Anyone tried out Granola?

dcc24 · 2010-04-24 19:29:15

Although the algorithms behind it are closed sourced, they sure promise a lot. I'm talking about Granola (http://grano.la/) Did anyone try it out?

falconindy · 2010-04-24 20:19:15

If anyone's curious about it, here's a PKGBUILD, as the installer.bash that they provide will only support deb or RPM based distros.

# Contributor: Dave Reisner <d@falconindy.com>

pkgname=granola
pkgver=2.1.3
pkgrel=1
pkgdesc="Granola is software that improves the energy efficiency of your PC or laptop."
arch=('i686' 'x86_64')
[ "$CARCH" == "i686" ] && ARCH=i386
[ "$CARCH" == "x86_64" ] && ARCH=amd64
url="http://www.grano.la"
license=('custom')
depends=('cpufrequtils')
makedepends=('deb2targz')
source=("https://download.miserware.com/linux/deb/ubuntu/lucid/${ARCH}/${pkgname}_${pkgver}-0lucid1_${ARCH}.deb")
[ "$CARCH" == "x86_64" ] && md5sums=('5f8819f85fdcebb56961cb0e86e076de')
[ "$CARCH" == "i686" ] && md5sums=('792085d52e147c10e1e05773ef917bdb')

build() {
  cd $srcdir
  deb2targz "${pkgname}_${pkgver}-0lucid1_${ARCH}.deb"
  bsdtar xf "${pkgname}_${pkgver}-0lucid1_${ARCH}.tar.gz"

  cp -a usr var etc "${pkgdir}"
  mv "${pkgdir}/etc/init.d" "${pkgdir}/etc/rc.d"

}

# vim:set ts=2 sw=2 et:

edit: fix PKGBUILD to include arch specific md5sums.

Last edited by falconindy (2010-04-30 17:38:56)

Berticus · 2010-04-24 20:24:38

So how do you know if it's actually working?

dcc24 · 2010-04-24 20:25:19

Berticus wrote:

So how do you know if it's actually working?

Exactly. That's why I'm asking for any experiences

Labello · 2010-04-25 09:19:16

sounds very suspicious to me.

i think phoronix should put it to a test

Angry · 2010-04-30 02:57:36

It works great.

I installed granola + granola-gui + cpufrequtils.
I added granola as deamon to /etc/rc.conf, added modules: acpi-cpufreq cpufreq-userspace and disabled: !acpi-cpufreq
Was some guesswork, because their wiki and manual are nearly useless.

Had some warnings like:
- "granola[3278]: Error changing file ownership: miserware: no such group. Does this user exist?" -> Added group miserware
- or something else with permission for a file, dont remember exactly -> But starting granola as deamon solved it.

Then i tweaked some settings in /etc/granola.conf
The CPU acts perfect. It stays at min levels and rises fast when more CPU is needed. I didnt recognize any speed deficit.

Conclusion: I like it a lot!

hokasch · 2010-04-30 09:41:51

how is this different to just using cpufreq-utils? it looks like it does exactly the same.

Angry · 2010-04-30 10:45:56

I think there are no important differences, but u have a gui that shows you how much power u save.

murfMan · 2010-04-30 13:42:08

falconindy wrote:

If anyone's curious about it, here's a PKGBUILD, as the installer.bash that they provide will only support deb or RPM based distros.

# Contributor: Dave Reisner <d@falconindy.com>

pkgname=granola
pkgver=2.1.3
pkgrel=1
pkgdesc="Granola is software that improves the energy efficiency of your PC or laptop."
arch=('i686' 'x86_64')
[ "$CARCH" == "i686" ] && ARCH=i386
[ "$CARCH" == "x86_64" ] && ARCH=amd64
url="http://www.grano.la"
license=('custom')
depends=('cpufrequtils')
makedepends=('deb2targz')
source=("https://download.miserware.com/linux/deb/ubuntu/lucid/${ARCH}/${pkgname}_${pkgver}-0lucid1_${ARCH}.deb")
md5sums=('5f8819f85fdcebb56961cb0e86e076de')

build() {
  cd $srcdir
  deb2targz "${pkgname}_${pkgver}-0lucid1_${ARCH}.deb"
  bsdtar xf "${pkgname}_${pkgver}-0lucid1_${ARCH}.tar.gz"

  cp -a usr var etc "${pkgdir}"
  mv "${pkgdir}/etc/init.d" "${pkgdir}/etc/rc.d"

}

# vim:set ts=2 sw=2 et:

 makepkg
==> Making package: granola 2.1.3-1 i686 (Fri Apr 30 08:41:48 CDT 2010)
==> Checking Runtime Dependencies...
==> Checking Buildtime Dependencies...
==> Retrieving Sources...
  -> Downloading granola_2.1.3-0lucid1_i386.deb...
--2010-04-30 08:41:49--  https://download.miserware.com/linux/deb/ubuntu/lucid/i386/granola_2.1.3-0lucid1_i386.deb
Resolving download.miserware.com... 174.143.111.2
Connecting to download.miserware.com|174.143.111.2|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 104228 (102K) [application/octet-stream]
Saving to: "granola_2.1.3-0lucid1_i386.deb.part"

100%[=============================================================================================================>] 104,228      333K/s   in 0.3s    

2010-04-30 08:41:49 (333 KB/s) - "granola_2.1.3-0lucid1_i386.deb.part" saved [104228/104228]

==> Validating source files with md5sums...
    granola_2.1.3-0lucid1_i386.deb ... FAILED
==> ERROR: One or more files did not pass the validity check!

kokoko3k · 2010-04-30 14:02:23

Yep, it would be fine if someone which already uses other power-saving methods could give it a shot and report if there are some advantages.

Last edited by kokoko3k (2010-04-30 14:02:34)

falconindy · 2010-04-30 14:47:18

murfMan wrote:

==> ERROR: One or more files did not pass the validity check!

so run `makepkg -g >> PKGBUILD` and then makepkg again... my fault for not including the checksum, but it's hardly a show stopper.

edit: snip quote

Last edited by falconindy (2010-04-30 14:47:55)

murfMan · 2010-04-30 17:35:46

thanks ill give it another try

edit:

pkgbuild works fine now but its still not working:

granola[2998]: Version 2.1.3 is starting.
granola[2998]: Can't open lockfile /var/run/granola.pid: Permission denied
granola[2998]: Failed to start, error obtaining lockfile
granola[2998]: Already running or do we not have have permission to write to '/var/run/granola.pid'?

 sudo granola 
granola[3003]: Version 2.1.3 is starting.
granola[3003]: Using config file '/etc/granola.conf'
granola[3003]: Set LogLevel to notice
granola[3003]: Set Policy to MiserWare
granola[3003]: No match for processor '              Intel(R) Pentium(R) 4 CPU 3.00GHz' in power table '/var/lib/miserware/powertable.dat'
granola[3003]: Could not retrieve power table information for this CPU. Using CPU datasheet value for power estimations.
granola[3003]: Error opening scaling governor file '/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor' in read mode
granola[3003]: Is cpufreq enabled in this kernel and do you have a CPU which supports DVFS?
granola[3003]: Can't manage DVFS for any CPUs
granola[3003]: Warning, could not set up a configuration for managing the CPUs
granola[3003]: Failed to start, error initializing

is there something better / lighter than this? like a console app?

Last edited by murfMan (2010-04-30 17:40:07)

falconindy · 2010-04-30 18:53:24

Its supposed to run as a daemon. There's an init script included (which probably needs modification for Arch). Secondly, read the error message. They tell you exactly what the problem is.

I don't use this, nor do I have any interest (my kernel doesn't even support cpu frequency scaling). I just felt like scraping together a PKGBUILD for anyone who felt the desire to mess around with it.

Is there something "lighter" and more Arch-like? Absolutely, but that wasn't the point of this thread.

Angry · 2010-05-02 18:21:17

murfMan wrote:

thanks ill give it another try

edit:

pkgbuild works fine now but its still not working:

granola[2998]: Version 2.1.3 is starting.
granola[2998]: Can't open lockfile /var/run/granola.pid: Permission denied
granola[2998]: Failed to start, error obtaining lockfile
granola[2998]: Already running or do we not have have permission to write to '/var/run/granola.pid'?

I had that problem too. As mentioned: Add granola to deamons.

Lemak · 2010-05-05 19:16:03

Hey guys we've gotten some requests and have seen these posts, so we have just released a 32 and 64-bit tarball of Granola that should be more usable on Arch. We hope that you will share this with your friends and community. If you have any improvements to make Granola more friendly to Arch, please send them back. Unfortunately we do not have the resources to make packages for every Linux distribution, so we need your help.

The tarball release can be found at http://grano.la/help/#tarball.

The more users in the Granola community there are, the better off our global community becomes. One desktop running Granola will save almost as much carbon as one tree. While one tree does not make much of a difference, an entire forest makes an incredible impact. Likewise, as the Granola community becomes larger so does its impact. So please, share this software with your community. (http://grano.la/community/impact.php)

Thanks,
Geoff Zelenka
MiserWare, Inc.

Angry · 2010-05-05 22:49:19

Nice Instructions. granola and granola-gui worked without problems for me (x86_64). Thx and stay tuned!

lucke · 2010-05-05 23:39:04

Do you have any data how it actually compares to cpufreq_ondemand, Lemak?

hokasch · 2010-05-06 13:03:40

lucke, I asked the same question per email:

miserware wrote:

We do not currently have a benchmark against the "ondemand" governor. This is worth looking into, and I have forwarded your request to
the development team.

Kind of odd... The whole numbers thing seems to be a bit shaky. They do not mention if the reported savings are calculated against os-native frequency scaling or running with frequency scaling turned off. Also one has to realize that the reported "XX% saved" does not relate to overall power consumption of the machine, but just the cpu. On modern laptops, that share is not too big - think about lcd backlight, wlan etc.
I have yet to see a real power saving benchmark that "adheres to industry standards" to believe all these claims about their "sophisticated workload modeling and prediction algorithms".

edit: removed part I misread. also, they do indeed mention how the power savings are calculated: against frequency scaling off. Given on-demand scaling is enabled by default on Windows and probably most Linux Distros (e.g., Ubunut), I find this quite dishonest. They should report "you saved XX percent compared to if you would have turned off power management" then.

Last edited by hokasch (2010-05-06 13:57:58)

lucke · 2010-05-06 14:08:44

I truly wonder indeed if their proprietary algorithms do (and can?) provide a measurable advantage over inbuilt OSes' solutions. In the pdf with benchmarks posted on their site, they set up the power profile in Windows 7 to high, which, I guess, means disabled cpu throttling.

I'd like to know the advantage of using it instead of cpufreq_ondemand/Windows' cpu throttling driver.

Well, it if saves some watts indeed, I applaud their efforts.

dcc24 · 2010-05-06 14:12:45

Still, with a closed-source algorithm, we have no way of knowing what it actually does. It could very well be harmful.

Lemak · 2010-05-06 15:49:03

lucke wrote:

Do you have any data how it actually compares to cpufreq_ondemand, Lemak?

The only benchmarks that we have published are the Windows ones mentioned earlier in this thread. You can find them at http://grano.la/software/benchmark.php.

We will be publishing more benchmarks soon.

hokasch · 2010-05-06 16:17:11

Really interested in more benchmarks. Having looked at the site again, I am pretty convinced for the moment that this is basically marketing blah, not an opportunity to save power:

Features:

# Most effective power saving software for Linux

pm-utils + ondemand is less effective? really?

# Sophisticated workload prediction improves upon built-in CPU governors

tough claim without having done any benchmarks...

To arrive at total savings, they used a laptop with the display shut off. Now, with the lcd consuming a great share of energy, this makes the measurements pretty useless for total savings on laptops. Also, other hardware not turned on in the tests (e.g. wlan, dedicated graphics) will inadvertabily cut down the numbers. Since laptops are build for low energy consumption, one can not apply the measurements to desktop systems - the components will most likely contribute very differently to the overall power consumption (graphics,hdd..). In short, it is understandingly hard to come up with any numbers because of differing hardware, but they choose to measure a particulary unrepresentative (and favorable) hardware setup. So anyway, with this, they arrive at total savings between 2,5% and 18%, depending on workload, compared to frequency scaling turned off.

On the Website however, the claims differ a lot from these results:

If you mostly surf the Internet and check email, the savings will be substantial and probably 30% or more.

No benchmarks to support "30% or more", nor any tests for that usage scenario.

If you are playing the latest intensive graphic game without a high-end graphics card, the savings will probably be less; perhaps as low as 10%.

The benchmarks with Half Life 2 resulted in a 2,6% reduction.

The savings reported by the program itself are based on the power consumption of your CPU in full speed, although if the cpu is not in their tables (and they don't look awfully big), 90w tdp is assumed. If you have a cpu consuming less power (25w tdp for me...), the numbers will approach pure fantasy value.

Next they compute the amount of CO2 saved by the Granola community to be equivalent to 27702 trees. This seems to come from the assumption that every download is equal to the "average saving" of 56.9kWh that is to be achieved if the software is used on a system that runs 24/7 for one year. WTF?

Last edited by hokasch (2010-05-06 16:33:40)

voidzero · 2010-06-16 18:01:20

Hi, I recently found this thread and since I'm using my laptop a lot lately without a power outlet close by, I was wondering about using this. But I'm also concerned about the closed source drivers, because it really defies the FOSS ideal. What do you guys think about this closed source application? And like more people have asked, "is it really worth it"™?

ngoonee · 2010-06-17 00:30:04

Use laptop-mode/cpufreq and maybe phc. This thing sounds more marketing than software.

Anikom15 · 2010-06-17 01:07:52

This thing sounds more pointless than software.

Arch Linux

#1 2010-04-24 19:29:15

Anyone tried out Granola?

#2 2010-04-24 20:19:15

Re: Anyone tried out Granola?

#3 2010-04-24 20:24:38

Re: Anyone tried out Granola?

#4 2010-04-24 20:25:19

Re: Anyone tried out Granola?

#5 2010-04-25 09:19:16

Re: Anyone tried out Granola?

#6 2010-04-30 02:57:36

Re: Anyone tried out Granola?

#7 2010-04-30 09:41:51

Re: Anyone tried out Granola?

#8 2010-04-30 10:45:56

Re: Anyone tried out Granola?

#9 2010-04-30 13:42:08

Re: Anyone tried out Granola?

#10 2010-04-30 14:02:23

Re: Anyone tried out Granola?

#11 2010-04-30 14:47:18

Re: Anyone tried out Granola?

#12 2010-04-30 17:35:46

Re: Anyone tried out Granola?

#13 2010-04-30 18:53:24

Re: Anyone tried out Granola?

#14 2010-05-02 18:21:17

Re: Anyone tried out Granola?

#15 2010-05-05 19:16:03

Re: Anyone tried out Granola?

#16 2010-05-05 22:49:19

Re: Anyone tried out Granola?

#17 2010-05-05 23:39:04

Re: Anyone tried out Granola?

#18 2010-05-06 13:03:40

Re: Anyone tried out Granola?

#19 2010-05-06 14:08:44

Re: Anyone tried out Granola?

#20 2010-05-06 14:12:45

Re: Anyone tried out Granola?

#21 2010-05-06 15:49:03

Re: Anyone tried out Granola?

#22 2010-05-06 16:17:11

Re: Anyone tried out Granola?

#23 2010-06-16 18:01:20

Re: Anyone tried out Granola?

#24 2010-06-17 00:30:04

Re: Anyone tried out Granola?

#25 2010-06-17 01:07:52

Re: Anyone tried out Granola?

Board footer