Stress Test (kernel compile)

[hmuller]

I ran a series of kernel compilations to stress test the system and to answer the following questions:

1. Can the RK3399 SoC handle the rigor of kernel compilation?
   
2. Do the thermal characteristics of the SoC remain in the recommended range throughout kernel compilation?
   
3. Do the SoC/GPU temperatures truly differ?
   
4. Does the system under heavy load discharge the battery WHILE attached to DC power supply? (https://forum.pine64.org/showthread.php?...9#pid52379)
   

I used an external USB 3.0 hard drive to both apply an additional power load on the system and to offload compilation writes (decrease writes to eMMC).

In summary:

1. The PBP performed well, no lockups or freezes experienced.
   
2. The max SoC temp measured during make -j6 was 74 degrees Celsius, within recommended operating range.
   
3. The SoC/GPU temperatures do differ throughout operation (to my surprise).
   
4. Data indicates that under kernel compilation with external hard drive attached, the battery does not discharge while attached to the DC power supply.
   

Subjectively, although the bottom of the laptop did get very warm it was not uncomfortably so. I did not think to measure the case temperature with anything other than my fingers.

UPDATE 02 DEC 2019:

The tools used to run the compile test can be found here:

• https://github.com/hxmuller/pbp-compile-test
  

The data I produced along with a report can be found here:

• https://github.com/hxmuller/pbp-compile-test-data
  

[as400]

Thanks for this. Very useful info.

Did you get any WiFi disconnections during the compile ? There are some reports on the forum that the temperature is responsible for the problems with WiFi.

[murak]

I ran a series of kernel compilations to stress test the system and to answer the following questions:

1. Can the RK3399 SoC handle the rigor of kernel compilation?
   
2. Do the thermal characteristics of the SoC remain in the recommended range throughout kernel compilation?
   
3. Do the SoC/GPU temperatures truly differ?
   
4. Does the system under heavy load discharge the battery WHILE attached to DC power supply? (https://forum.pine64.org/showthread.php?...9#pid52379)
   

I used an external USB 3.0 hard drive to both apply an additional power load on the system and to offload compilation writes (decrease writes to eMMC).

In summary:

1. The PBP performed well, no lockups or freezes experienced.
   
2. The max SoC temp measured during make -j6 was 74 degrees Celsius, within recommended operating range.
   
3. The SoC/GPU temperatures do differ throughout operation (to my surprise).
   
4. Data indicates that under kernel compilation with external hard drive attached, the battery does not discharge while attached to the DC power supply.
   

Subjectively, although the bottom of the laptop did get very warm it was not uncomfortably so. I did not think to measure the case temperature with anything other than my fingers.

Thank you! Compiling is a good real-world test and it shows the quality of the pinebook pro hardware. 

Is there some other gpu heavy test that can be run simultatiosly to stress both cpu and gpu at the same time? Maybe a timedemo in Quake3? 

I am olso looking for a smal program to display cpu/gpu load and temp. It was many years since I´ve done any stresstesting on Linux, but back then I used conky. Is there another option nowadays? Maby something even more lightweight?

[hmuller]

Did you get any WiFi disconnections during the compile ? There are some reports on the forum that the temperature is responsible for the problems with WiFi.

Reviewing /var/log/syslog shows quite a few messages related to WiFi but they appear to be interspersed throughout the log and not isolated to periods with increased temperatures. So I cannot confirm that correlation.

Is there some other gpu heavy test that can be run simultatiosly to stress both cpu and gpu at the same time? Maybe a timedemo in Quake3? 

I am olso looking for a smal program to display cpu/gpu load and temp. It was many years since I´ve done any stresstesting on Linux, but back then I used conky. Is there another option nowadays? Maby something even more lightweight?

A Google search on this provides interesting results regarding GPU stress testing. It's not something I have looked into. And given my lack of familiarity with that topic, caution would be my watch word.

I don't have any recommendations on cpu/gpu monitoring for the PBP. But I use the i3wm DE on my desktop computer and have configured i3bar to provide that information. Conky, as far as I know, would still be an option to consider.

[mfritsche]

Hmm. Mine won't survive a -j6 kernel compile without signs of severe heat stroke. Guess it's time to get the screwdriver out

[Thra11]

When compiling for a long time on all cores, mine can get warm too. I've noticed that if it gets too hot, the red LED next to the power cable flashes instead of being steady, and it doesn't appear to charge (or only charges very slowly). I assume this is the charging circuit refusing to charge the battery when it's too hot (IIRC, LiPo batteries can catch fire/explode if they overheat). if someone can confirm that this is what the flashing red LED means, we could add it to the wiki. Whether the laptop overheats and gradually discharges, or manages to slowly charge seems to vary depending on how well it can dissipate heat from the case: Sitting on my lap or on an insulating surface -> overheating, discharging, Resting in a position with the base exposed to the air -> gradual charging.

[agD0i7rY]

I haven't tried a kernel yet but I have tried to compile some big projects* with -j6 and each time the Pbp gets hot during compilation and shuts down.
On the other hand for small compilations no problem with -j6 but gain is negligible anyway.
From memory I had no problem with -j4 for Retroarch.

* They are big for me

[mfritsche]

Hmm. Mine won't survive a -j6 kernel compile without signs of severe heat stroke. Guess it's time to get the screwdriver out

Added thermal pads on the casing around the cpu and thermal paste on the cpu thermal pad. Without a cooling pad under the pinebook I still get compilation errors with -j6 - but temperature seems to be generally lower.

Cross compiling is annoying (try ghc), emulation of aarch64 is slooow (not even going there), so compiling on the pinebook pro is the only viable option.

[nobodyinperson]

I recently compiled hledger (Haskell) and really had problems keeping my Pinebook Pro alive! It wouldn't charge (red LED flashing, power supply not even getting warm). In the end I was able to complete the compilation by reducing display brightness and using a cooling pad. That kept the battery at around 9%. Was rather a stress test for me than my Pinebook ;-)