• After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.

Root For the Crazies out there: Super low volt kernel test

A

adrynalyne

Android Expert
http://dl.dropbox.com/u/2056318/1216_test2_voodoo_fascinate.zip


-25mV off top end.
-50mV off midrange.
-100mV off low end.

http://dl.dropbox.com/u/2056318/1217_adr_test2_voodoo_fascinate.zip



-50mV off top end.
-75mV off midrange.
-100mV off low end.

http://dl.dropbox.com/u/2056318/0101_adr_test_voodoo_fascinate.zip





based off jt1134's source. Yes, its voodoo. If you don't know how to install it, please don't. This is only for the experienced, seasoned, and otherwise slightly insane people out there.


Support is nada, this is JUST a test.
 
I'm actually testing nemesis 1.2 Ghz version he made using your same voltage settings right now. The 100-800 Mhz steps have your same degree of uv. Only his top step is 1200 vs 1000 of course. I will report back in like an hour. If his is stable I'm sure yours will be too. Or maybe I can try yours too after. :)
 
OK its late and Im tried so I hope this copy pasta is good enough for you adryn:
Its a portion of my mini review to nemesis this is the relative part dealing with the 800-100Mhz stuff where you guys are the same.
Basically its all good to go, except some times it almost seems like on conservative its waiting for fractions of a second to ramp up, this hesitation virtually disappears on interactive however

its solid and stable at all the 100-800Mhz increments[23:33] <saps> no problems
[23:33] <saps> however
[23:33] <saps> using the conservative governor
[23:34] <saps> at time there almost seemed to be moments when it seemed like the cpu was waiting for more juice as it accelerated up the chain from 100Mhz to a higher step
[23:34] <saps> I noticed this in things like just opening quadrant
[23:34] <saps> there would be what some might call lag
[23:34] <saps> just from pressing the icon to it actually opening
[23:35] <saps> now mind you this lag last a fraction of a second
[23:35] <saps> and once the app is cached its gone
[23:35] <saps> but using interactive governor eliminates this lag
[23:35] <saps> play movies was fine too
[23:36] <saps> now in my gameloft golf game Im on record saying only the 1.2Ghz-ul [1250mv] is completely stable even the 1275mv gives me freezes
[23:37] <saps> with this test kernel I was stable using conservative governor but not interactive
[23:37] <nemesis2all> i see
[23:37] <saps> using conservative the lag seemed to lag or hesistate more between screeens and even with some of the initial player animation
[23:37] <nemesis2all> are you feeling really brave?
[23:38] <saps> OTB ;)
[23:38] <nemesis2all> i'm making another attemp at steps
[23:38] <saps> so my personal feelings is the steps 800 and down are fine even with the tremendous uv, I would not go lower on the bottoms steps though as they seem to take fractionally longer ramping up
[23:39] <saps> using interactive does seem to help this aforementioned lag though, like a lot
[23:39] <saps> for me to run 1200Mhz in Lets Golf 2 I either need no more than 1250mv on that top step or to only use conservative governor
[23:40] <saps> I was about to test the 1.1 you recently released in Golf2 with interactive, normally I just use conservative

Sorry if thats too much info or not enough didnt feel like re-typing at 1230am
 
ive been running this for about an hour or so now...im gonna charge all the way up and wipe battery stats for battery life comparison

but for games, browsing, texts so far seems good and responsive :)
 
adrynalyne said:
Can't do much with the minimal feedback I am getting.
Click to expand...

Adryn....i've been running it since you released it....id have posted feedback sooner if i wasn't working. :)

But i've noticed some very minor lag when the processor revs up such as the start of baseball superstars and when long pressing on the homescreen immediately after waking the phone

When i flashed it around 3 pm est i was at 81% battery....now im at 68% and that includes 20 mind of gaming....10 or so minutes on the web and a few texts

So it seems to be working out ok....but id need to make it through a full day of use to really know

Performance otherwise is solid
 
I have not even had a chance to test this out yet myself but I know this morning on IRC it was reported going very well. Im still doing the drain test on nemesis kernel using the numbers from 12/16 test2. Pretty remarkable. I started watching the A team on my phone at 11pm last night prior to that I was at 72% [at what 16 hours off charger and medium use]. Using the conservative governor the cpu will run at 400Mhz for videos [vs full speed with interactive - Ive confirmed this]. So watching the 1h50m+ movie and drained it to 54%. The phone basically then slept like from 1am-7-8am and now Im at 52%.

So big time confirmation on the incredibly slow rate of drain when its sleeping at 100Mhz. I know the -100mv @ 100Mhz isnt changed from 12/16test2 to 12/17test2 which is also what nem has. This is a definite win win win for battery performance and everything
 
kernel performs very well when sleeping...stopped using at 8 am at 91% now at 11:50 am still 89% (same as when i checked at around 11am)
 
nitsuj17 said:
kernel performs very well when sleeping...stopped using at 8 am at 91% now at 11:50 am still 89% (same as when i checked at around 11am)
Click to expand...

And that's key. According to set cpu on the typical day I'm at minimum more than any other step
 
Hey Adrynalyne -

this kernel has been running great for us mere mortals as well ;-)

The wife and I are both running it for the average day to day things and it's great. She can totally see the changes in performance and in the improvements in the battery life.

We've been running since the 1216 and ultimately the 1217. Just fine.

Very smooth and consistent.

Also fwiw, we are not running setcpu for changing the gov. We are running with what ever the default is.

Thanks for the time spent on the kernel.

;-)
 
Yeah it seems the only time i use setcpu is to run stresstest and long bench when first flashing a kernel
 
On the n1 i did a interactive governor test. Do the search on Xda for Interactive Governor Test by me ... the battery life while sleeping is a key factor while using any kernel of course. but, the more important thing is your BACKGROUND DATA. I bet if you have bground data on and your losing 2% you might not lose but .5% with background data off.

I so badly want to try this kernel but i'm not up for going back to voodoo unless its a JT variant so I might try this one since you based yours from that... also, interactive governor on the n1 runs faster, all the time, that's why watching movies is lower battery usage on conservative.

conservative?!? why is that the governor choice here? what happened to on demand? someone needs to code the ondemand governor. its faster than conservative, and doesn't use as much battery from running fullspeed too much of the time (it jumps from the bottom step up to the top when you do something instead of needs-based. using this 1150 kernel from geeknik the only thing i want is flash... and him to stable this up a little more
 
akaine2001 said:
On the n1 i did a interactive governor test. Do the search on Xda for Interactive Governor Test by me ... the battery life while sleeping is a key factor while using any kernel of course. but, the more important thing is your BACKGROUND DATA. I bet if you have bground data on and your losing 2% you might not lose but .5% with background data off.

I so badly want to try this kernel but i'm not up for going back to voodoo unless its a JT variant so I might try this one since you based yours from that... also, interactive governor on the n1 runs faster, all the time, that's why watching movies is lower battery usage on conservative.

conservative?!? why is that the governor choice here? what happened to on demand? someone needs to code the ondemand governor. its faster than conservative, and doesn't use as much battery from running fullspeed too much of the time (it jumps from the bottom step up to the top when you do something instead of needs-based. using this 1150 kernel from geeknik the only thing i want is flash... and him to stable this up a little more
Click to expand...

theres nothing wrong with voodoo5 whatsoever, really theres nothing wrong with voodoo in general (including the old varieties) other than user error
 
akaine2001 said:
On the n1 i did a interactive governor test. Do the search on Xda for Interactive Governor Test by me ... the battery life while sleeping is a key factor while using any kernel of course. but, the more important thing is your BACKGROUND DATA. I bet if you have bground data on and your losing 2% you might not lose but .5% with background data off.

I so badly want to try this kernel but i'm not up for going back to voodoo unless its a JT variant so I might try this one since you based yours from that... also, interactive governor on the n1 runs faster, all the time, that's why watching movies is lower battery usage on conservative.

conservative?!? why is that the governor choice here? what happened to on demand? someone needs to code the ondemand governor. its faster than conservative, and doesn't use as much battery from running fullspeed too much of the time (it jumps from the bottom step up to the top when you do something instead of needs-based. using this 1150 kernel from geeknik the only thing i want is flash... and him to stable this up a little more
Click to expand...



I won't address this here, but you are completely wrong about how these governors work, and confused about SGS phones in general (where the ondemand governor is broken for ALL of them). You are welcome to swing by IRC and we can discuss it there. I don't want this to turn into a flame war.
 
akaine2001 said:
On the n1 i did a interactive governor test. Do the search on Xda for Interactive Governor Test by me ... the battery life while sleeping is a key factor while using any kernel of course. but, the more important thing is your BACKGROUND DATA. I bet if you have bground data on and your losing 2% you might not lose but .5% with background data off.

I so badly want to try this kernel but i'm not up for going back to voodoo unless its a JT variant so I might try this one since you based yours from that... also, interactive governor on the n1 runs faster, all the time, that's why watching movies is lower battery usage on conservative.

conservative?!? why is that the governor choice here? what happened to on demand? someone needs to code the ondemand governor. its faster than conservative, and doesn't use as much battery from running fullspeed too much of the time (it jumps from the bottom step up to the top when you do something instead of needs-based. using this 1150 kernel from geeknik the only thing i want is flash... and him to stable this up a little more
Click to expand...
both adryn's and nemesis' kernels were/are originial jt kernels just tweaked a bit with voltage and what not.
we dont have on demand for SF never have. jt tinkered with interactive to ramp up faster and down slower, its the best we got if you're looking for a combo of battery and performance.
you did kinda answer your own question as to why conservative is the governor of choice here = lower battery usage


FWIW using nem's latest test which now mimicks adryn 12/17 test2 volts 800 and down I didnt lose a single % on battery from 1am-7am while I slept and the phone slept this morning. However, by 830 I noticed I had missed 3 calls somehow and those 3 calls dropped my 3% since 7am. Overall, still the uber uv is clearly the way to go
 
What is interesting or will be interesting is to see who is willing to drive the floor down even lower. Im semi on record thinking -100mv at 100Mhz is as low as she can go based on my observations of the stuttering but who knows if 1200Mhz genuinely only needs 1250mv you'd think 100Mhz could maybe live with a bit less than 900mv. I suppose we'll have to wait and see. I will say even if we dont go any lower on voltage Im very pleased with the settings/voltage table of 12/17 test 2.

Having said we truly wont know how low is too low until we literally find the lower limit
 
saps said:
What is interesting or will be interesting is to see who is willing to drive the floor down even lower. Im semi on record thinking -100mv at 100Mhz is as low as she can go based on my observations of the stuttering but who knows if 1200Mhz genuinely only needs 1250mv you'd think 100Mhz could maybe live with a bit less than 900mv. I suppose we'll have to wait and see. I will say even if we dont go any lower on voltage Im very pleased with the settings/voltage table of 12/17 test 2.

Having said we truly wont know how low is too low until we literally find the lower limit
Click to expand...

knowing my fussy phone, this is probably as low as it will tolerate (and im surprised it is frankly) but finding the actual floor voltage-wise is always good :)
 
nitsuj17 said:
knowing my fussy phone, this is probably as low as it will tolerate (and im surprised it is frankly) but finding the actual floor voltage-wise is always good :)
Click to expand...
Agreed. Geeknik has a test out with every 800 and down step at -100mv from stock so there is a bit more room to go down in the middle. Im inclined to think maybe even -125mv at the bottom step wont fly but if it does why the heck not. Again though if this is as good as its gets its already very very good
 
I'm not sure how to find the floor voltage. I dropped a kernel 450mV and it still ran...which tells me there is a hard wall to how far we can go before its ignored.
 
adrynalyne said:
I'm not sure how to find the floor voltage. I dropped a kernel 450mV and it still ran...which tells me there is a hard wall to how far we can go before its ignored.
Click to expand...
Agreed, the -100mv definitely works, I suppose someone will just have to try 125 or 150 till it dont worky or dont takey right ;)
 
You must log in or register to reply here.
Back
Top Bottom