Saving battery power with underclocking, low voltage etc. - Page 13 - maemo.org

Active Topics

The N900 is now 15 years old! (9)
to Nokia N900 by Arno_11 - 5 hrs, 55 mins ago
Sailfish OS for the Motorola Moto G7 Power (XT1955-5) - (ocean) (4)
to SailfishOS by Maemish - 5 days, 18 hrs ago
more...

Page 13 of 20

Thread Tools

Tigerite	2010-05-07 , 14:52
Posts: 284 \| Thanked: 320 times \| Joined on May 2010 @ Peterborough, UK	#121

Originally Posted by Tigerite

That's what I thought too regarding 0 Mhz, however the 500 Mhz setting shown for me was completely different to any in FREQS which is rather perplexing (almost as if it isn't overwriting the previous setting)? As you say, I do hope titan can shed some light and/or fix the bug..

As per my reply on the overclocking thread, it's a bug only in show and is fixed by changing line 61 from:

Code:

    for f in $allfreq; do

Code:

    for f in 0 $allfreq; do

Easy peasy

Quote & Reply |

oskarmat	2010-05-07 , 15:03
Posts: 41 \| Thanked: 39 times \| Joined on Jan 2010 @ ireland	#122

e sti cazzi

Quote & Reply |

	hawaii	2010-05-07 , 17:32
	Posts: 1,030 \| Thanked: 792 times \| Joined on Jun 2009	#123

the staving conf causes immediate reboot on my device when I place or receive a phone call.

Will increase voltage levels and report back.

Quote & Reply |

	automagic68	2010-05-08 , 06:59
	Posts: 415 \| Thanked: 161 times \| Joined on Apr 2010 @ San Francisco, CA	#124

Originally Posted by tonism

Just reporting that my battery lasted ~32h with moderate usage (2.5G, always WiFi, ~5h heavy testing due making of starving conf, ~1h of mp3, little video recording and some phonecalls).

It seems to me that battery lasts little better than with default configuration. I would have expected 20-24h with default configuration. I'm using 125 - 1000 MHz starving conf.

Your battery lasted 32hours w/ 125-1000mhz? Min frequency at 500mhz or 125? I'm at 250-750mhz w/ battery at 36% after 10hrs. Should I change the minimum to 125 or 500? Also is 850 faster than 750 to the extent that processes would run faster and save me energy? Thanks again the starving conf its still running smooth.

Quote & Reply |

	tonism	2010-05-08 , 10:29
	Posts: 20 \| Thanked: 37 times \| Joined on May 2010 @ Estonia	#125

Originally Posted by automagic68

Your battery lasted 32hours w/ 125-1000mhz? Min frequency at 500mhz or 125? I'm at 250-750mhz w/ battery at 36% after 10hrs. Should I change the minimum to 125 or 500? Also is 850 faster than 750 to the extent that processes would run faster and save me energy? Thanks again the starving conf its still running smooth.

32h with 1 GHz starving conf (MINFREQ=125, MAXFREQ=1000): http://tonism.2bit.ee/N900/starving

Actually I'm not sure what to set as the minimum frquency. Since 125 MHz voltage is lower by pretty huge amount then I tried that one. But 250 MHz vs 500 MHz voltage difference is too small to favor 250 MHz.

And as for maximum speed, I think you're safe to use at least 850 MHz. Processes get done faster and CPU can resume to idle state quicker. If 1 GHz is stable and you won't mind exceeding the voltage limit where Igor said it is okay to hold your devices constantly then go for it. I'm using that one

After all, most of the time my device sits idle and when it needs to do anything it can handle the voltage just fine and does the task @ 1 GHz. Hell, the stock 600 MHz had even higher voltage!

No problem. I think we should consider ourselves among the few lucky ones who can run the starving conf stable. The statistics so far seem to be 50:50?

__________________
starving configuration for titan kernel

Last edited by tonism; 2010-05-08 at 10:33.

Quote & Reply |

titan	2010-05-08 , 11:22
Posts: 946 \| Thanked: 1,650 times \| Joined on Oct 2009 @ Germany	#126

Originally Posted by tonism

If 1 GHz is stable and you won't mind exceeding the voltage limit where Igor said it is okay to hold your devices constantly then go for it. I'm using that one After all, most of the time my device sits idle and when it needs to do anything it can handle the voltage just fine and does the task @ 1 GHz. Hell, the stock 600 MHz had even higher voltage!

There is no "voltage limit". high voltage is just one of many factors that can damage your device. lower voltage with high frequency can be equally bad as high voltage with lower frequency.

Quote & Reply |

The Following 2 Users Say Thank You to titan For This Useful Post:
moepda, tonism

	tonism	2010-05-08 , 13:05
	Posts: 20 \| Thanked: 37 times \| Joined on May 2010 @ Estonia	#127

Originally Posted by titan

There is no "voltage limit". high voltage is just one of many factors that can damage your device. lower voltage with high frequency can be equally bad as high voltage with lower frequency.

The limit, as I've understood, is (72 * 0.0125 V) + 0.6 V = 1.5 V, and is set by software. Hardware wise there is no limit. The limit I was referring to was that somewhere Igor said that do not overclock your devices over 500 MHz for constant use. Assuming he meant 500 MHz @ def voltages.

Hmm.. Thank you for this information. I was thinking that is the voltage danger affected by its speed. Would seem logical. So the "danger factor" goes linearly, basically voltage * speed?

Meaning that at default settings its 1.35 V * 600 MHz = 0.81 GVHz?

With my 1 GHz settings it would be then 1.2875 V * 1 GHz = 1.2875 GVHz.

If that is correct then the closest value for 0.81 GVHz would be something between:
700 MHz @ 1.1125 V = 0.77875 GVHz
750 MHz @ 1.125 V = 0.84375 GVHz

And the "safe" limit Igor was referring is then 500 MHz @ 1.2 V = 0.6 GVHz.

The closest settings with my conf would be then:
550 MHz @ 1 V = 0.55 GVHz
600 MHz @ 1.0125 V = 0.6075 GVHz

I'm sure there are other factors, like heat etc, but is the basic logic behind voltages and speeds like this?

In that case I should make new setups. starving550, starving700 and starving1150 in order for ppl who use them can have more reliability with increased speed and decreased battery consumption?

Also, titan, can you comment on the idle frequencies? Is it wise to idle @ 125 MHz? Or should I go for 500 MHz? Or 550 MHz with starving conf, perhaps, as the GVHz rating is lower than default @ 500 MHz? 250 Mhz, as seems to me, has too little voltage difference and 500 MHz wins the day by quickly doing idle tasks and resume to sleep state? Also, wasn't there a some sort of bug @ 125 MHz?

__________________
starving configuration for titan kernel

Last edited by tonism; 2010-05-08 at 13:20.

Quote & Reply |

	qwerty12	2010-05-08 , 13:12
	Posts: 4,274 \| Thanked: 5,358 times \| Joined on Sep 2007 @ Looking at y'all and sighing	#128

titan, kernel-power-settings 0.6 is missing "update-sudoers" in its postinst. Can you add it, please?

Quote & Reply |

titan	2010-05-08 , 14:49
Posts: 946 \| Thanked: 1,650 times \| Joined on Oct 2009 @ Germany	#129

please have a look at http://wiki.maemo.org/Overclocking#U...voltage_tables
voltage has a quadratic effect.
we only know the TI specs for their recommended voltage levels.
no one knows the real long-term effects of overclocking with undervolting.
we simply have no evidence and only extensive tests by TI would quality as sufficient evidence.

would you mind if I add your starving config to the templates in kernel-power-settings?

I had worse battery life with 125 than with 250 and 125 is supposed to be unstable,
so I'd not recommend to enable it.
It might be useful, however, if you setup some kind of deep-sleep state with all
connectivity disabled and all processed reniced or suspended.

Originally Posted by tonism

The limit, as I've understood, is (72 * 0.0125 V) + 0.6 V = 1.5 V, and is set by software. Hardware wise there is no limit. The limit I was referring to was that somewhere Igor said that do not overclock your devices over 500 MHz for constant use. Assuming he meant 500 MHz @ def voltages.

Also, titan, can you comment on the idle frequencies? Is it wise to idle @ 125 MHz? Or should I go for 500 MHz? Or 550 MHz with starving conf, perhaps, as the GVHz rating is lower than default @ 500 MHz? 250 Mhz, as seems to me, has too little voltage difference and 500 MHz wins the day by quickly doing idle tasks and resume to sleep state? Also, wasn't there a some sort of bug @ 125 MHz?