Wandboard Quad Rev. C1 - Temperature failure/System freeze

Discussion of your EDM baseboard, your add-on boards or other peripherals for your wandboard.

Wandboard Quad Rev. C1 - Temperature failure/System freeze

Postby lmendes » Thu Mar 02, 2017 8:41 am

Hi,
I own a Wandboard Quad from 2014 which is working nicely inside the box reaching about 75ºC under load conditions and is able to keep working under these conditions for days, but the same is not happening with 4 new Wanboard Quads produced in 2016, which I acquired from MOUSER distributor like the one in 2014.
Essentially all of them boot Linux and are able to enter the accelerated X GUI, however if I run a small stress load composed of glxgears, glxheads, a Vivante Qt test application and the screensaver, two of them fail after a few minutes and the other two before two hours under test. The tests are the same as the ones performed with the 2014 wandboard which works as expected. All tests share the same UBoot, same Linux distribution, same applications and same Vivante driver, which can even be YOCTO Krogoth with core-image-minimal-xfce, plus mesa test applications and Vivante test applications, Except that the 2016 Wandboards are tested without the box for better dissipation.
The failure behaviour observed is system freezing with last generated image frame displayed on the screen. The temperature of the CPU when failure occurs is 45 degrees Celsius for the two Wandboards that fail after 2 hours, the other two fail even at a lower temperature. The CPU temperature was checked by logging cat /sys/class/thermal/thermal-zone0/temp at each second to a log file.
If a 40x40mm FAN is placed on top of the heatsink right above the CPU the two Wanboards that would otherwise fail after 2 hours, then they are able to pass the test and work for days, with the CPU reaching at most 27 degrees Celsisus. The other 2 Wandboards fail after a few minutes even with the FAN placed on top of the heatsink.
All wandboards can be restarted and boot into X GUI after failure, just by power cycling with a few seconds interval.
Before the Wandboard freezes, at least, the ones lasting 2 hours, it is observed that the frames freeze for a few instants and them resume, these freezes become more frequent until the whole system freezes.
I suspect that the oscillator maybe defective failing to oscillate if the temperature reaches 45ºC.
Can you provide some information about what can be done to fix it? Maybe replacing a component? Is the problem known?
Thanks & Best regards,
Luís
lmendes
 
Posts: 3
Joined: Thu Mar 02, 2017 8:12 am

Re: Wandboard Quad Rev. C1 - Temperature failure/System free

Postby ClayMontgomery » Tue Apr 04, 2017 8:14 pm

I have seen problems like this on other i.MX6 quad boards that have no heatsink. They tend to work fine with just a linux command shell, but fail when the GPU and/or VPU are used for hours. Have you identified any difference between your boards other than their ages? I'm curious if you could measure any difference in the heatsinks using an external thermometer?

Regards, Clay
ClayMontgomery
 
Posts: 67
Joined: Tue Nov 19, 2013 9:14 pm

Re: Wandboard Quad Rev. C1 - Temperature failure/System free

Postby lmendes » Sun May 07, 2017 12:12 pm

Hi Clay,

Sorry for the delayed reply, but I've been busy lately.

I wasn't able to find further differences between the boards. They are the same revision level C1. It may be that at some point an electronic component was replaced by one with a different revision or from another manufacturer that may have introduced this behaviour.

The last test I did was to place K type thermocouple between the thermal tape and the chip die and what I found is that all the new Wandboards fail at low temperatures. Some at 40 degress Celsisus, others at 65 degrees Celsius, some take only a few seconds to freeze, while others take up to an hour to freeze, but all of them freeze.
The 2014 Wandboard reaches a temperature of 85 degrees Celsisus between the die and the thermal tape and it works happily for days.
All wandboards had the heatsink in place.
From this test I find that the problem isn't caused by SoC overtemperature, but rather other source.

Other test I did was to place a fan on top of the heastink, just above the i.MX6 SoC and what I found is that the Wandboard that fail at 40 degrees Celsius, fail just the same and in a matter of seconds, while the ones that would fail at 65 degress now keep working for days, because that temperature is never reached when the fan is installed. Anyway 65 degrees Celsius is much below the SoC maximum temperature which is 105 degrees Celsius. As such I think this cannot be a SoC overheat problem, but rather other compoonent.
I suspect that some component near the SoC may be susceptible to temperature variations, maybe some batch of faulty components. I don't know...

Regards,
Luís
lmendes
 
Posts: 3
Joined: Thu Mar 02, 2017 8:12 am

Re: Wandboard Quad Rev. C1 - Temperature failure/System free

Postby lmendes » Wed Jun 07, 2017 9:34 pm

I've tested a Wandboard Quad Plus rev. D1 and it doesn't suffer from this problem. Which is good news.
Regarding the rev. C1 Wandboards, I still don't have a solution for them.
lmendes
 
Posts: 3
Joined: Thu Mar 02, 2017 8:12 am


Return to Hardware and peripherals

Who is online

Users browsing this forum: No registered users and 13 guests