Get the temperature of LAN Chip for P3B+

[Shawn1913]

Hi all,
I had a post at viewtopic.php?f=28&t=217092 which is asking a case of P3B+ always shutdown with my application running for hours, and I've logged performance counter with command

Code: Select all

/opt/vc/bin/vcgencmd

and

Code: Select all

top -n 1

along the test, I can see the temperature (<65'C), Memory usage (<10), Cpu usage(<30) are all good, so I strongly suspect the LAN Chip heat is the root cause since from the official doc Operating Temperatur
said the LAN Chip dead temp is just 70'C, and my application is doing most of job about the TCP/IP and Usb serial communication.
So just want to know is there any API like vcgencmd I can use to read the LAN Chip temperature ? if there's no, how can I sure the problems like this? any system logs could help?

[epoch1970]

I don't know the answer to that question, I imagine there is no such sensor.
However, are you talking of 70°C ambient or 70°C CPU temp?

70°C CPU temp is nothing to write home about.
At 70°C ambient temp, all the ancillaries like ethernet or usb cables are likely to start melting I think. I don't know of many ethernet cables rated to operate over 60°C ambient, for starters.

[hippy]

So just want to know is there any API like vcgencmd I can use to read the LAN Chip temperature ?

There appears to be no LAN9515 datasheet publicly available, almost no data at all, but none of the LAN9512, LAN9513 and LAN9514 devices have an on-chip temperature sensor mentioned so it would be reasonable to presume it is the same and there is no way the chip can report its own temperature, and therefore no way for vcgencmd or anything else to know or report what it would be.

if there's no, how can I sure the problems like this?

About the only thing you could do is take a thermal imaging camera or temperature probe and measure what the chip's temperature is.

You could cautiously put your finger on it. You'll soon know which side of 50C it is.

[jamesh]

If I remember, I'l try and get a IR image of the board under heavy network load tomorrow.

[Shawn1913]

If I remember, I'l try and get a IR image of the board under heavy network load tomorrow.

Hey jamesh, I can't wait for you testing result!
My case is:

• open 2 TCP connections to external device, send 5 messages(and will receive 5 response as well) in 1 seconds in each connection, and each message is about 50 bytes

• open 2 USB serial(by USB to RS232 converter) connections to another external device, send 3 message(and will receive 3 response as well) in 1 seconds in each connection, and each message is about 20 bytes

[hippy]

It seems hard to imagine that such a small amount of data being transferred would lead you to the suspicion that it's the LAN chip overheating and stopping working.

Obviously something is going wrong with your system but I wouldn't suspect 'thrashing the LAN chip' would be causing that.

[jamesh]

It seems hard to imagine that such a small amount of data being transferred would lead you to the suspicion that it's the LAN chip overheating and stopping working.

Obviously something is going wrong with your system but I wouldn't suspect 'thrashing the LAN chip' would be causing that.

Indeed. From a networking point of view, that's practically idle. Seriously doubt there are any problems with overheating due to traffic levels.

[jamesh]

OK, had a play with the IR camera on the ethernet chip whilst under a bit of load (copying from Samba share to SD card). Went from 54-55ºc to 55-56ºc, so not a huge rise. Hottest chip on the board though, although nowhere near any critical temperatures.

[Shawn1913]

OK, had a play with the IR camera on the ethernet chip whilst under a bit of load (copying from Samba share to SD card). Went from 54-55ºc to 55-56ºc, so not a huge rise. Hottest chip on the board though, although nowhere near any critical temperatures.

thanks for the testing, so for my case, do you have any suggestion for further narrow down?

[jamesh]

Since the Pi is apparently actually shutting down, I have no idea. Cannot think of a reason why that might happen.

Have you any logs or possible symptoms? Power issues?

[Shawn1913]

Since the Pi is apparently actually shutting down, I have no idea. Cannot think of a reason why that might happen.

Have you any logs or possible symptoms? Power issues?

Unlikely the power issue since my power adapter is from a famous brand with quality (but will still try another by your suggestion).
I only have the application logs which basically no help (it just suddenly stopped).
Can I check any system logs to help? could you give a hint.

Last thought, As you tested, the LAN Chip is always(?) the hottest, and now the SOC temp reading is 63'C, can I say the LAN Chip must hotter than SOC, that means it almost reached its dead temp 70' C???

[jamesh]

Since the Pi is apparently actually shutting down, I have no idea. Cannot think of a reason why that might happen.

Have you any logs or possible symptoms? Power issues?

Unlikely the power issue since my power adapter is from a famous brand with quality (but will still try another by your suggestion).
I only have the application logs which basically no help (it just suddenly stopped).
Can I check any system logs to help? could you give a hint.

Last thought, As you tested, the LAN Chip is always(?) the hottest, and now the SOC temp reading is 63'C, can I say the LAN Chip must hotter than SOC, that means it almost reached its dead temp 70' C???

No. My tests were with an idle CPU. It will get hotter under load. There is no direct link between CPU and ethernet chip temps.

[Shawn1913]

Since the Pi is apparently actually shutting down, I have no idea. Cannot think of a reason why that might happen.

Have you any logs or possible symptoms? Power issues?

Unlikely the power issue since my power adapter is from a famous brand with quality (but will still try another by your suggestion).
I only have the application logs which basically no help (it just suddenly stopped).
Can I check any system logs to help? could you give a hint.

Last thought, As you tested, the LAN Chip is always(?) the hottest, and now the SOC temp reading is 63'C, can I say the LAN Chip must hotter than SOC, that means it almost reached its dead temp 70' C???

No. My tests were with an idle CPU. It will get hotter under load. There is no direct link between CPU and ethernet chip temps.

Ok...
So could you give some suggestion on this kind of 'automatically shutdown' case? there's must have somewhere logged some clue.
BTW, Does RaspeberryPi.org provide business consultant service, like consultant for technical issues?

[jamesh]

Unlikely the power issue since my power adapter is from a famous brand with quality (but will still try another by your suggestion).
I only have the application logs which basically no help (it just suddenly stopped).
Can I check any system logs to help? could you give a hint.

Last thought, As you tested, the LAN Chip is always(?) the hottest, and now the SOC temp reading is 63'C, can I say the LAN Chip must hotter than SOC, that means it almost reached its dead temp 70' C???

No. My tests were with an idle CPU. It will get hotter under load. There is no direct link between CPU and ethernet chip temps.

Ok...
So could you give some suggestion on this kind of 'automatically shutdown' case? there's must have somewhere logged some clue.
BTW, Does RaspeberryPi.org provide business consultant service, like consultant for technical issues?

No, we have no paid business service. Use the forums and github issues.

TBH, the help you get from the userbase is pretty good. And since I work for Raspberry Pi as an engineer, it's not like we don't see and comment on things like this.

With regard to the issue, automatic shutdown is something I have rarely heard of. Overheating simply causes the CPU to lower is performance until the heat level drops - it does not cause a reboot.

You could look in syslog to see what happens just before the shutdown, but if this is some weird HW fault or odd driver failure, its unlikely to show anything.

Having just look through the linked thread - does it actually reboot, or does it lock up? They are different things. I wonder if this is either the test escape SOC issue , or perhaps the EEE issue , or even the tcp segmentatation offload problem. In all cases you need the latest Raspbian to mitigate the issue.

Do an

Code: Select all

sudo apt update
sudo apt upgrade

to get the latest bug fixes.

Someone has also recently reported a further problem with EEE as well, you can try and turn it off if the update above doesn't work. You'll need to do this each time you reboot, until we know if it fixes the problem.

Code: Select all

sudo ethtool --set-eee eth0 eee off

[epoch1970]

(If the OP had installed the watchdog package -a recent version at that-, I think there is a way to get it to halt the machine when the Pi temp sensor reaches a certain value... But usually you don't install watchdog inadvertently and indeed there are traces in syslog.)

[jamesh]

(If the OP had installed the watchdog package -a recent version at that-, I think there is a way to get it to halt the machine when the Pi temp sensor reaches a certain value... But usually you don't install watchdog inadvertently and indeed there are traces in syslog.)

I really don't think this is an overheating problem. There are other more likely causes.