Overclocking questions


I had assumed that just set the theoretical maximum, that same way you could set the maximum speed you could overclock to?
 
Well the mouse pointer returned of its own choice. Frelling bizarre... It would not come back yesterday, no matter what I changed on mouse settings, display settings, multiple reboots, etc.


Oh well, Linux just seems to be self-healing in some respects I suppose.


It does seem that OPP is constant, but it also seems, according to TI's website, that the OMAP scales its speed according to usage needs. I'm looking forward to checking out changes in batter usage. Running OPP 5, 1000 MHz now.
 
It does seem that OPP is constant, but it also seems, according to TI's website, that the OMAP scales its speed according to usage needs. I'm looking forward to checking out changes in batter usage. Running OPP 5, 1000 MHz now.
I also can't remember if it has actual CPU scaling or not, but I do remember that the NOP instruction (or something like it) uses basically no power: clocked to 1000Mhz executing nearly a billion NOP instructions uses about the same power as clocked to 14Mhz executing 14 million actual instructions (after accounting for a lot of hand wavy rounding errors)


"What's the point in using a lower clock then?" you ask.


Because not everything uses the no-power NOP instruction. If a program isn't written with idle time in mind, it could be using near 100% CPU just doing nothing, and it's better to use near 100% CPU doing a lot of things at 500Mhz than using near 100% CPU at 1000Mhz doing the same things in the same time but with twice the energy.
 
ED has said in the past that the opp scales and has always scaled with an increase or decrease in mhz.


Sorry, posting from my phone and can't find quote at the moment.
 
ED has said in the past that the opp scales and has always scaled with an increase or decrease in mhz.
MWeston also said that trickle charging the battery was normal behavior and wouldn't cause a problem. :p


ED builds the OS image, but I don't know how intimate he actually is with the kernel. It could just be he's guessing based on what he's heard as well.


Now if Notaz said that the OPP scales (presumably because he's the one that made it do that) then I think it's reasonable to take that as positive proof. :)
 
ED has said in the past that the opp scales and has always scaled with an increase or decrease in mhz.
MWeston also said that trickle charging the battery was normal behavior and wouldn't cause a problem. :p


ED builds the OS image, but I don't know how intimate he actually is with the kernel. It could just be he's guessing based on what he's heard as well.


While DVFS and SmartReflex are innate features of the OMAP chipset, they are not currently enabled in the current Linux kernel. The current kernel being used by the Pandora is quite old and does not support these features. It is my understanding, however, that there is currently an effort to get this updated but I do not think it will be anytime soon.


Here is more information about OMAP Power Management


And here is a PM response from EvilDragon from a few months ago on this topic:

EvilDragon said:
Sent 06 April 2011 - 03:08 PM

Natro, on 06 April 2011 - 10:41 AM, said:


I am by no means an expert in this area, but I have been very interested in exploring the OMAP power management capabilities and DVFS aspects of the OMAP3530. From what I can tell, the Pandora currently is not utilizing any of this capability. From further investigation, it seems the kernel used by the Pandora is old and may not even have the power management options available to it. Doing this may very well extend the already great battery life of the Pandora and possibly address other issues.


So my question is... when do you anticipate an update to the OS that will include a newer kernel and enabled OMAP power management options? I know building and shipping the Pandoras is priority #1 for now.


Well, I'm not working on the kernel, so building and shipping won't stop anything here.


notaz is the main guy working on the kernel, DJWillis the one on the OS.


There's already and Alpha version of the next big OS version being tested, featuring XFCE4.6 and the .37-kernel.


There are still some bugs to fix, but work is being done on that for a while already. Nobody knows when a final release will be though. There's still a lot to test and work on.

Now if Notaz said that the OPP scales (presumably because he's the one that made it do that) then I think it's reasonable to take that as positive proof. :)

The features I am referring to are the DYNAMIC (read: automatic and on demand) scaling of voltage and frequency. What Notaz is referring is his script that will change the OPP levels manually based on the STATIC frequency selected by the user. He is not stating that the OPP levels change based on current CPU usage or frequency.
 
The features I am referring to are the DYNAMIC (read: automatic and on demand) scaling of voltage and frequency. What Notaz is referring is his script that will change the OPP levels manually based on the STATIC frequency selected by the user. He is not stating that the OPP levels change based on current CPU usage or frequency.
When did Notaz say anything about scaling OPP?
 
The features I am referring to are the DYNAMIC (read: automatic and on demand) scaling of voltage and frequency. What Notaz is referring is his script that will change the OPP levels manually based on the STATIC frequency selected by the user. He is not stating that the OPP levels change based on current CPU usage or frequency.
When did Notaz say anything about scaling OPP?

It seemed that you were implying that he did in your earlier statement. My apologies if I misinterpreted that. I was just trying to state that the currently delivered Pandoras do not have the dynamic power management features enabled and cannot until the kernel is updated. Currently the only way the OPP levels are changed is manually by the user. I know there is a script that will lets you select your CPU frequency and will adjust the OPP levels based on the selection but both the frequency and OPP levels will not change based on use. Since Notaz is the one working on an updated Kernel, its entirely possible that he has some of these power management features working on his Pandora.
 
It seemed that you were implying that he did in your earlier statement.
Oh! No, I said "if". IF Notaz said it, I would take it as evidence to be true, since I trust him to have a very in depth knowledge of the kernel being used and any patches that were applied.
 
sorry to throw this off-topic but i don't want to make a new thread for this simple question: if i increase the OPP level >3 does that void the warranty? and does going over 1ghz void it too? (i can get 870 at OPP3 so i feel good about OC)
 
sony, as a moderator, you better have some quoting facts from ed or craig to back that up
 
Ok, I'll add. Keep it under 1Ghz. No higher than OPP5.
okay i'll go with that. i don't know how to set it higher than OPP5 in the GUI so i can't over-volt it too much.
sony, as a moderator, you better have some quoting facts from ed or craig to back that up
i thought there was a post by craig or ed saying all/most pandoras can go OPP5 and 1ghz without harming it. (it will reduce the life span) that has nothing to do about the warranty but i thought it was interesting, bummers i can't find it. :(
 
OPP5 and 1Ghz is the max we allow.


We have spoken about defaulting to OPP5 and 800Mhz in the past, but nothing has been decided.
 
if it dies in those 30sec then you just lost your pandora and if it breaks later you will have to pay for the repairs since it is no longer under warranty.
 
Back
Top