A Look At Nouveau Driver Power Usage

[phoronix]

Phoronix: A Look At Nouveau Driver Power Usage

There's been many individuals asking how the work is going in tracking down the major Linux kernel power regression I brought to light late last month (actually, there's at least two power regressions in the kernel). Not much progress has been made since then as I've been out of the office (and country) so I've been preoccupied with other matters, but I do happen to have another power test today to satisfy other reader requests.

http://www.phoronix.com/vr.php?view=15991

[Ansla]

What was this benchmark about?

Since noveau does not have power management enabled it was clear from the beginning it will perform similar to the blob with power management disabled. I was expecting a test with power management enabled (if noveau has this implemented, otherwise maybe for radeon that does have at least a basic implementation).

[Cyborg16]

Well, these do tell us that it's possible to save 5-7 Watts on idle/desktop usage, which is significant. It's a shame we can't see how much of the system power usage is due to the GPU, though.

[MuPuF]

Well, these do tell us that it's possible to save 5-7 Watts on idle/desktop usage, which is significant. It's a shame we can't see how much of the system power usage is due to the GPU, though.

5-7W is what we experienced too. Michael could have tried changing the perflvl, nouveau "supports" that but it would have saved 2 to 3W, not more.

Anyway, focus is on fan management in the nouveau PM team. We may issue a call for help on this issue as this is a big WTF at the moment.

[Nobu]

The Nouveau driver with its adaptive power optimizations left the GPU in its lowest performance state the entire time, even during 1080p video playback via X-Video.

Wait, what? You mean the nVidia driver, right? Of course, I would be happy if that was actually the case.

[jcgeny]

have you tried to slim "the kernel" down , step by step .
removing module-drivers one by one ...

[RealNC]

have you tried to slim "the kernel" down , step by step .
removing module-drivers one by one ...

With the reason for doing that being...?

[jcgeny]

the power bug should come from one module - driver that interact badly with the cpu .
may be it is a combination or two or more that makes it .
i mean that if he remove modules like an onion ,, the bug should disappear at one time .

[Nobu]

That would work, but only if it's a bug in a module. If it's in core kernel code, then you could have removed all the modules and still experience the bug.

Still, it's not a bad idea, but it might take a while.