Hi
We have a motherboard that has Thermal Throttling in the BIOS (which
we cannot disable). This causes the CPU usage to go up and down when
the CPU temperature reaches (and stays around) the Throttling
temperature point.
What we would like to know is whether this will affect the timings in
drivers, eg the wireless drivers we are using. What can we check in
drivers' code that will tell us that its operations may be affected
the throttling?
In the past few days, we noticed that some of the linux units we
deployed freezes after deveral hours of operation, we are now trying
to reproduce the problem in our test environment. Some insight on the
affect of throttling will help us narrow down the search.
Regards
Keith
On Wed, 2006-10-04 at 03:29 +1300, Keith Chew wrote:
> Hi
>
> We have a motherboard that has Thermal Throttling in the BIOS (which
> we cannot disable). This causes the CPU usage to go up and down when
> the CPU temperature reaches (and stays around) the Throttling
> temperature point.
>
> What we would like to know is whether this will affect the timings in
> drivers, eg the wireless drivers we are using. What can we check in
> drivers' code that will tell us that its operations may be affected
> the throttling?
>
> In the past few days, we noticed that some of the linux units we
> deployed freezes after deveral hours of operation, we are now trying
> to reproduce the problem in our test environment. Some insight on the
> affect of throttling will help us narrow down the search.
Hi,
in general linux should be ok with this happening. However for specific
cases... you'll need to provide more information; you're not mentioning
which drivers you are using for example. Or even which versions of the
kernel etc etc....
(also: if you actually HIT throttling, there is something very very
wrong; you're not supposed to hit that unless the fan is defective, but
never in "normal" healthy operation. If you do hit it without hardware
defects then there is most likely a fundamental airflow problem you'll
want to fix urgently)
--
if you want to mail me at work (you don't), use arjan (at) linux.intel.com
Hi!
> > We have a motherboard that has Thermal Throttling in the BIOS (which
> > we cannot disable). This causes the CPU usage to go up and down when
> > the CPU temperature reaches (and stays around) the Throttling
> > temperature point.
> >
> > What we would like to know is whether this will affect the timings in
> > drivers, eg the wireless drivers we are using. What can we check in
> > drivers' code that will tell us that its operations may be affected
> > the throttling?
> >
> > In the past few days, we noticed that some of the linux units we
> > deployed freezes after deveral hours of operation, we are now trying
> > to reproduce the problem in our test environment. Some insight on the
> > affect of throttling will help us narrow down the search.
> in general linux should be ok with this happening. However for specific
> cases... you'll need to provide more information; you're not
> mentioning
Really? AFAICT P4 will happily slow down behind our backs, making at
least udelays() with interrupts disabled sleep for too long.
> (also: if you actually HIT throttling, there is something very very
> wrong; you're not supposed to hit that unless the fan is defective, but
> never in "normal" healthy operation. If you do hit it without hardware
> defects then there is most likely a fundamental airflow problem you'll
> want to fix urgently)
At least in toshiba notebook, I was hitting thermal throttling even
through both fans were okay. There are many misdesigned machines out
there, I fear.
Pavel
--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
On Sun, 2006-10-08 at 23:18 +0200, Pavel Machek wrote:
> Hi!
>
> > > We have a motherboard that has Thermal Throttling in the BIOS (which
> > > we cannot disable). This causes the CPU usage to go up and down when
> > > the CPU temperature reaches (and stays around) the Throttling
> > > temperature point.
> > >
> > > What we would like to know is whether this will affect the timings in
> > > drivers, eg the wireless drivers we are using. What can we check in
> > > drivers' code that will tell us that its operations may be affected
> > > the throttling?
> > >
> > > In the past few days, we noticed that some of the linux units we
> > > deployed freezes after deveral hours of operation, we are now trying
> > > to reproduce the problem in our test environment. Some insight on the
> > > affect of throttling will help us narrow down the search.
>
> > in general linux should be ok with this happening. However for specific
> > cases... you'll need to provide more information; you're not
> > mentioning
>
> Really? AFAICT P4 will happily slow down behind our backs, making at
> least udelays() with interrupts disabled sleep for too long.
it will during thermal throttle yes. but udelay() and co as API never
have been super accurate; they mostly promise at least this much delay
Hi
> > in general linux should be ok with this happening. However for specific
> > cases... you'll need to provide more information; you're not
> > mentioning
>
> Really? AFAICT P4 will happily slow down behind our backs, making at
> least udelays() with interrupts disabled sleep for too long.
>
I believe we are seeing similar behaviours with the drivers. As a
temporary workaround, we have asked the manufacturers to supply us
with a BIOS that disabled THRM-Throttling. So far, things are looking
good, but we are still doing more testing.
> > (also: if you actually HIT throttling, there is something very very
> > wrong; you're not supposed to hit that unless the fan is defective, but
> > never in "normal" healthy operation. If you do hit it without hardware
> > defects then there is most likely a fundamental airflow problem you'll
> > want to fix urgently)
>
> At least in toshiba notebook, I was hitting thermal throttling even
> through both fans were okay. There are many misdesigned machines out
> there, I fear.
In our case, we are running the unit in mobile environment, with not
much ventilation. In our test environment, the unit runs fine at
temperatures of 80-90 degrees. In the real environment, we only expect
temperatures of 50-70 degrees.
Regards
Keith