![]() |
Re: [Announce] kernel-power unstable v49
Thanks freemangordon, now i understand what it is going about. In situation, where we're close to optimal values, I always prefer a little "too high" voltage, than one "tiny-bit too low", even if it may appear only during certain, specific situations - but, AIUI, You're trying to achieve exactly same thing, 100% safe values, that are - at the same time - saving battery. The idea is very appealing - I wonder, how up times on one battery will look with dual-scud 3000 mAh I'm using ;)
Only one small question remain - do you think that - with new values - current idea of "race-to-idle" (i.e, that limits 500-900, are more power-efficient that 250-900 or 1250-900, due to device finishing tasks faster on 500-900, and setting on 0 for more time), are outdated with new Sr, i.e 125-900 is now more power-efficient (or will be, after finalizing SR vdd2 also)? "Common sense" still tells me, that it's better to finish task on 500 mhz, and sit on 0 mhz (saving on power leakage), than doing it longer on 125 mhz, even extra-low-voltaged. but, I may miss something, that's why I'm asking. /Estel |
Re: [Announce] kernel-power unstable v49
@freemangordon...
i tried lowering dsp rates but same results.. and one more thing...it seems like sr_vdd1 decreases the voltages too much as the video that starts playback smoothly starts stuttering in between, when i pause at that moment and resume it, the same sequelae is repeated and ultimately the same notice (saying media format not supported) pops up...the same video is played with my custom profile 250-805 limits with dsp 520 at 805 and 430 at 720 and much reduced voltages i.e 50 at 805 and 720, 48 at 500 550 and 600... ps vdd1=0 vdd2=1 thanks.. |
Re: [Announce] kernel-power unstable v49
@knobby could you please lock frequencies one by one (starting @500) to check which frequency is problematic.
|
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
locking frequency = device is all the time sitting on one frequency, no matter of load. You can achieve this from kernel-settings. See:
http://wiki.maemo.org/Kernel_Power |
Re: [Announce] kernel-power unstable v49
yep.
kernel-config load dsp and after that ssh into device while playing video and start: kernel-config lock 500 kernel-config lock 550 ... kernel-config lock 805 and report on which frequency video stops EDIT: should we move on IRC? |
Re: [Announce] kernel-power unstable v49
Is KP profile voltage the minimum voltage that is used and SR enable higher voltages when needed (e.g SR cannot set VDD1 below profile voltages)?
Fast testing w/o SR (w SR) Incl. File browsing, web surfing, leafpad editing, etc... 17 -> OK (17) 16 -> OK (16) 15 -> OK (18) 14 -> OK (18) 13 -> OK (18) 12 -> OK (18) 11 -> OK (18) 10 -> reset Quote:
Use case: night time, phone not in use at all. EDIT: I'll do more testing with 125:12,90 setting. I'll make this a new default setting and w/o SR |
Re: [Announce] kernel-power unstable v49
It is the starting point, i.e. initial voltage.
EDIT: So it seems we can go as lower as 11-12 (~740mV). That are very good news re battery life |
Re: [Announce] kernel-power unstable v49
@freemangordon...
at 500 550 and 600 video is stuttering all the way but no such message (media format unsupported)popped up..(maybe it is normal bcoz of low dsp) but at 720 and 805 the video showed the same phenomenon but at 720 it was far less than 805... |
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
Quote:
EDIT: Is it HD video you are trying to play? |
Re: [Announce] kernel-power unstable v49
Quote:
EDIT: Night time: 125:12,90 500:30,360 (w/o SR) Day time: 500:30,360 850:49,600 (w SR) |
Re: [Announce] kernel-power unstable v49
Quote:
dsp rates.. 720=520 805=600 And yes i am using stock dsp profile without any modifications... i have a custom profile very similar to dsp profile but with reduced voltages and vdd1=0 it has.. 500=48 360 550=48 400 600=50 430 720=50 520 805=52 600 850=52 600 900=54 600 vdd1=0 vdd2=1 sampling rate=300000 upthreshold factor=95 ign nice load = 0 power saver bias=0 this plays all videos smoothly.. IMHO i think vdd1 is the confounding factor and yes it hd 1280x720 xvid |
Re: [Announce] kernel-power unstable v49
@knobby: The problem is that your DSP cannot handle much overclocking.
Ok, could you please make a custom profile based on dsp, make maximum frequency to be 900 and change DSP frequencies as follows (leaving SR enabled): 720 - 430 805 - 500 850 - 560 900 - 600 and try again. |
Re: [Announce] kernel-power unstable v49
@freemangordon
If it's of any assistance, I was just testing my voltage limits. 500/360=28 fully stable, 27 gave a reboot, but I keep mine at 29 550,600/520=39 fully stable, 38 gave playback error, no reboot though; i keep it at 40 I'll be testing 250 and 125 and let you know how it'll fare ;) EDIT: SR is disabled @Gusse What are your nightime min max? |
Re: [Announce] kernel-power unstable v49
Quote:
I WANNA GIVE U A TIGHT HUG FREEMANGORDON.... video played flawlessly at 850 and 900 though stuttered at 720 and 805... awesomely awesome playback at 850 and 900...no irritating messages and not a single framedrop... thanks a lot... :) |
Re: [Announce] kernel-power unstable v49
Here running at 125:10,90
will report tomorrow, leave it this way for a night with a charger plugged in. EDIT usually running at this Code:
current kernel configuration: |
Re: [Announce] kernel-power unstable v49
Quote:
/Estel |
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
Quote:
550/520=36 600/520=39 I tested also 550 at 36, immediate error, doesn't even think about it. 39 for 600 is stable as I said in the post above. 33 for 500 seems pretty high, though. |
Re: [Announce] kernel-power unstable v49
Quote:
*edit* , i see on the kp49 from 1st november that no bridgedriver is supplied in kernel-power-modules_2.6.28-10power49_armel.deb so is this new addition needed for my 720p ? |
Re: [Announce] kernel-power unstable v49
@Estel
supl.google.com works. I guess Elop is really trying to kill the N900 by taking them away under warranty and now canceling a-gps server. Good thing it's not KP fault ;) |
Re: [Announce] kernel-power unstable v49
Quote:
In short, don't try some exotic frequency profiles with SR enabled, it won't work. Use dsp profile and if it is not stable re videos, move DSP overclocking to higher frequencies(look at knobby's posts). Anyway thanks for your help testing SR. |
Re: [Announce] kernel-power unstable v49
Quote:
BTW, I don't use SR, just tested for you. Oh....and you're welcome! |
Re: [Announce] kernel-power unstable v49
Quote:
These are sr_vdd1_voltage values: 250 MHz sr_vdd1_voltage 23-34 500 MHz sr_vdd1_voltage 36-48 550 MHz sr_vdd1_voltage 35-54 600 MHz sr_vdd1_voltage 38-60 720 MHz sr_vdd1_voltage 52-60 805 MHz sr_vdd1_voltage 48-60 850 MHz sr_vdd1_voltage 52-60 900 MHz sr_vdd1_voltage 53-60 How can I check what voltage is best to use, since current OPP voltage as calculated by SR is lower than set by dsp profile? Thanks in advance... |
Re: [Announce] kernel-power unstable v49
Quote:
BTW are you sure you wrote correct values for 500(36) and 720(52)? |
Re: [Announce] kernel-power unstable v49
I HAVE A BUG!!!!!!
A SERIOUS ONE TOO. where is the freemangordon beer fund? |
Re: [Announce] kernel-power unstable v49
Turning display on from side switch at 125M with low voltages cause reboot (32wd_to).
Voltages below 20 seems to be problematic when device is waked with side switch. Sometimes it wake -up OK, but time to time nothing happens on screen and 32k WD kick-in causing reboot. Also if side switch is pulled several times (display on & off) then this is also causing reboot. Otherwise 125:12,90 is OK if you won't let it go sleep/deep-sleep or what it is doing when display is off. EDIT: Verified with 2 devices. No matter if SR is enable or not. |
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
@freemangordon
With your expertise and all, would you mind explaining as to why I should leave dsp overclocking for higher frequencies like knobby when it actually works on my device at low? I even had 600/600, perfectly stable, but didn't see much of a difference with 520, so lowered back. Thanks in advance. |
Re: [Announce] kernel-power unstable v49
Quote:
I'm also not 100% sure about risk of filesystem corruption because of "sneaky" undervoltage (one that doesn't result in reboots, but may result in CPU calculating with errors), but I do believe, that I encountered it once, long time ago. Despite being careful with things I install, many things started to act bizarrely, or just get plain broken, and it was getting worse over time. I was using so-called "ideal" undervolting profiles - without random reboots - and it was one and only time, when I had to reflash - after that, installing the same packages etc, but using custom profile with much higher voltages, I *never* encountered such problems again. So, as I've said - I can't be 100% sure, but it seems to me that it isn't urban legend. Saving battery life is cool, but lets don't get over-excited with it, and remain in sane, 150% stable values, +1 value up, just in case ;) /Estel |
Re: [Announce] kernel-power unstable v49
seems stable for me with 250-900 with SR enabled.
Could someone explain how the initial votages used - eg starving profile or ulv etc etc) will effect the smart reflex? cause my ideas on SR is that it intelligently adjusts the voltage? thanks in advance |
Re: [Announce] kernel-power unstable v49
@freemangordon
I've successfully enabled/used 720p in the very beginning of the announcement, all was fine, just battery started to drain too fast (~10-12 hours max with moderate use). Therefore I restore my previous kp48 and everything was fine, but without 720p of course. I can see that you've managed to fix SR, so it suppose to save some battery while kp49 + 720p, so I decide to try again with 720p/kp49. I can see everything playing fine. I'm still measuring my battery duration, but I'm optimistic. I want to help you, so this become a stable change, here are some info that could be in use for you. Let me know if you need more info, I would be glad to share or test anything. I've tried playing 720p content @500 and @805, both played ok: # cat /proc/cpuinfo Code:
Processor : ARMv7 Processor rev 3 (v7l) # cat /sys/power/Efuse Code:
0089a5b5 # cat /sys/power/idcode Code:
IDCODE: 4b7ae02f # kernel-config show Code:
current kernel configuration: |
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
Quote:
|
Re: [Announce] kernel-power unstable v49
Quote:
Second:then comment and finally:nothing personal but very stupid opinion |
Re: [Announce] kernel-power unstable v49
That is offtopic here, please stop it.
|
Re: [Announce] kernel-power unstable v49
Ok.No problems
|
Re: [Announce] kernel-power unstable v49
Quote:
|
All times are GMT. The time now is 10:03. |
vBulletin® Version 3.8.8