Running mprime295-linux64 test 2 = In-place large FFTs (maximum heat and power consumption, some RAM tested) with 4 threads. Core 0: +55.0?C (crit = +95.0?C) Core 1: +58.0?C (crit = +95.0?C) Core 2: +58.0?C (crit = +95.0?C) Core 3: +55.0?C (crit = +95.0?C) ?CPU Temp: +47.0?C (high = +85.0?C, hyst = +81.0?C) ?Sys Temp: +43.0?C (high = +85.0?C, hyst = +81.0?C) ???? Temp: +47.0?C (high = +70.0?C, hyst = +68.0?C) ACPI THRM temperature: 47 C [so this is probably just CPU Temp] So '?CPU Temp' is probably indeed CPU Temp, but the other two are still unclear (especially the third one, which does oscillate a little, seen 45-48). The Core Temps do seem to be consistently about ~10 C higher than CPU Temp, which suggest that TjMax is probably closer to 90 C [instead of the assumed 95 C]. Since usually core temps are about 5 C higher than cpu case temp, assuming that that is indeed what is being measured here. Which also is in compliance with the thermal spec of an atom 330 being 85.2C [ http://processorfinder.intel.com/details.aspx?sSpec=SLG9Y ] (+~5C case max to junction max gives a 90C TjMax). But it's really all just guess-work... Unfortunately my setup has rather good cooling (ran into issues with the 2 HDDs overheating and opened up the case and added an extra huge fan). And as I don't currently have physical access to it (it's on the other side of the pond) I'm not willing to dial it down. Anyway... to actually include some useful content in this email. How are we going to proceed with the atom coretemp patch? [forward ported from http://bugzilla.kernel.org/show_bug.cgi?id=11161 ] Cheers, Maciej 2009/6/15 Maciej ?enczykowski <zenczykowski at gmail.com>: > (In case that wasn't clear, the thing marked CPUFAN in the BIOS and in > the config is actually on the northbridge, and the exhaust from the > northbridge blows over the cpu) > > 2009/6/15 Maciej ?enczykowski <zenczykowski at gmail.com>: >> Fair enough, but the intel atom 330 chip is only 8W TDP (passive >> cooling, although using the exhaust heat from the northbridge for >> cooling) and the northbridge chipset is ~20W TDP (active cooling). ?As >> such it's not clear exactly what to expect to get hotter faster. >> They're marked the way I believe the BIOS marks them, but on an idle >> machine with plenty of cooling: >> >> Core 0: ? ? ?+40.0?C ?(crit = +95.0?C) >> Core 1: ? ? ?+41.0?C ?(crit = +95.0?C) >> Core 2: ? ? ?+42.0?C ?(crit = +95.0?C) >> Core 3: ? ? ?+41.0?C ?(crit = +95.0?C) >> >> CPUFAN: ? ? 1920 RPM >> SYSFAN1: ? ?3667 RPM >> SYSFAN2: ? ?2008 RPM >> ?CPU Temp: ? +31.0?C ?(high = +85.0?C, hyst = +81.0?C) >> ?Sys Temp: ? +41.0?C ?(high = +85.0?C, hyst = +81.0?C) >> ???? Temp: ? +47.0?C ?(high = +70.0?C, hyst = +68.0?C) >> >> # cat /proc/acpi/thermal_zone/THRM/temperature >> temperature: ? ? ? ? ? ? 31 C >> >> Unfortunately the core temps may or may not be correct (they may need >> an offset of +-X degrees, since it's not clear what TjMax actually is >> for an atom 330, and I'm 'kind of' guessing it to be 95), and the cpu >> temp being lower than the sys temp actually makes some sense... >> >> The last temperature isn't even shown by the BIOS, but does seem to >> show a kind of reasonable value (maybe it's the NB?)... >> >> I'm guessing that the CPU does actually run cooler than the >> northbridge - it's just a tad counter-intuitive. >> >> 2009/6/15 Jean Delvare <khali at linux-fr.org>: >>> On Fri, 5 Jun 2009 13:29:38 +0200, Maciej ?enczykowski wrote: >>>> submitting a sensors.conf file for a Jetway NC92-330-LF motherboard >>>> with f71882fg monitoring chip. >>>> >>>> # uname -a >>>> Linux gaia 2.6.29.4-168.mz1.fc11.x86_64 #1 SMP Fri Jun 5 11:10:05 CEST >>>> 2009 x86_64 x86_64 x86_64 GNU/Linux >>>> >>>> # lsmod | head -n 5 >>>> Module ? ? ? ? ? ? ? ? ?Size ?Used by >>>> f71882fg ? ? ? ? ? ? ? 27120 ?0 >>>> coretemp ? ? ? ? ? ? ? ?6768 ?0 >>>> >>>> Also including a slight patch to coretemp.c for atom support. >>>> >>>> Hope you include the conf in the wiki. >>> >>> Done, thanks for your contribution: >>> http://www.lm-sensors.org/wiki/Configurations/Jetway/NC92-330-LF >>> >>> Note: you should really be able to tell which is the CPU temperature >>> sensor by putting significant load on the system and watching which >>> temperature value raises highest and fastest. >>> >>> -- >>> Jean Delvare >>> >> >
