From mboxrd@z Thu Jan 1 00:00:00 1970 From: Guenter Roeck Date: Tue, 27 Mar 2012 23:28:53 +0000 Subject: Re: [lm-sensors] [PATCH] hwmon: (it87) Add support for IT8781F, IT8782F, IT8783E/F Message-Id: <20120327232853.GA21100@ericsson.com> List-Id: References: <6.2.5.6.2.20120324091434.02773d68@roeck-us.net> In-Reply-To: <6.2.5.6.2.20120324091434.02773d68@roeck-us.net> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable To: lm-sensors@vger.kernel.org On Tue, Mar 27, 2012 at 03:21:35PM -0400, Bj=F6rn Gerhart wrote: > Hi Jean, >=20 > Am 26.03.2012 um 19:01 schrieb Jean Delvare: > > [..] > >> If actually it is the job of the module to do such calculations, I wou= ld offer to create an it87 patch for IT8783 fixing this - analog to an exis= ting module with a similar behavior (if you know one). > >=20 > > You assume that the PECI value always refers to the CPU. I do not think > > this is true, other parts (e.g. north bridge) could report their > > temperature through PECI as well. > >=20 > > And if it is the CPU temperature, we already have the coretemp driver > > providing the same information, so there's little point in implementing > > the same in it87. The only benefit of handling PECI sensors in it87 and > > the like would be so that they can be used as sources for automatic fan > > speed control, but if the chip itself has no idea about the absolute > > value of the temperature then I doubt it can use it for automatic fan > > speed control. > >=20 > Yes, with the coretemp driver even both core temperatures are displayed c= orrectly. So I agree, it's nonsense to implement our special design into th= e generic driver ;-) Depends. As Jean mentioned, it would make sense if the temperature can be u= sed for automatic fan control. The PECI specification says (I qoute Wikipedia) "PECI reports a negative va= lue expressing the difference between the current temperature and the thermal throttle poi= nt (at which the CPU reduces speed or shuts down to prevent damage due to overheating) i= nstead of the absolute temperature". So the negative value and slope is to be expected. Question is if there is a way to tell the chip the how to calculate the rea= l temperature from the raw value reported by PECI. The Nuvoton chips have a set of regist= ers for that purpose. Since the raw PECI value is not very useful, it might well be that it is po= ssible to program the ITE chip to adust its readings. No idea how to do that, thou= gh. If your HW folks have a contact at ITE, maybe you could ask them to try to = find out. Or maybe there is some BIOS code available from ITE and/or from BIOS vendor= s. Alternatively, it might be possible to program fan control to take the nega= tive slope into account; at least Wikipedia suggests that as possibility (to have the = fan run faster as the reported temperature approaches zero from below). No idea how to do = that either, though. There is also register 0x0a, bits 4-6. Is that set to PECI on your board ? Either case, I think it would make sense for us to report the sensor type a= s PECI if so configured. Yet another patch ... Thanks, Guenter _______________________________________________ lm-sensors mailing list lm-sensors@lm-sensors.org http://lists.lm-sensors.org/mailman/listinfo/lm-sensors