<div dir="ltr"><div><div>OK, I confirm that the 15.04 second stable release behaves exactly the same as the TI SDK does with respect to idle prompt power consumption. cpufreq with ondemand governor also works. </div>Regards, </div>Gabor </div><div class="gmail_extra"> <div class="gmail_quote">On Thu, Jul 30, 2015 at 2:23 PM, Oliver Grawert <<a href="mailto:ogra@ubuntu.com" target="_blank">ogra@ubuntu.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">hi, Am Donnerstag, den 30.07.2015, 12:52 +0200 schrieb Gábor Paller: > > Linux localhost.localdomain 3.19.0-18-generic #18-Ubuntu SMP Tue May ah, that might be the issue ;) (BeagleBoneBlack)ubuntu@localhost:~$ uname -a Linux localhost.localdomain 3.19.0-23-generic #24-Ubuntu SMP Tue Jul 7 18:56:34 UTC 2015 armv7l armv7l armv7l GNU/Linux > > > This is a 15.04 Snappy on BBB without any kernel hacking. > > "is this under the same userspace too (i.e. are you sure it is kernel > induced and not some bad userspace component) ?" > > > My test is the following. I created an SD card from the stock Snappy > image, flashed it and booted it. I waited for the boot sequence to > finish and waited for the idle prompt. I measure 340 mA of > consumption. I did the same with the TI SDK and I measure 250 mA. > Hence the 40% difference. > > > As with any Linux distribution it is not only the "kernel" that is > running in any of these cases but a bunch of drivers/daemons too. But > there is no userspace component in these cases, at least I did not > install/run any. > well, i'm just trying to find out if "TI SDK" means you have your own rootfs or if you just put vmlinuz and the modules from the TI SDK into an already installed snappy system ... there can be plenty differences on kernel level (does the TI kernel actually have all needed apparmor patches for example). but i would expect userspace difference too if you dont use the very same rootfs for measuring both. to identify if the higher power consumption is a kernel or userspace thing you can indeed only measure reliably by using the same environment to run the different kernels in, else you might for example have userspace bits that turn different knobs in sysfs on boot etc which will definitely get you different hardware behavior. ciao oli </blockquote></div> </div>