Power Consumption on Thinkpad P14s AMD G2

[stettberger]

Hi!

I'm using the latest HEAD (febb6391e7e6ca8f04adbf2af56ad123defb4d77) with an Debian unstable kernel (Linux obelix 5.10.0-8-amd64 ). And it looks like the driver/the device requires a lot of energy. At least powertop says:

Power est.              Usage       Events/s    Category       Description
  22.7 W     39.5 pkts/s                Device         Network interface: wlp3s0 (rtw89_pci)

In the dmesg, I see:

[33713.163160] rtw89_pci 0000:03:00.0: firmware: direct-loading firmware rtw89/rtw8852a_fw.bin
[33713.164878] rtw89_pci 0000:03:00.0: Firmware version 0.13.24.0, cmd version 0, type 1
[33713.164880] rtw89_pci 0000:03:00.0: Firmware version 0.13.24.0, cmd version 0, type 3
[33713.184564] rtw89_pci 0000:03:00.0: chip rfe_type is 1
[33713.189958] rtw89_pci 0000:03:00.0 wlp3s0: renamed from wlan0
[33721.789295] wlp3s0: authenticate with f0:b0:14:a0:e9:83
[33721.985640] wlp3s0: send auth to f0:b0:14:a0:e9:83 (try 1/3)
[33721.987043] wlp3s0: authenticated
[33721.987233] wlp3s0: associate with f0:b0:14:a0:e9:83 (try 1/3)
[33721.990160] wlp3s0: RX AssocResp from f0:b0:14:a0:e9:83 (capab=0x1511 status=0 aid=1)
[33722.098853] wlp3s0: associated
[33722.098884] wlp3s0: Limiting TX power to 20 (23 - 3) dBm as advertised by f0:b0:14:a0:e9:83

23W seems a little much for my understanding of mobile devices. Is this a known issue?

[lwfinger]

I was not aware of PowerTop.. On my system (7-year old Toshiba i7 laptop), powertop shows power consumption about 7.5W. Running s flood ping to my router with over 1000 packets per sec, the power usage goes to 10W. It appears that the usage of the 8852AE is about 3W when if is busy sending and receiving.

On my system, the 8852AE is on an extender. It does not feel warm. I never found out how to get the wattage use by devices with powertop.

[stettberger]

I'm not sure if this power consumption is an actual issue. However, I have the problem that the Laptop drains (for my feeling) too much energy and the rtw89 driver is the only driver that I use from a secondary source.

My device is build in and lspci says:

03:00.0 Network controller: Realtek Semiconductor Co., Ltd. Device 8852
	Subsystem: Lenovo Device 4852
	Physical Slot: 0
	Flags: bus master, fast devsel, latency 0, IRQ 88, IOMMU group 13
	I/O ports at 3000 [size=256]
	Memory at fd700000 (64-bit, non-prefetchable) [size=1M]
	Capabilities: [40] Power Management version 3
	Capabilities: [50] MSI: Enable+ Count=1/1 Maskable- 64bit+
	Capabilities: [70] Express Endpoint, MSI 00
	Capabilities: [100] Advanced Error Reporting
	Capabilities: [148] Device Serial Number 00-e0-4c-ff-fe-88-52-01
	Capabilities: [158] Latency Tolerance Reporting
	Capabilities: [160] L1 PM Substates
	Kernel driver in use: rtw89_pci
	Kernel modules: rtw89pci

I'm not sure how I can give more useful information about this.

[covertg]

@stettberger, perhaps try recalibrating powertop to make sure that its calculations are relatively accurate? From my understanding, powertop's wattage calculations are all estimates, which rely on it taking a lot of readings from the device while going through different types of load (all while on battery power). See:

• https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/monitoring_and_managing_system_status_and_performance/managing-power-consumption-with-powertop_monitoring-and-managing-system-status-and-performance
• https://01.org/sites/default/files/page/powertop_users_guide_201412.pdf

I had also been experiencing significant battery drain on a new ideapad x AMD laptop, so did some exploring with powertop as well. For me though, the primary culprit was hybrid graphics. Setting up tlp helped a good bit as well.

Now, after running powertop --calibrate (and letting powertop run in the background for some hours after that by now, as well), the ~idle power consumption reported by powertop for rtw89_pci ranges from 1.4-1.5W (this is on 2.4ghz, a bit away from the router). A ping flood to my router (on 5ghz, nearby) pulls 2.4-2.8W with about 550 packets/sec. Note this is all on battery, not AC.

[stettberger]

@covertg: Thank you for this information. I didn't know about --calibrate. I did run the calibration, but still, my network interface with the rtw89_pci driver is reported to use 22.7 W at 23.8 pkts/s. The dmesg is free of any errors or warnings. The only printk() that could be associated to this issue is:

[193915.914665] wlp3s0: Limiting TX power to 20 (23 - 3) dBm as advertised by f0:b0:14:a0:e9:83

EDIT: I only have a single graphics card (amdgpu, Cezanne, rev d1).

EDIT2: Flood Pining has no influence on the Powertop Output. It always sticks to 22.7W

[lwfinger]

I have no idea what is going wrong with powertop, but I am fairly certain that your chip is not using 10 times the power of mine and @covertg's devices.

That message about limiting the TX power is a consequence of the regulatory rules in your location.

Are you using the Bluetooth part of the chip?

[stettberger]

I'm using it in the sense that bluetooth is activated. However, when I deactivate it in the UI, such that the device disappears in hcitool dev, strange things happened. First of all: The reported energy consumption did not change.

But my device became hell of unstable. iwlist sc sporadically said that it cannot see any networks. All networks reported with a signal strength of 0 db and my notebook could not authenticate to the accesspoint anymore. It said that the AP did not answer after 3 tries: timeout.

Reloading the modules did not help. Rebooting did not help. I almost went crazy, as there was no indicator why things were going downhill.

Then I rebooted to Windows, connected to a Wifi, and everything ist fine by now.

[domenkozar]

Possible fix https://patchwork.kernel.org/project/linux-wireless/patch/[email protected]/

[lwfinger]

If you had looked at 'git log', you would have seen that the patch was merged last Saturday.

I work closely with Ping-Ke Shih, khew that the patch was coming, and merged it here as soon as he pushed it to [email protected].