3.1 kernel with a new selective power regression

[Oxwivi]

Phoronix says (http://www.phoronix.com/scan.php?page=article&item=linux_31_power_regress) 3.1 introduces another power regression, but this time it's selective - only affecting PCs with a yet-unknown common factor.

Mobile Linux is gonna take a beating, dammit.

[Erik1984]

Bad news :( Not to good for desktop users either (probably more noisy fans and higher energy bill). Just keep your finger crossed and hope you are amongst the lucky hardware configuration when 3.1 kicks in or this problem gets fixed before that day. Also a good reason to keep old kernels around.

[3rdalbum]

Phoronix says (http://www.phoronix.com/scan.php?page=article&item=linux_31_power_regress) 3.1 introduces another power regression, but this time it's selective - only affecting PCs with a yet-unknown common factor.

Mobile Linux is gonna take a beating, dammit.

Unfortunately, power management is done fundamentally wrong at the hardware level on normal PC architecture. It's amazing that suspend and hibernate work at all on Linux, really.

As for the actual power regressions, I'd probably doubt that this is anything new. Phoronix only recently added a thingy to their test suite that measures power consumption, so now they can actually see differences on certain affected pieces of hardware. Unfortunately, a 10% power regression on some hardware might be the fix to a "kernel panic" problem on some other hardware, such is the state of x86 PC architecture.

I'd rather lose 10% battery life than lose the ability to run Linux.

But this stuff is nothing new, it's just that Phoronix can measure it now.

[Oxwivi]

Comparison to older kernel's power consumption hasn't got anything to do with Phoronix Test Suite's new features.

[mips]

I just wish they would fix the reboot bug I have.

[JDShu]

Phoronix says (http://www.phoronix.com/scan.php?page=article&item=linux_31_power_regress) 3.1 introduces another power regression, but this time it's selective - only affecting PCs with a yet-unknown common factor.


The regression he was talking about last time was selective as well - it was specific to a BIOS software.

On a broader note, Phoronix is notorious for making a big issue about bugs that he finds. Bugs that he finds often don't affect most people, and my guess is that there are bugs that he doesn't find that will affect others.

[AloofObserver]

Wow. That is depressing. I'm kinda glad my laptop hasn't been affected by the ASPM regression and apparently won't be affected by the 3.1 regressing.

Depressing nonetheless.

[Oxwivi]

Bugs that he finds often don't affect most people, and my guess is that there are bugs that he doesn't find that will affect others.

I'd say there's equal chance of them reporting either types of bugs.

[JDShu]

I'd say there's equal chance of them reporting either types of bugs.

Yep, my point is that bugs happen, unfortunately. Phoronix just comes across as profiteering by overemphasising bugs that he found, when it's really just one of many.

[Oxwivi]

All is forgiven as long as they keep developing PTS and find more bugs. :P

But I think they rightly label the power regressions as 'major', since energy prices are never going down and the loss of mobile convenience.

Well, I maybe biased - I happen to highly value efficiency.

[nrundy]

Unfortunately, power management is done fundamentally wrong at the hardware level on normal PC architecture. It's amazing that suspend and hibernate work at all on Linux, really.


Are these regressions that Phoronix keeps finding only occurring on "windows pc" hardware (computers that come with Windows preinstalled)? Is there also increased power draw on "Linux pc" hardware, like on computers from System 76 and Zareason, etc?

[JDShu]

This is not a matter of relative importance. Phoronix found a power regression and wrote an article that essentially generates FUD over it. The problem is that they don't own every hardware configuration there is, meaning that they can only write articles about bugs regarding their own hardware configurations. But there is likely to be a regression (including power regressions) on some specific hardware every cycle - we just get sensational headlines if Phoronix happens to own that hardware.

[nrundy]

Ubuntu does use more power than Windows though. Lucid used more power than Win7. And Natty runs at like 12/13 watts. Whereas Win7 is well below this, like 8 watts or so I think.

How would one figure out whether this is due to a regression on their specific hardware or whether Linux just uses more power than Win7?

[Johnb0y]

Well, I maybe biased - I happen to highly value efficiency.

Well... you might like this desktop picutre i quickly made up ages ago...:P

[Thewhistlingwind]

How would one figure out whether this is due to a regression on their specific hardware or whether Linux just uses more power than Win7?

Linux just uses more power, the kernel doesn't have very good power management. (Yet)

[Oxwivi]

Well... you might like this desktop picutre i quickly made up ages ago...:P
I'm sorry, but a simple dark background is more eco-friendly and certainly more effort-less to apply! :P