Re: Off-Topic: G5 Open Firmware instability
On Fri, May 13, 2022 at 08:56:00AM -0700, tonyj@suse.de wrote:
> On Fri, May 13, 2022 at 03:58:08PM +0200, Frank Scheiner wrote:
> > Hi Tony,
>
> Hi.
>
> > Have you already looked below the cover of the cooling system? Is
> > there any visible leakage (remove power before looking into that)?
>
> Yes and no.
>
> > not, it could be that the water cooler is missing some liquid after that
> > much time which could limit its performance.
>
> /sys/devices/platform/windfarm.0 reports
>
> backside-fan: 1100
> backside-temp: 57.000
> cpu-current-0: 7.566
> cpu-current-1: 8.054
> cpu-current-2: 9.277
> cpu-current-3: 9.886
> cpufreq-clamp: 0
> cpu-front-fan-0: 3104
> cpu-front-fan-1: 3104
> cpu-power-0: 9.291
> cpu-power-1: 9.891
> cpu-power-2: 12.706
> cpu-power-3: 13.179
> cpu-pump-0: 3584
> cpu-pump-1: 3584
> cpu-rear-fan-0: 3200
> cpu-rear-fan-1: 3200
> cpu-temp-0: 39.265
> cpu-temp-1: 39.625
> cpu-temp-2: 66.046
> cpu-temp-3: 70.515
> cpu-voltage-0: 1.228
> cpu-voltage-1: 1.228
> cpu-voltage-2: 1.369
> cpu-voltage-3: 1.333
>
>
> Clearly cpus 2/3 are running hotter. I need to try disabling
> them (kernel cmd line or at runtime) and see if cpus 0/1 start
> running as hot.
>
> > * Have you checked that the radiator of the cooling system is not
> > cluttered with dust?
>
> Yes.
>
> > Say do you hear the fans speeding up over time?
>
> No. The onty time I hear the fan speed up is if the machine hangs,
> then about 45-60 seconds later it will go full.
>
> > You can "force" the fans to run on full speed by removing the
> > transparent cover.
>
> I have tried this. No change.
>
> > or longer that way, I'd say there's definitely something wrong with
> > your cooling system.
>
> Clearly there is. My concern however is that I have two issues and
> the OF one is distinct from the cooling one. The machine can boot into
> the existing Linux (from SATA) and stay up indefinitely (if load av <
> 1.5) but from a cold boot, will hang in OF in under 1 minute. Maybe
> Forth is a real resource hog :-)
It is certainly not unusual for firmware to run on a single core at 100%
(implementing idle handling and interrupts and such is just considered
too much hassle for something that isn't normally going to run).
So yeah it is actually not unlikely that openfirmware would cause the
CPU to run hot compared to linux at low load.
--
Len Sorensen
Reply to: