Rockchip running hot while idle (and if x86 fix would be applicable to aarch64)

For Developers
1

I have two NanoPis (R5S and R6S, RK3568B2 and RK3588S respectively) both running 24.10 snapshot.

Even while idle (light or no internet activity via LAN<->WAN), the case is quite warm to the touch. CPU temperature monitoring (collectd sensors and thermal) is showing temperatures around 50°C (122°F) with room temperature around 26°C (78°F)

The only additional packages I have installed are adguardhome and iperf3, but I tried also default install (without these pacakges) and it made no difference.

I noticed the following commit for x86 CPUs that reduced power consumption and also idle temperatures.

Would this change also be applicable to aarch64 CPUs?

https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=4c22224bcc08c9bd18e6e9921280c73e2abcd7c3

-CONFIG_HZ_PERIODIC=y
+CONFIG_NO_HZ=y
+CONFIG_NO_HZ_COMMON=y
+CONFIG_NO_HZ_FULL=y
1 Like
2

Armbian agrees with you

1 Like
3

Thanks. When I have time I will try a custom build with this change to see if it has any impact on the Rockchip idle temperatures.

EDIT: it seems that Rockchip’s config-6.6 already includes the following settings:

CONFIG_NO_HZ_COMMON=y
CONFIG_NO_HZ_IDLE=y

However it is missing some other settings such:

CONFIG_NO_HZ=y
CONFIG_NO_HZ_FULL=y

Not sure if these are/would be required…

4

try the stable version instead of snapshot

5

There is no 23.05.x official stable version for the R5S/R6S. Also notice I am using a 24.10 snapshot (not main snapshots).

6

Then you should wait until a stable version is released.
Meantime you can run alternative os which supports the device fully.

7

Sorry, you are not helping with such remarks. This is a developers’ area where such kind of experimentation and results are welcome. This is the way we users can contribute to the OpenWrt development team.

8

nohzfull is not needed. What i posted is from armbian, not openwrt. Try to align settings.
used to have r2s, nice to do nat alone at ideal speeds, but heat under usage felt dangerous.

1 Like
9

Is there anything stange in dmesg logread top lsof?

10

You see that on 23.05.5 x86_64 in a VM that it steadily eats CPU unlike eg ubuntu dektop in a vm. Kind of fixed for x86 now, but probably not in ideal state in all platforms.

2 Likes