Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
I managed to squeeze an extra 1h .. 2h out of my 13" MBA's battery using CoolBook to undervolt the CPU, not sacrificing performance.

During my normal usage (alternating between browsing lightweight .. heavy web pages, watching 360p .. 720p YouTube videos (Flash, not HTML5), watching 720p / 4 mbps H.264 videos via VLC, coding in Terminal):

Before:
  • Battery life: barely over 4 hours
  • Core temperature: 50 .. 80°C
  • Fan speed: 2000 .. 6200 RPM
After:
  • Battery life: 5 .. 6 hours easily (that's +25 .. 50%)
  • Core temperature: 45 .. 65°C
  • Fan speed: 2000 .. 4500 RPM
Beware: Reaching such a gain isn't as easy as downloading and running the program. It involves a lengthy testing process consisting in determining the lowest voltage that the particular MBA's CPU can run with (at each SpeedStep frequency) while remaining stable under maximum stress load. It took me 1h to complete, but the longer the better (to ensure stability over extended periods of stress load).

Each CPU is unique, some are better than others at running at lower voltages, so there's no magic set of parameters to enter for each configuration.


1. CoolBook:

Buy, download and install CoolBook: unfortunately, I don't know any other way to tweak CPU voltage.

It's not too expensive given the benefits: $10, or $12.5 for us lucky EU residents… Bear in mind that the license is strictly tied to the one single computer that the serial is first entered in. A pity.

screenshot20110715at733.png


The GUI is not intuitive. Nor does it always behave as expected. I won't go into details, anyone can figure out the shortcomings and workarounds. But once I got the hang of it, I managed to make it behave as desired.

screenshot20110715at711.png


Thankfully, given the tasks it handles, it's quite clean. Composed of 3 components:
  • CoolBook2.kext: kernel extension (permanent)
  • CoolBookSwitcher: daemon process (permanent), runs as root, but that's acceptable given the low level settings it manages
  • CoolBookController.app: GUI (temporary)
It's transparent resource-wise:
  • Doesn't clobber the status bar nor the dock
  • The daemon is stable: no crashes in 2 days so far, monitored by launchd anyway
  • Negligible memory footprint and free of memory leaks: stuck at 664 KB since launch

2. Configuration: finding out optimal voltages:

screenshot20110715at852.png


a) Take note of the maximum voltage out of all default frequencies listed. In the example above (CoolBookController window): 1.0875V.

b) Clear the list, and re-add them one by one (tick B/2 for the lower ones) with the same voltage for each: the max as noted in a). Click "No" when prompted to perform a recommended stability test.

c) Run CPUTtest with Test type = small / Instances = 2 (number of cores) / Repetitions = 10 / default checkboxes. If the test passes, repeat with Test type = big.

d) If the test passes again, decrease all voltages by one notch and repeat c) until failure, to determine the lowest stable voltage. Keep track of results in a spreadsheet.

e) Move on to the the next highest frequency and repeat c) to e) until reaching either the lowest frequency, or lowest voltage (because, for example, if your lowest voltage is 0.925V and the CPU is stable at 1862 MHz, then it's also going to be stable at 1596 MHz).


3. Final settings:

The final settings should end up looking something like the following (those are for my 13" MBA):

screenshot20110715at728.png


(Exact same settings in the "Battery" tab.)

$12.5 well spent :cool:. Though I wonder if CoolBook's author is still maintaining his application. The last update is 4.5 months old at the time of writing. And so far, incompatible with successors to Core 2 Duos (incompatible with Core i3, i5, i7, first and second generation [Sandy Bridge]). At that pace, I doubt it's going to be compatible with the the refreshed MBAs (rev. E) for months, if ever :rolleyes:.
 
Last edited:

rmbrown09

macrumors 6502a
Jan 25, 2010
942
0
USA
Damn want to try this out, just sold my Air though and unfortunately CoolBook doesn't work on i5's and i7's (I'm 99% sure, that was the case when I emailed them like 3 months ago) and since the new Airs will have them, this won't work.


That is a really long sentence.
 

PaulWog

Suspended
Jun 28, 2011
700
103
Wow I didn't know Macbooks supported modifications of the clockspeeds & voltages of the hardware.

I'll definitely keep this piece of software bookmarked. It's amazing how much voltage a CPU can automatically set itself to, compared to what it actually needs.
 

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
Damn want to try this out, just sold my Air though and unfortunately CoolBook doesn't work on i5's and i7's (I'm 99% sure, that was the case when I emailed them like 3 months ago) and since the new Airs will have them, this won't work.
Yes, the inability to tweak the CPU voltage is something to keep in mind when considering the new MacBook Airs.

I'll definitely keep this piece of software bookmarked. It's amazing how much voltage a CPU can automatically set itself to, compared to what it actually needs.
Exactly. I'm always baffled at how much CPUs can be undervolted at stock frequencies. Same thing on desktop CPUs.
 
Last edited:

Bigmacduck

macrumors regular
Feb 15, 2009
228
5
I confirm CoolBook works great with Core 2 Duos

I can confirm that CoolBook controller works perfectly well on my MBA 2009 2.13GHz Intel Core 2 Duo. With the lower voltage the CPU power consumption and therefore the heat generation are significantly reduced. It runs much cooler and battery lasts longer.

I just use 4 frequency steps as this is more than sufficient. Here are the voltages that work for me:

1064 MHz 0.9000 V
1596 MHz 0.9250 V
1862 MHz 0.9250 V
2128 MHz 0.9250 V

Core i3 / i5 / i7 unfortunately do not support under-volting by design.
 

Bigmacduck

macrumors regular
Feb 15, 2009
228
5
I'm always baffled at how much CPUs can be undervolted at stock frequencies. Same thing on desktop CPUs.

Reason is that computer manufacturers use the Intel specified voltages. At these levels the CPU's are guaranteed to work under all circumstances. Most CPUs also run well with lower voltages and many work also with very low voltages. Problem is that the words "most" or "many" do not mean the same thing as "all".

Due to manufacturing tolerances, as subtile they might be, every CPU behaves differently when operated outside the specifications for voltage, frequency, temperature, altitude, etc.

Neither Apple, HP, Dell or any other manufacturer wants to risk that 10% of their PCs crash because of aggressive under-volting, that's why they stick to the specs.
 

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
@Bigmacduck
2.13 GHz at 0.925V! Plus you can go as low as 0.9V. You got a good one :D.
Reason is that computer manufacturers use the Intel specified voltages. At these levels the CPU's are guaranteed to work under all circumstances. Most CPUs also run well with lower voltages and many work also with very low voltages. Problem is that the words "most" or "many" do not mean the same thing as "all".

Due to manufacturing tolerances, as subtile they might be, every CPU behaves differently when operated outside the specifications for voltage, frequency, temperature, altitude, etc.

Neither Apple, HP, Dell or any other manufacturer wants to risk that 10% of their PCs crash because of aggressive under-volting, that's why they stick to the specs.
OK. That's what I imagined happened but since Intel already test each individual piece during their binning process, I think they reach the optimal voltages at some point before failure at a given frequency. Though even if that was true, I don't know how they would go about storing these settings and have them applied in the end host system.
 
Last edited:

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
@Bigmacduck
OK. That's what I imagined happened but since Intel already test each individual piece during their binning process, I think they reach the optimal tensions at some point before failure at a given frequency. Though even if that was true, I don't know how they would go about storing these settings and have them applied in the end host system.

Yeah, you hit it on the head why this isn't done from the factory. Intel would have to pass this info onto the OEMs to implement, and OEMs would need an OS + BIOS/EFI to make it stick.

Plus you have the public behavior once you start finding out that yours doesn't get quite as good as the neighbor's CPU because of normal variances in manufacturing. Much better to produce a consistent product and let the tweakers try to get every last drop out instead of shipping an inconsistent product that gets you backlash, even if all your estimates were conservative.
 

jackyyeow

macrumors 6502
Jul 5, 2011
262
0
The ability to undervolt the CPU with coolbook is one of the strong reason why I bought my MBA last 2 week instead of waiting for the new MBA.

I tried the lowest voltage on my ultimate on all clock speed, so far haven't encounter any errors. I'll wait until I see error message pop up then only I'll raise the voltage a little.

As for battery life, doing light surfing with clicktoflash, about 40%~50% brightness, it gives me about 3.5~4 hours when I'm on about 50% brightness. I believe I'll get more if I set throttle on CPU speed so it'll not exceed the baseline speed 1.86GHz or even down to 1.6GHz I can get some extra 10%. Didn't do proper test though, I'm happy with the result.
 

jackyyeow

macrumors 6502
Jul 5, 2011
262
0
Just want to update about the battery life with coolbook installed. With 6 bars of brightness, bluetooth off and AirPort on, I'm getting about 10mins out of 2% juice, so that's about 50mins 10% juice, by just doing light surfing and using clicktoflash. Sometimes I surf for like 30mins and I'm seeing the battery drop only 5%, and that'll actually matching the battery usage of iPad!

Pretty darn impressive piece of machine, together with minimal battery usage when sleep, this machine is like a dream to use!:apple:

PS: I don't have reference comparing before and after coolbook, it's one of the first app I purchased and installed immediately when I first power on the machine.
 

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
@jackyyeow
Thanks for sharing your experience regarding battery life with CoolBook installed.

"clicktoflash" is a lifesaver (now part of Chrome as an "experiment", see about:flags). I can't stand those crappy Flash ads that nobody even skims, yet they suck more battery to render than most regular activities.
 

jackyyeow

macrumors 6502
Jul 5, 2011
262
0
@jackyyeow
Thanks for sharing your experience regarding battery life with CoolBook installed.

"clicktoflash" is a lifesaver (now part of Chrome as an "experiment", see about:flags). I can't stand those crappy Flash ads that nobody even skims, yet they suck more battery to render than most regular activities.

No problem! I'm new to Mac so I just use safari at the moment. Flash block on FireFox doesn't seems to work so well so I just stick to safari. Never try chrome though I think I'll trade battery life for convenient and speed (if it ever applies for Chrome lol, I think I don't miss anything switching from PC's FireFox to Safari so far).

Interesting to note that when idling or when I'm typing this reply, the machine only sucking 5.x watt of battery according to coconutbattery application, and would go up to about 6~8+ watt depends on the web activity but mostly 7+- watt, light youtube will result about 9watt of usage. I assumed this is very low power usage? I've read about Sony X series Atom processor using about 6~8 watts, I think the C2D is actually more power efficient despite having bigger screen.
 

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
No problem! I'm new to Mac so I just use safari at the moment. Flash block on FireFox doesn't seems to work so well so I just stick to safari. Never try chrome though I think I'll trade battery life for convenient and speed (if it ever applies for Chrome lol, I think I don't miss anything switching from PC's FireFox to Safari so far).
I'm sorry, I'm not sure I understand your comment about Chrome: are you saying it's slower and less convenient than Safari?

Either way, I recommend you try it, you might fall in love with it like I did :D (on OS X, Windows and especially Linux where it's lightning fast). Also, you might want to watch your memory usage with Safari and Firefox, they both leak memory big time. Chrome doesn't since each tab is isolated in its own process.

Interesting to note that when idling or when I'm typing this reply, the machine only sucking 5.x watt of battery according to coconutbattery application, and would go up to about 6~8+ watt depends on the web activity but mostly 7+- watt, light youtube will result about 9watt of usage. I assumed this is very low power usage? I've read about Sony X series Atom processor using about 6~8 watts, I think the C2D is actually more power efficient despite having bigger screen.
Oh that's impressive (if coconutBattery is accurate). This reminds me that the new 13" MBAs are likely to get 17w processors (this TDP includes the GPU). If we compare that to the 17w C2D + 10-20w 320M (no definite number on Google) of the current revision, and taking into account that the SB processors are 30-50% more efficient clock for clock, then this new revision is going to have one heck of a battery life! I reckon 10+ hours for the 13".
 

paulyras

macrumors 6502a
Dec 3, 2006
504
19
Singapore
Yep. I've used this since the revA MBA. It's fantastic. It probably adds on average about 45 min to an hour to my battery life (very unscientific tests).

Bummer about lion though. I'm anxiously awaiting an update there.
 

Dowjohnny

macrumors 6502a
Jul 5, 2011
508
248
Germany
..

This thread reminds me of a question i always wanted to ask here, and since i assume this thread is read by battery life experts i simply dare to ask in here:

What is the best way to keep the battery of my MBA in good shape when using it at home on my desk?

I use it a lot over the day (5h) when i'm at home (mostly on my desk) an wonder, if its best to keep it connectet to the magsafe during that time an just run it all day long or to disconnect as soon as its fully loaded and only reconnect when the battery is completely empty?

Just wondering becaus if its connectet to the powerline the battery is never discharged, will this harm it? Maybe i'm wrong, but i somehow have in mind, that batteries should be emptied completely regulary (i do this on my ipad e.g.)

Can someone give me a short help? Thx


(PS: Sry for my english, i am no native speaker ;) )
 
Last edited:

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
@Dowjohnny
I have always wondered about the exact same questions so I am not in a position to answer them, sorry. There's an official article by Apple that you must have read, but just in case:

"Apple Portables: Calibrating your computer's battery for best performance"

The main thing to remember:
The battery has an internal microprocessor that provides an estimate of the amount of energy in the battery as it charges and discharges. The battery needs to be recalibrated from time to time to keep the onscreen battery time and percent display accurate and to keep the battery operating at maximum efficiency. You should perform this procedure when you first use your computer and then every few months after that. If you normally leave your Apple portable computer connected to AC power and very rarely use it on battery power you may want to perform this process once a month.
The rest of the page explains how to perform the calibration.
 

jackyyeow

macrumors 6502
Jul 5, 2011
262
0
I'm sorry, I'm not sure I understand your comment about Chrome: are you saying it's slower and less convenient than Safari?

Either way, I recommend you try it, you might fall in love with it like I did :D (on OS X, Windows and especially Linux where it's lightning fast). Also, you might want to watch your memory usage with Safari and Firefox, they both leak memory big time. Chrome doesn't since each tab is isolated in its own process.
Sorry I misread about your post and was under the impression that Chrome didn't handle flash block well lol. I meant in my last post that if Chrome is faster and more convenient but not handling flash content well then I'd rather use safari. I'm still experimenting more alternatives of the applications that came with factory install. I'll definitely give chrome a try.



Oh that's impressive (if coconutBattery is accurate). This reminds me that the new 13" MBAs are likely to get 17w processors (this TDP includes the GPU). If we compare that to the 17w C2D + 10-20w 320M (no definite number on Google) of the current revision, and taking into account that the SB processors are 30-50% more efficient clock for clock, then this new revision is going to have one heck of a battery life! I reckon 10+ hours for the 13".
Doesn't seems to be always the case, I think I've read that user's report real life usage of last year's 2.4GHz C2D on MBP 13" with 320m giving 10 hours of surfing, while not possible with 2011 MBP 13".

Some results from Anandtech too:
http://www.anandtech.com/show/4205/the-macbook-pro-review-13-and-15-inch-2011-brings-sandy-bridge/15
The SB MBP proven that the CPU is much more efficient than C2D, but C2D still wins out in light usage test. That's stock setting I'm supposed, however there's this coolbook, C2D should be able provide more battery life albeit slower CPU for intensive tasks. Notice how much battery life you can get out of MBA 13" doing light surfing and listening to music!

Interestingly, the HD3000 seems to contribute in big improvements on battery life when it comes to video watching.





What is the best way to keep the battery of my MBA in good shape when using it at home on my desk?

I use it a lot over the day (5h) when i'm at home (mostly on my desk) an wonder, if its best to keep it connectet to the magsafe during that time an just run it all day long or to disconnect as soon as its fully loaded and only reconnect when the battery is completely empty?

Just wondering becaus if its connectet to the powerline the battery is never discharged, will this harm it? Maybe i'm wrong, but i somehow have in mind, that batteries should be emptied completely regulary (i do this on my ipad e.g.)
I'm no expert, but battery should have regular charge cycles to maintain the health. You can use the battery every 3~4 days, discharge it and charge it up again. My PC laptop's battery died after less than a year of holding the charge constantly.
 
Last edited:

Roman2K~

macrumors 6502a
Original poster
Mar 11, 2011
552
16
OP, what do you mean in instruction number 2B by "(tick B/2 for the lower ones)"?
On my 2.13 GHz C2D, the lowest non-idle (> 798 MHz) candidate frequency listed in the drop down menu was 1596 MHz. In order for this menu to list frequencies between 798 and 1596 MHz, I had to tick the B/2 checkbox next to the menu.

I don't know what frequencies you have with your 1.4 GHz C2D but I'm sure ticking B/2 will enable more intermediate frequencies.

And will this make as much of a difference on a 1.4 ghz cpu (geez, sounds like I have a g4 typing this)?
If you manage to lower voltages, especially across the upper range of frequencies, you will most certainly notice a difference in battery life. Once you got those, a good benchmark is to close all apps, launch a movie and compare estimated remaining battery life with and without CoolBook.
 

ZipZap

macrumors 603
Dec 14, 2007
6,112
1,467
The OP instructions really make no sense to me.

If I clear the list am I to add back all entries at once then test?

I dont understand what the purpose of the spreadsheet is...as in the instances is not configurable in coolbook

I am never prompted to perform a stability test

My assumption is that I need to find the lowest votage that is stable for each frequency tracking results in a spreadsheet. When I finish with one frequency, I need to clear the list and add the next highest.

Your instructions should be rewritten as they are not clear in this regard.
 

clyde2801

macrumors 601
The OP instructions really make no sense to me.

If I clear the list am I to add back all entries at once then test?

I dont understand what the purpose of the spreadsheet is...as in the instances is not configurable in coolbook

I am never prompted to perform a stability test

My assumption is that I need to find the lowest votage that is stable for each frequency tracking results in a spreadsheet. When I finish with one frequency, I need to clear the list and add the next highest.

Your instructions should be rewritten as they are not clear in this regard.

I think the spreadsheet is to keep track of the values being used to ensure what's stable.
 

-tWv-

macrumors 68000
May 11, 2009
1,583
2
Ohio
I've been using this for about 8 months. I love it, keeps the heat down and definitely helps battery life.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.