Power Consumption

[juniorm33]

Yea, wasn't the boards that was slower, it was the kernal.
But easy to see why the confusion happened.
Switched the kernal, suddenly we're all seeing lower clocking. And it probably even happened around the time of the board revision.

Where the very high original overclocks were at OPP5, and any we've seen recently seem to have been at OPP3 so aren't directly equivalent.

No voltage scaling in current kernal.

All very good news, and really nice bit of info.

I'm sure we can find some way to cause hassle from it though, thinking hat's on people!

 

[Monk]

I'm sure we can find some way to cause hassle from it though, thinking hat's on people!

Pfft, easy! Thinking hats? I don't even need socks on to make trouble from that!

 

[WizardStan]

and quickly add "this seems to suggest that OpenPandora are using mostly OPP3, including at 600MHz".

Indeed.
His remark that they've removed the voltage scaling because it kernel panics kind of disturbs me just a little bit. It's one thing to deny something by warranty, but OPP5 is still technically within spec and if the kernel actually panics when run at it, it makes me fear deeper problems.
Not that I ever had any intention of running at OPP5, of course.

 

[MWeston]

His remark that they've removed the voltage scaling because it kernel panics kind of disturbs me just a little bit.

That's not what I meant. It panics because it is running on newer hardware and it is the old, very first kernel to ever touch Pandora. It was the only one to ever have voltage control. It requires the code to be modified to match newer board revisions and I don't know where to look and didn't spend any time trying to find out. There were major design changes from rev1 to rev2.

It crashes while the board is booting and when it gets to the UART init step (even though UART is already working from the uboot setup). You don't even get a chance to change the voltage.

 

[WizardStan]

It crashes while the board is booting and when it gets to the UART init step (even though UART is already working from the uboot setup). You don't even get a chance to change the voltage.

Oh, I get it. It's the old kernel (which had voltage scaling) that doesn't work on the new boards, and you simply haven't activated voltage scaling in the new kernel? Ok, that's reasonable and understandable and totally not what I thought you meant. I thought you meant that the act of adding voltage scaling to the kernel was causing it to panic, which is a very troublesome thought since it shouldn't crash simply because it's possible to change the voltage.
Thanks, glad to be mistaken.

 

[aq21]

What kernel version are we talking about here? What version of OMAP patches? Is the kernel source available somewhere?

 

[Monk]

His remark {...} kind of disturbs me just a little bit.

That's one thing about MWeston's posts that I don't recall ever worrying me, actually. With any post there is always - ALWAYS - the chance for misinterpretation, there is pretty much always "something" not said or not said in an ... unambiguous manner, but I don't recall ever feeling that MWeston was deliberately hiding things, deliberately misleading, or deliberately winding us up (with the possible exception of once where he looked EXACTLY like he was saying "nudge nudge wink wink, a nod's as good as a wink to a blind man". IIRC). He comes across as very... forthright? I think - I hope - if he knew the Pandora was borroxed in some way, he'd point it out and start looking for a way to un-borrox it.

Plus, I think I read him pretty much as he meant, more or less, which probably helped. It won't be that funny if OPP5, when it's tested, doesn't work, but I'm a not all that bothered by the insides. The performance counts more than the meal the actors ate before the play.

 

[urjaman]

What kernel version are we talking about here? What version of OMAP patches? Is the kernel source available somewhere?

git.openpandora.org

 

[notaz]

DVFS is not implemented in the kernel of first firmware release, hopefully that'll come with newer kernels and firmwares.

 

[LineOf7s]

Disclaimer: I know nothing.

With the remarkable difference in potential performance between the OPPs in the rev 1 boards (625 vs 900MHz), what cost does running the Pandora at OPP5 bring? Battery life? Heat? Loss of chip lifetime? All of the above?

Or are there other considerations my feeble mind is yet to comprehend? (this is likely)

 

[fusion_power]

DVFS is not implemented in the kernel of first firmware release, hopefully that'll come with newer kernels and firmwares.

Would be cool because battery life and underclocking is more important for me than Overclocking records. And it would be awesome if I no longer have to manualy set a CPU clock like it was onto the GP2X.
I wonder why this important techniques are not already in the general Linux kernels since years.