Hello, everyone!

I have done some test on a couple of different kernels, and Liquorix seemed to give the best results in both terms of speed and power saving:

Liquorix (recompiled to work on Squeeze)
kernel: linux-image-3.2.7-liquorix-686
glxgears: 572.322 FPS
power: 16W
battery: 40 min.

Debian Squeeze (it is more server-oriented)
kernel: linux-image-2.6.32-5-686
glxgears: 224.190 FPS
power: 19W
battery: 27 min.

Squeeze Backports
kernel: linux-image-3.2.0-0.bpo.2-686-pae
glxgears: 655.923 FPS
power: 20W
battery: 9 min.

Real-Time
kernel: linux-image-2.6.33.7-rt29-1-686
glxgears: 210.966 FPS
power: 20W
battery: not tested

GRML
kernel: linux-image-2.6.35-grml
glxgears: 202.164 FPS
power: 21W
battery: not tested

Also, the Opera browser is almost unusable in some of them, while it works perfectly on Liquorix (GPU). And sound has gaps in the Squeeze kernel.

Are my tests trustworthy in your opinion? Would you recommend Liquorix for laptops?

Would setting CONFIG_HZ_300 and/or CONFIG_PREEMPT_VOLUNTARY do any good to me, or would they just bring more slowness problems, just like the Debian kernel? It is a laptop but is powered on for a lot of time during the day.
Back to top

just out of curiousity, can you show: inxi -bxxx

wondering what the underlying hardware there is.

The squeeze kernel isn't a surprise, I mean, that's really server optimized, so that's not unexpected. But your other tests are interesting.
Back to top

Hello, thanks for the quick reply! My biggest surprise was that I "was sure" server optimized kernels would spend much less power.



<edited, removed serials, sorry, should have remembered -xxx shows the serials. Need to remember to request -z option to mask some data, ie, -bzxxx thanks techAdmin>
Back to top

If you're in the mood to test, try this:

open a terminal, open two tabs, in each tab, start this:



ie, send as many zeros to /dev/null as the system is capable of. To stop it, do ctrl+c, after you get your wattage readings.

do two to get both cpu cores going fast, or 4, in 4 tabs, that's an easy way to get cpu spikes that are consistent and based on the same thing. 2 of the processes at the same time should do it, unless that cpu maybe tries to use hyperthreading or something. You can check kernel load with htop as it runs.

then measure wattage with each kernel again and see what the wattage is with all cores fully working.

Just curious, but I believe that server stuff is not built around the idea of saving laptop battery life, but desktop optimized kernels are. Or should be. I am curious however to see if server optimized kernels are optimized for efficient high cpu load. And how those numers are different in the kernels you tried.
Back to top

I think these seem a bit odd:

Liquorix: 19W
Debian: 33W
BPO: 34W
RT: 34W
GRML: 33W
Back to top

I think that all the other kernels are missing some important ASPM patches that made it in to 3.3 and 3.4. Even before then, I was backporting the patches from the LKML so that you guys wouldn't get a strong power draw when idle. It seems to me that there's more to it than being idle, even during intense CPU activity, some unimportant hardware can be shut down.
Back to top

OK, thanks for you answer and most of all, thank you for making such a great kernel!
Back to top