LM_Sensors Finally Gets Better Intel CPU Support

[phoronix]

Phoronix: LM_Sensors Finally Gets Better Intel CPU Support

While the Linux hardware compatibility of modern processors and motherboards have normally been spot-on, as talked about in our many reviews, one of the areas that still causes annoyances with modern hardware can be the hardware sensors support. For motherboards this commonly means being able to monitor sensors for the temperatures, fan speeds, voltages. For the CPUs, their integrated temperature sensor(s) also aren't commonly supported on just-released Intel and AMD products...

http://www.phoronix.com/vr.php?view=ODI1Nw

[Cyborg16]

And i3/i5/i7 CPUs? You say recent, but I don't think those are included in the code names mentioned. Well, that's getting somewhere at least..

[xeros]

And i3/i5/i7 CPUs? You say recent, but I don't think those are included in the code names mentioned. Well, that's getting somewhere at least..

Yeah, when I've read the article I've searched for the date... ;-)

[Wingfeather]

Cyborg16,

The newest Intel Core 2 Duo and Core 2 Quad CPUs along with those from the Nehalem and Westmere families have been tested with this code.

Doesn't that have you covered?

[Cyborg16]

Sorry, obviously didn't know the code names as well as I thought!

[Fixxer_Linux]

I was surprised by my relatively up values of temp on my Core2Quad (50°C when it's suppposed to run @ 40°C or even below). I understand now that this was just some outdated code that could explain that.

This is a good new.

[NeoBrain]

I was surprised by my relatively up values of temp on my Core2Quad (50°C when it's suppposed to run @ 40°C or even below). I understand now that this was just some outdated code that could explain that.

This is a good new.

IIRC, some Intel CPUs had a faulty temperature detection, they always report about 10% too much. Affects my Core2Duo, not sure whether it's the case for Core2Quad as well though.

EDIT: i.e., it's not LM_Sensors fault but rather the hardware's one

[curaga]

My C2D reported temps went up by 10 C after a kernel update ~2 years ago, I think they changed the scale or something.

[mugginz]

Instead the driver now checks a CPUID instruction (CPUID.06H.EAX[0]) to determine if thermal sensors are available.

Well I'm not so sure it's wise to use this method unless there's no analogue to the way AMD CPU's expose this info over the PCI bus.

Unless there's some other mitigating code, if you're running a multi socket board then this instruction will interrogate the CPU currently running the driver code and not necessarily the CPU you want the info from.

[IsawSparks]

Intel owners : The scale wasn't changed. Tjunction was. In early Core2's it's 90degrees Celsius, whereas in E6xxx series and later it's 100degrees Celsius. I've never heard of faulty temp sensors en masse, but it may be true.