275 I have. Both latops are the identical framework laptop only the network card is different and the performance difference is significat:
Mediathek: 20-26 MBits/s (Exchanged Firmware, 5GHz)
Intel: 184-192 MBits/s (5GHz)
Mediathek: 43-58 MBits/s (Original Firmware on Bazzite, 2,4 GHz)
Mediathek: 18-25 MBits/s (Original Firmware on Bazzite, 5GHz)
The laptops are placed back to back and use the same desk and use the same wifi connection. In theory Meidathek should be a lot faster as it can use MIMO and Intel can not, but as we can see, that doesn’t really matter if you driver/firmware is broken.
In a nut shell, I don’t expect any significant improvement of the Mediathek with kernel 6.19 and thus I will try to give back the Mediathek chip and request another chip.
This is the same wifi card default on the Framework Desktop and I’ve been going crazy trying to fix different issues with it. Sent some 11 patches upstream already today.
Yeah, I think patches fix most of these issues, too. If Framework sends me a laptop, I would be happy to test it for them. 
Do you have a driver build with all of these changes implemented? I would be happy to test it out ahead of these commits making it to main.