Eee Atom CPU control - testing

[jemimah]

eeecontrol .3 now works for me. I think it would be helpful to somehow specify that this isn't changing the cpu frequency scaling algorithm but rather the fsb speed. Setting it to powersave seems to save about a Watt on my machine. Might be nice to have the tool change both the fsb and the scaling algorithm in the same window.

[01micko]

Edit: Mick, if you have solved your problem and you can laugh again: Hey, my CPU runs faster than yours

a whole 2.49MHz! But that just may be the key I need Rolf. Because they are different I can differentiate. Thankyou

Jemimah, thanks for your report.

I'm not out of the woods yet though. Celeron owners, especially 700, 701 owners: I need your output from

Code: Select all

#cat /proc/eee/fsb

and

Code: Select all

#cat /proc/cpuinfo 

The first command must be run immediately after you have booted

Going by Rolf's results there maybe 4 cases to cater for, if not more.

(Atom, Celeron in Eee 900, 701, 701SD)

Thanks in anticipation.

[rhadon]

MHz Mick, Millions of Hz
Although I write this with a big smile I don't mean it serious.

~ Rollf

[01micko]

Atom testers,

On the first page of this thread mawebb88 and hokal reported that their Atoms, without any outside help, ie before we started playing with commands to alter the fsb, that their result of running "cat /sys/devices/platform/eeepc/cpufv" were "768". From that, we deduce that the Atoms start from boot in Performance mode. I'd like to see some tests after the gui is installed and the fsb altered to see if the Atoms still boot in Performance mode by default . I suspect this will be the case.

Help please

[01micko]

MHz Mick, Millions of Hz
Although I write this with a big smile I don't mean it serious.

~ Rollf

Thanks Rolf! (edited post)

[01micko]

Made a Eee-widget for Pwidgets. See http://www.murga-linux.com/puppy/viewto ... 837#348837

Cheers

[tempestuous]

The chipset in the Eeepc 701SD is in fact the Intel Celeron M 353. The cpu on these models defaults to 900MHz.

Thanks dawnsboy,
That explain's why rhadon's Eee 900 (Celeron M 353) is at 900 MHz at bootup.

Users on the forum at eeeuser.com report that they have been unsuccessful in overclocking it with the exception of those who have tried the Super Hybrid Engine (701SD apparently came with it installed) report being able to adjust clock speeds to 630MHz (underclock), 900MHz (default) and 960MHz (overclock).

OK, this confirms that the new 353-Celerons definitely should not be using the "/proc/eee/fsb" method for CPU FSB control.
They should be using the "/sys/devices/platform/eeepc/cpufv" method which is, in fact, SHE (Super Hybrid Engine) control.

Just to be clear here:
rhadon's Eee 900 and 01micko's Eee 701SD both have the new "353"-Celeron.
In theory, the Atom gui is the correct one for them. It uses the /sys/devices/platform/eeepc/cpufv
file for FSB speed.

The question remains: how do we differentiate between the two Celerons to help install the correct gui?

[tempestuous]

... OK, I just Googled for the output of "cat /proc/cpuinfo" of an Eee 700 (with older Celeron).
We already know that "model name" is the same as the faster Celeron
... but I see a difference:
"stepping : 6"
whereas the 353-Celeron has "stepping : 8"

Of course, Puppy users with older Celerons need to confirm this.

[01micko]

As soon as we can confirm the output from

Code: Select all

#cat /proc/cpuinfo

for an older Celeron a new gui will be ready.

A little strange the 2.49MHZ difference between rhadon's and my results but I guess voltages would be slightly different, different draws of current for different hardware and different power supplies.

[tempestuous]

The fan control script complains that the proc/eee directory does not exist.

First check that the eee module is installed -

Code: Select all

modinfo eee

If you see no result, go and install eee-0.2-k2.6.30.5.pet again. Be aware that I revised this dotpet on Oct 4 to include the depmod-FULL command at post-installation.

If/when you do see the eee module reported, next check that it's loaded -

Code: Select all

lsmod

If it's not loaded, go ahead and do so -

Code: Select all

modprobe eee

... but check /etc/rc.d/rc.local to see a "modprobe eee" entry, because all gui dotpets in this thread should create that line. It ensures that the eee module loads at bootup.

[01micko]

Hmmm..

Did some benchmarking ...

Doesn't look like there is any change..

I'm going to take a look inside Xandros (yuk!) to look for any clues

[dawnsboy]

Doesn't look like there is any change..

And there will not be a change if you are using an Eeepc 701 SD. Unlike its 701 series counterparts the 701SD is set at 900MHz by default.However you should find the Super Hybrid Engine on the default Xandros that should allow you to underclock to 630MHz, return to the default 900MHz or overclock to 960MHz.

Look for my post near the bottom of the page at this link: http://www.murga-linux.com/puppy/viewto ... 3&start=30 for more on this subject. Hopefully it will be of some help to you.

[01micko]

Thanks dawnsboy,

Read all the links you pointed me too and it certainly seems the only way to control the cpu freq on the 701SD is with SHE, (which looks like is owned by ASUS and probably closed source).

The p4-clockmod I suppose then is the stumbling block here? Xandros has a very old kernel, 2.6.1 but SHE does seem to work, no difference from performance to super-performance but powersaving shows a notable difference, however Puppy is faster overall. (No surprise there)

Cheers

[tempestuous]

it certainly seems the only way to control the cpu freq on the 701SD is with SHE, (which looks like is owned by ASUS and probably closed source).

No no.
The eeepc-laptop kernel module fully supports the SHE function, and exposes it as /sys/devices/platform/eeepc/cpufv

I started this forum thread with the express purpose of testing this new form of CPU FSB control.

In fact, with earlier versions of the eeepc-laptop module the FSB configuration file was called "she". From kernel 2.6.30 onward it's now called "cpufv".

[01micko]

Hmmm..... No difference with the benchmark though but it clearly works in xandros. Perplexing. Only probable answer is incompatibility with the p4-clockmod as reported by dawnsboy.

Can you or anyone recommend a good open source cpu benchmarking tool? There are a few, but I don't know where to start.

Cheers

[prehistoric]

@tempestuous,

I apologize if this post is verbose, but I'm trying to avoid misunderstanding. I went back and downloaded those pets just before I ran this yesterday to get that depmod_FULL change you mention. I'm holding the installation which showed the problem on an SD card, so I can slip it in and exactly reproduce things. I'm hoping this will keep it from being contaminated by separate experiments.

First, results of modinfo:

Code: Select all

:# modinfo eee
filename:       /lib/modules/2.6.30.5/kernel/drivers/acpi/eee.ko
description:    Support for eeePC-specific functionality.
author:         Andrew Tipton
license:        GPL
depends:        i2c-core
vermagic:       2.6.30.5 SMP mod_unload modversions 486 
#  
# modinfo eeepc_laptop
filename      /lib/modules/2.6.30.5/kernel/drivers/platform/x86/eeepc-laptop.ko
license:        GPL
description:    Eee PC Hotkey Driver
author:         Corentin Chary, Eric Cooper
alias:          acpi*:ASUS010:*
depends:        hwmon,rfkill
vermagic:       2.6.30.5 SMP mod_unload modversions 486 
# 

Here's my complete lsmod output:

Code: Select all

# lsmod
Module                  Size  Used by
parport_pc             29828  0 
lp                      9476  0 
parport                34508  2 parport_pc,lp
snd_pcm_oss            37440  0 
snd_seq_dummy           2608  0 
snd_seq_oss            27648  0 
snd_seq_midi_event      6892  1 snd_seq_oss
snd_seq                48464  5 snd_seq_dummy,snd_seq_oss,snd_seq_midi_event
snd_seq_device          6968  3 snd_seq_dummy,snd_seq_oss,snd_seq
snd_mixer_oss          15820  1 snd_pcm_oss
eee                     5188  1 
i2c_core               23776  1 eee
iptable_mangle          2380  0 
iptable_nat             4684  0 
nf_nat                 17888  1 iptable_nat
ipt_REJECT              2828  1 
nf_conntrack_ftp        7120  0 
nf_conntrack_irc        5136  0 
iptable_filter          2348  1 
xt_state                1836  4 
nf_conntrack_ipv4      13336  7 iptable_nat,nf_nat
nf_conntrack           63764  6 iptable_nat,nf_nat,nf_conntrack_ftp,nf_conntrack       _irc,xt_state,nf_conntrack_ipv4
nf_defrag_ipv4          1708  1 nf_conntrack_ipv4
ip_tables              11228  3 iptable_mangle,iptable_nat,iptable_filter
fan                     4048  0 
arc4                    1612  2 
ecb                     2508  2 
ath5k                 117296  0 
mac80211              166056  1 ath5k
led_class               4112  1 ath5k
serio_raw               5168  0 
cfg80211               64972  2 ath5k,mac80211
pcspkr                  2284  0 
snd_hda_codec_realtek   193584  1 
atl1e                  30688  0 
snd_hda_intel          24232  0 
snd_hda_codec          64620  2 snd_hda_codec_realtek,snd_hda_intel
snd_pcm                72496  3 snd_pcm_oss,snd_hda_intel,snd_hda_codec
snd_timer              20340  2 snd_seq,snd_pcm
snd                    56516  10 snd_pcm_oss,snd_seq_oss,snd_seq,snd_seq_device,       snd_mixer_oss,snd_hda_codec_realtek,snd_hda_intel,snd_hda_codec,snd_pcm,snd_time       r
soundcore               6912  1 snd
snd_page_alloc          8852  2 snd_hda_intel,snd_pcm
shpchp                 31616  0 
pci_hotplug            28140  1 shpchp
intel_agp              25788  1 
agpgart                34188  2 intel_agp
video                  18560  0 
output                  2796  1 video
battery                10032  0 
thermal                12712  0 
evdev                   9472  0 
eeepc_laptop           11680  0 
rfkill                 10672  2 eeepc_laptop
button                  5148  0 
hwmon                   2344  1 eeepc_laptop
processor              34592  1 
ac                      3952  0 
fuse                   53800  0 
aufs                  137092  1 
nls_iso8859_1           3724  1 
nls_cp437               5356  1 
usbhid                 26112  0 
usb_storage            51584  1 
squashfs               22928  2 
uhci_hcd               21564  0 
ehci_hcd               32856  0 
usbcore               138160  5 usbhid,usb_storage,uhci_hcd,ehci_hcd
#

Modification of rc.local

Code: Select all

# cat /etc/rc.d/rc.local
#this file called from rc.sysinit
#you can edit this file
#When firewall is installed, will append lines to this file...

if [ -x /etc/rc.d/rc.firewall ]; then
  /etc/rc.d/rc.firewall start
fi
modprobe eee
/usr/sbin/eee-fan-ctrl.sh &
# 

Does that cover all the questions?

[01micko]

Ok, sorry for getting a little off topic here.. but my pain with the p4-clockmod module does not exist, that is because the module is non-existent in 430-small.iso...(a bug reported in 430-small) (thanks anyway dawnsboy). The current compile of SHE is just not working on Puppy with the Eee-pc-701SD.

Back to the gui, which is the point here.. where are the other Celeron testers? I suppose this is an Atom thread so some are scared off. I will start a new thread to see if some folks with Celerons want to run some commands so I get a better picture. I may provide a minor upgrade in the meantime, with a few refinements.

Cheers.

[tempestuous]

prehistoric, the result from your "modinfo eee" command confirms that the eee module is loaded on your system.
So the fan control script should not complain about the /proc/eee directory being missing.

[01micko]

Ok testers,

Here is 0.4
Changelog
-changed "OK" button to close
-added a splash which shows up for a few seconds letting the user know something is happening
-changed the pinstall script to detect "stepping :6" (older Celerons) and will install the appropriate invoking script to /usr/bin... newer Celerons will use the atom GUI now (*rhadon's and my 701SD)

To Do
-there is no built in mechanism to detect what mode the machine is in at boot.
-it seems later Celerons are in "Normal" mode at boot, older Celerons "Powersave" and I'm not sure about Atoms.

I'm making a new thread for Celeron users, no older Celeron users have posted here so I'm just relying on tempestuous info and google.
Please offer any feedback

Enjoy

SEE NEWER POST V-0.5

[prehistoric]

prehistoric, the result from your "modinfo eee" command confirms that the eee module is loaded on your system.
So the fan control script should not complain about the /proc/eee directory being missing.

I agree, but I'm still getting the error reported earlier. I'm going to do a new installation on another SD card to test the 0.4 script in case this install has been corrupted during testing. We'll see what happens.