Bug 219429

Summary:	mt7925e Bad wireless performance in low signal conditions
Product:	Drivers	Reporter:	gm34
Component:	network-wireless	Assignee:	drivers_network-wireless (drivers_network-wireless)
Status:	REOPENED ---
Severity:	high	CC:	lkml, may, simon
Priority:	P3
Hardware:	AMD
OS:	Linux
Kernel Version:		Subsystem:
Regression:	No	Bisected commit-id:
Bug Depends on:	219173
Bug Blocks:

Description gm34 2024-10-26 09:39:35 UTC

I have Asus px13 laptop with internal wifi controller using the mt7925e and experiencing connection and transfer issues.

I have run a series of tests walking around different spots with an android phone, another laptop (with intel wifi) and with this mediatek laptop. On all devices I watched the "iw dev xxx station dump" information.

Mediatek laptop seems to consistently report up to 10 dbm less (worse) signal in similar spots than the other two devices. Because of that, mediatek has lower negotiated rx/tx speeds or even disconnects or refuses to connect.
The signal diff increases with signal attenuation, usually between -70 and -80. Especially at the edge of operating range mediatek reports -80 dbm and stops working, while other two devices have for example -70 dbm or even -68 and work happily at high speeds.
However, where the signal is good (-70 or better according to mediatek), the negotiated speed is ok, stable, and works to expectation.

In all cases I tested the reported speeds against a >1GB file transfer from local FTP server.

mt7921e.disable_aspm=Y does not seem to have any effect on the issue.

I have booted windows 11 from usb disk to test the mediatek laptop hardware and the signal reception seemed much better. At places where Linux disconnected, Windows driver would report -72 dbm (comparable to mobile phone). At places where Linux would not connect, Windows would report between -75 and -85 but still work. I tested the transfers using winscp with the local FTP server and they were functional.

This is a 5ghz only network with 802.11ac.

Linux version: 6.11.5

Any help would be appreciated!

Comment 1 Artem S. Tashkinov 2024-10-26 11:18:58 UTC

Very similar to my experience in bug 219173.

Comment 2 Artem S. Tashkinov 2024-10-26 11:20:49 UTC

Would be great if your posted your iperf3 results both in normal and reverse ( -R ) modes under Linux and Windows.

Comment 3 Artem S. Tashkinov 2024-10-26 11:24:12 UTC

Sadly mediatek developers are not on this bug tracker and are not aware of these bug reports :-(

Comment 4 Artem S. Tashkinov 2024-12-30 11:57:14 UTC

Please update the firmware and retest using Linux 6.12.7.

Comment 5 Antheas Kapenekakis 2025-03-05 17:52:10 UTC

Hi Artem, can you re-open this? I am getting this issue. I am on 6.13.5 and I cherry picked

pick feb268e8c40c wifi: mt76: mt7925: config the dwell time by firmware
pick 9613fb788c56 wifi: mt76: mt7925: replace zero-length array with flexible-array member
pick c432aa344028 wifi: mt76: mt7925: fix the unfinished command of regd_notifier before suspend
pick 18c64f7627d7 wifi: mt76: mt7925: fix CLC command timeout when suspend/resume
pick 264a2b7587e7 wifi: mt76: mt7925: add handler to hif suspend/resume event
pick 11f390565379 wifi: mt76: mt7925e: fix too long of wifi resume time

With no difference. If you can point me to new firmware I can test that too.

Comment 6 Antheas Kapenekakis 2025-03-05 17:55:55 UTC

Here is the dmesg snip from the wifi card

```
[    6.205710] mt7925e 0000:c2:00.0: ASIC revision: 79250000
[    6.282993] mt7925e 0000:c2:00.0: HW/SW Version: 0x8a108a10, Build Time: 20250113153001a
[    6.623260] mt7925e 0000:c2:00.0: WM Firmware Version: ____000000, Build Time: 20250113153106
[    7.312475] mt7925e 0000:c2:00.0 wlp194s0: renamed from wlan0
```

In the linux-firmware main I can currently see the same build time in the whence
```
Driver: mt7925 - MediaTek MT7925 Wireless MACs

File: mediatek/mt7925/WIFI_MT7925_PATCH_MCU_1_1_hdr.bin
Version: 20250113153001a
File: mediatek/mt7925/WIFI_RAM_CODE_MT7925_1_1.bin
Version: 20250113152854
```

Comment 7 may 2025-03-18 21:49:35 UTC

I can confirm that there are still noticeable differences between Windows & Linux.
Linux release: 6.14.0-rc7

Speedtest Linux: 200Mbit/s (ISP max 500Mbit/s)
Speedtest Windows 500Mbit/s (ISP max 500Mbit/s)

It is also suspicious, that txpower seems to be limited to 3.00 dBm.

❯ sudo iw wlp9s0 info
Interface wlp9s0
	ifindex 3
	wdev 0x1
	addr 44:fa:66:65:fd:c5
	ssid <REDACTED>
	type managed
	wiphy 0
	channel 100 (5500 MHz), width: 80 MHz, center1: 5530 MHz
	txpower 3.00 dBm
	multicast TXQ:
		qsz-byt	qsz-pkt	flows	drops	marks	overlmt	hashcol	tx-bytes	tx-packets
		0	0	0	0	0	0	0	0		0

Comment 8 Simon Allen 2025-04-16 04:06:25 UTC

Exact same issue on an integrated MT7925 built into a desktop Asus B850-I motherboard, on 6.14.2 (Arch). So it's not just chips build into laptops/USB.

At first I was under the impression that this was maybe related to Bluetooth as I was seemingly experiencing the issue only when Bluetooth headphones were connected, but after more testing this doesn't appear related.

I've tried disabling 2.4GHz/5GHz/6GHz and using each independently, and I've also tested manually selecting the channels at each of those frequencies, to no effect.

Comment 9 gm34 2025-04-16 06:45:49 UTC

Hi, after reading various online discussions, I believe the 3dbm transmit power information is just a placeholder and does not reflect the real value. It can't be changed either. 

FWIW I "fixed" this problem with $15 and 15 minutes of work: I replaced the card for intel ax210. The Broadcom card sits in my closet patiently waiting for its second chance in the future :)