From: Arthur Machlas on
On Fri, Jul 30, 2010 at 12:52 AM, Stan Hoeppner <stan(a)hardwarefreak.com> wrote:
> Arthur Machlas put forth on 7/29/2010 12:01 PM:
>
>> Things are running nicely, but the problem I hoped
>> to resolve hasn't been. Namely, the lowest frequency my cpu can reach
>> is 1Ghz... instead of the 800Mhz that it reaches on windows and in the
>> spec sheets.
>>
>> Advice on how to proceed from here is greatly appreciated.
>
> I led you to the well but it's up to you to drink Arthur.  You didn't read the
> help screens.  All the frequency info you need is there.  It is the key to you
> succeeding at this.  You may have to experiment some, but that's a requirement
> when rolling one's own kernels.  Welcome to the club.  It's rarely "easy". ;)
>
Asking on here isn't my first attempt at figuring things out. My
kernel config work is based on Greg KM's book "The Linux Kernel In A
Nutshell". So, not only have I read the help screens, but I've also
read a pretty decent book by one of the kernel's prime maintainers.

I should have been more clear about the advice I was hoping for, since
there is nothing wrong with my kernel config, I was looking for advice
about how to go about further debugging this issue of not having full
frequency range. I wanted to give fixing it a solid try before filing
a bug against the kernel itself.

> BTW, I'm curious as to your motivations for this.  Is this basically a
> "Windows can do 800MHz, so $deity dammit, Linux should be able to do it as
> well!" thing?

Not as such. More like a my processor is supposed to scale from 800Mhz
to 1.6Ghz, and its strange that it doesn't. I wonder why.

> In practical terms Arthur, you will not notice a meaningful
> difference in thermal output or current draw (battery consumption) between
> 800MHz and 1GHz with the n450.  The n450 has a TDP of 5.5 watts at 1..66GHz.
> Thus you won't even save 1 watt going from 1GHz to 800MHz in power save mode.
>  It'll be something like 300 milliwatts or less.  This exercise of yours is
> futile if your goal is a _practical_ difference in system operation.

I suppose that will have to count as advice on how to proceed, but I
hope you'll forgive me if I continue to search for an answer. Your
comments are reassuring to me though, that it isn't a serious problem,
and I do thank you for that.

At this point, I'm thinking it's a problem with the kernel or a
problem with my bios, and I think there are some kernel command line
parameters I can use to test the latter in Greg's book.

Best wishes,
AM


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org
Archive: http://lists.debian.org/AANLkTik4hXnhDfFabuJJjnPiaksp-b0emRvL3P157Czw(a)mail.gmail.com
From: Arthur Machlas on
On Thu, Jul 29, 2010 at 8:20 PM, Christian Jaeger <chrjae(a)gmail.com> wrote:
> How do you read the possible cpu frequencies?
>
> Your kernel needs cpufreq support and ondemand, powersave, etc.
> governors; check with
> cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
> cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
> cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
>
> Although with some hardware AFAIK other drivers than cpufreq are used,
> I don't know for Atom.
>

Hi Chris,

Available frequencies are taken from
/sys/devices/system/cpu/cpu0/freq/available_scaling_frequencies, in
addition powertop shows the same information.

Governors are all compiled into the kernel, and switching between them
by echoing "conservative" or "ondemand" to
/sys/devices/system/cpu/cpu0/freq/current_scaling_governor works. I
haven't tried user space yet, because I don't have any userspace tools
installed right now to change cpufreq. I'll remedy that this weekend
and see if it makes any difference.

Currently I'm using the acpi_cpufreq built into the kernel, however a
few other intel related ones like p4 and even those marked as
deprecated are built as modules.

Appreciate the input,
AM


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org
Archive: http://lists.debian.org/AANLkTinwXFWGRmvP+BghixQF8oWQGpRHrs3d3xt=bF0H(a)mail.gmail.com
From: Stan Hoeppner on
Arthur Machlas put forth on 7/30/2010 9:04 AM:

>> BTW, I'm curious as to your motivations for this. Is this basically a
>> "Windows can do 800MHz, so $deity dammit, Linux should be able to do it as
>> well!" thing?
>
> Not as such. More like a my processor is supposed to scale from 800Mhz
> to 1.6Ghz, and its strange that it doesn't. I wonder why.

That's what I figured. I had a similar mindset when I was a younger man. Let
me share a relevant story:

Back in 1997 I bought a beautiful used Midnight Blue 1993 Ford Probe GT with
the hopped up 24 valve quad cam V6 and a 5 speed manual transmission. It was
Motor Trend Magazine's 1993 Car Of The Year. The tested top speed was 137
miles per hour. I was dead set on achieving that published top speed. I
didn't have a racetrack available for my testing. Over the course of about 3
weeks I would find open stretches of Interstate 370 with little to no traffic.
On 5 occasions I was able to hit 135 mph, but no more. I was determined to
hit the magical 137 that Motor Trend said it would do. On the six occasion on
a cool afternoon I had just hit 136 mph when I rocketed past a hidden Missouri
Highway Patrolman manning a radar gun. As soon as I saw the cherries in my
rear view mirror I slowed down and pulled over. Evading wasn't an option even
with that kind of speed and lead: you can't outrun a radio or a helicopter. ;)

Due to traveling at almost double the posted speed limit I spent a mandatory
night in jail. The attorney cost me $450 and the ticket for doing 76 mph
_over_ the posted speed limit was $548, for a grand total of $998. The
attorney got the reckless driving charge dropped and massaged the speeding
charge so I only took a 2 point hit on my license instead of 5. A reckless
driving conviction along with the additional points would have put me into
what the insurance industry calls the "high risk" category, which would have
quadrupled my insurance premiums. That would have driven my monthly insurance
payment to just under $500/month, which was more than my split of rent at the
time.

The moral of the story: Even if "they" say it will do it, and others have
achieved it, it's not always wise, nor cost effective, to maniacally chase the
numbers. :)

> I suppose that will have to count as advice on how to proceed, but I
> hope you'll forgive me if I continue to search for an answer. Your
> comments are reassuring to me though, that it isn't a serious problem,
> and I do thank you for that.

There's nothing to forgive. It's your time and effort going into this. I'm
simply trying to save you large amounts of both as I've "been there and done
that".

> At this point, I'm thinking it's a problem with the kernel or a
> problem with my bios, and I think there are some kernel command line
> parameters I can use to test the latter in Greg's book.

And before filing a kernel bug, it would be very wise to ask the questions you
have here on lkml before you submit a bug. Debian-users isn't really the
proper venue for discussing your problem. This is a user list, not a
developer list. There are very few, if any, kernel hackers on this list with
the knowledge/answers you are seeking. And no I'm not one of them. Having
more than cursory experience rolling my own kernels is a far cry from being a
kernel hacker. I'm not a C programmer.

My advice: expend your energy on aspects of Debian/Linux that will yield
practical, tangible benefits. This isn't one of them.

--
Stan


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org
Archive: http://lists.debian.org/4C52FD83.3080901(a)hardwarefreak.com