Mobile Users Beware: Linux Has Major Power Regression

[yotambien]

Has the commit you are researching been identified yet?

Yes, I think that's what we all want to know. Any update on this?

[bwat47]

I've noticed in natty using the official 11.4 drivers rather than the one in the repos runs several degrees cooler. My fan is not going to max speed nearly as much.

[autono]

No heat here

I am a novice mechanic and Linux challenged. But I decided to run a test. Using the PartedMagic liveCDs 5.10 and 6.0 (with 2.6.37 and 2.6.38) I used the Psensor Gui to determine that core temperature readings at idle were virtually identical. (The 5th reading, Mobo?, was frozen at 30* in all tests.)

ECS K8 board with MPC 67, BE2300 (X2 @1900 and 45W TDP).

Next I opened System Stability Tester (GUI) and ran Pi to 4 Megs. Core temps increased but were nearly identical across the 2 kernels. This data could be unreliable for a number of reasons.

------------------------------------------

I do see that AMD's AM3+ products will have a radical new power flow that increases amperage. (See bit-tech.net review of future Asrock AM3+ boards.) If AMD is doing this then it is quite possible that Intel has similar changes planned. Could the kernel team be anticipating future hardware in 2.6.38?

[squirrl]

Do they test?

This tells us that nobody is testing power consumption except Phoronix

If distributions aren't going to test these kinds of things out then that looks seriously bad on the distributions.

After 15/20 years and still carrying regressions like baggage; it's unbelievable.

[devsk]

I haven't seen a regression in 2.6.38. In fact, 2.6.38.5 is the best battery life I can remember. I get around 6hrs of full use from Latitude, earlier it was somewhere between 5-6hrs. With videos, I can run it for 4 hours.

So, this post looks wrong to me, particularly because it has "Beware" word in it. But I will wait for the identification of commits which was promised in this thread.

[fewt]

I watched a couple of movies during my flight(s) this week and saw fantastic battery life. At Summit, I used my Netbook almost every day with WIFI on for hours at a time sometimes, and it still had ~60% battery life left when I put it on the charger. When I landed at home last night after watching 2 movies, it still reported ~7 hours of juice

Linux mechan 2.6.38.5-1.fc14.i686 #1 SMP Wed May 4 10:52:45 CDT 2011 i686 i686 i386 GNU/Linux

This "issue" .. doesn't exist.

[RealNC]

With so many threads about this issue, it's difficult to find whether Michael posted the reason for this regression or not.

Did he? I remember him saying that he would post about it on "Monday", which I think was last month.

[curaga]

I think Michael's still traveling in Euroland, and will continue this when back.

[Tim Chen]

Have not seen regressions in our testing

Odd. In our testing of 2.6.37, 2.6.38, ... 3.0 kernels, we have seen power on a Lenovo T410 (core i3) remains fairly steady for idle workload and also for openarena. We have not seen any regressions.

[gururise]

Jupiter depends on Mono, yes. It also uses less than 20MB of resident memory, and 0 CPU time.

20MB is quite a bit of memory on my eeepc. What advantage does Jupiter provide over laptop-mode-tools? Seems like most everything Jupiter does, so can laptop-mode-tools, without requiring an extra 20MB of memory usage.. and if Jupiter uses 20MB, how much extra memory does the mono-runtime use? Most mono programs I've used seem to require additional memory for the mono-runtime.