[lm-sensors] better temp2 computation for asus CUSL2C

[lm-sensors] better temp2 computation for asus CUSL2C

[gimeshell at web.de]

Hi,

i'm using this line in sensors.conf for temp2 computation:
compute temp2 (@*30/43)+25, (@-25)*43/30

But resulting temperature is about 8 degrees too low.
Does anyone have a better temp2 computation for cusl2c?

I think it's not wise just to add 33 instead of 25 in computation. ;)

regards,
gimshell

[lm-sensors] better temp2 computation for asus CUSL2C

[Jean Delvare]

> i'm using this line in sensors.conf for temp2 computation:
> compute temp2 (@*30/43)+25, (@-25)*43/30
> 
> But resulting temperature is about 8 degrees too low.

Compared to what?

http://www.lm-sensors.org/wiki/FAQ/Chapter3#MyBIOSreportsamuchhigherCPUtemperaturethanyourmodules
(In particular the last paragraph.)

> Does anyone have a better temp2 computation for cusl2c?
> 
> I think it's not wise just to add 33 instead of 25 in computation. ;)

-- 
Jean Delvare

[lm-sensors] better temp2 computation for asus CUSL2C

[gimeshell at web.de]

On Sun, 18 Jun 2006 11:31:17 +0200
Jean Delvare <khali at linux-fr.org> wrote:

> > But resulting temperature is about 8 degrees too low.
> 
> Compared to what?
> 
> http://www.lm-sensors.org/wiki/FAQ/Chapter3#MyBIOSreportsamuchhigherCPUtemperaturethanyourmodules
> (In particular the last paragraph.)
> 

Yes that's it!
Now what would be better way to correct this? Is there any difference
between both ways?

a) multiply @ with bigger factor eg 31 instead of 30
compute temp2 (@*31/43)+25, ...

b) adding bigger factor, eg 31 instead of 25 to @.
compute temp2 (@*30/43)+31, ...

Why isn't compute line so easy to only add ONE value to @. No
multiplikation anymore. Only addition. Is this a practical thought?

regards,
gimeshell

[lm-sensors] better temp2 computation for asus CUSL2C

[Jean Delvare]

Hi (no name),

> > http://www.lm-sensors.org/wiki/FAQ/Chapter3#MyBIOSreportsamuchhigherCPUtemperaturethanyourmodules
> > (In particular the last paragraph.)
> 
> Yes that's it!
> Now what would be better way to correct this?

Actually my point was that there is probably nothing to correct.

>                                                Is there any difference
> between both ways?
> 
> a) multiply @ with bigger factor eg 31 instead of 30
> compute temp2 (@*31/43)+25, ...
> 
> b) adding bigger factor, eg 31 instead of 25 to @.
> compute temp2 (@*30/43)+31, ...

How are we supposed to guess? We would need many data points to decide
which is best, and you did provide only one (which may not even be
correct.)

> Why isn't compute line so easy to only add ONE value to @. No
> multiplikation anymore. Only addition. Is this a practical thought?

We use the formula which gives the best result for the applicable range
of temperatures. There's little point in having a slighly more simple
equation if it gives less accurate results.

-- 
Jean Delvare

[lm-sensors] better temp2 computation for asus CUSL2C

[gimeshell at web.de]

On Sun, 18 Jun 2006 19:02:30 +0200
Jean Delvare <khali at linux-fr.org> wrote:

Hi,

> Actually my point was that there is probably nothing to correct.

my BIOS does have such an idle loop, too.
Temperature is increasing from about 35 degrees immidiately after
entering BIOS up to 46 degrees met after a few minutes.
Looking at sensors output while linux is idle, i get shown about 30
degrees with sensors default computation line for asus CUSL.

> 
> We use the formula which gives the best result for the applicable
> range of temperatures. There's little point in having a slighly more
> simple equation if it gives less accurate results.
> 

But what's the mathematic principle behind this method (first
to multiply with value, then to add another value)?
Isn't the bottom line outcome the same if only adding fixed value
instead of first to multiply and after that to add another value?

regards,
gimshell

[lm-sensors] better temp2 computation for asus CUSL2C

[Jean Delvare]

> > Actually my point was that there is probably nothing to correct.
> 
> my BIOS does have such an idle loop, too.
> Temperature is increasing from about 35 degrees immidiately after
> entering BIOS up to 46 degrees met after a few minutes.
> Looking at sensors output while linux is idle, i get shown about 30
> degrees with sensors default computation line for asus CUSL.

So I'd say the compute line is correct and you don't need to change it.

> > We use the formula which gives the best result for the applicable
> > range of temperatures. There's little point in having a slighly more
> > simple equation if it gives less accurate results.
> 
> But what's the mathematic principle behind this method (first
> to multiply with value, then to add another value)?

We simply try to reflect the physical reality. Given that this physical
reality is complex we limit ourselves to a linear interpolation in the
temperature range we are interested in. Usually this is good enough.

> Isn't the bottom line outcome the same if only adding fixed value
> instead of first to multiply and after that to add another value?

Sorry, there's not much I can do for you at this point. You seem to
lack the basics of linear algebra.

-- 
Jean Delvare

[lm-sensors] better temp2 computation for asus CUSL2C

[gimeshell at web.de]

On Sun, 18 Jun 2006 22:17:15 +0200
Jean Delvare <khali at linux-fr.org> wrote:

> > my BIOS does have such an idle loop, too.
> > Temperature is increasing from about 35 degrees immidiately after
> > entering BIOS up to 46 degrees met after a few minutes.
> > Looking at sensors output while linux is idle, i get shown about 30
> > degrees with sensors default computation line for asus CUSL.
> 
> So I'd say the compute line is correct and you don't need to change
> it.
> 

Hi,

i did make some tests (sorry for late reply):

CPU busy
====
Linux: (mprime cpu stressing): lm-sensors: 44? C
BIOS: (instantly reset switch): 43? C

CPU idle
====
Linux: lm-sensors: 38? C
BIOS: (instantly reset switch): 36? C

BIOS
==
(probably cpu load, loop) 45? C

sensors.conf compute line:
compute temp2 (@*30/43)+34, (@-25)*43/30 

instead of default:
compute temp2 (@*30/43)+25, (@-25)*43/30 

Motherboard: Asus CUSL2C

Or did i overlook anything?

regards,
gimeshell