Switching between thermal sensors in Brisbane core

[VVJ]

Hi, there!

As we can see, a topic about the broken CPU temperature sensor in Brisbane core is extremely popular today. And I decided to pour some oil on the flames too.

So, in short. Everybody knows that the AMD64 processor has an internal thermal sensor to measure a CPU die temperature. But according to the latest BKDG document the dual-core NPT processor family has the second thermal sensor in the CPU core. This means the dual-core AMD64 processor revision F and later has 4 thermal sensors - each core has a pair of thermal sensors. Unfortunately, we can’t find any software that could provide a selection for the thermal sensor. I created the simplest app that enables you to switch between two sensors of the CPU core. I want to believe that one from the sensors is really working. So, here is the link to CoreSensor.

Good luck and thanks!

[papatsonis]

just tested it , and reports only 2 sensors , sensor #0 , sensor#1 , same results with coretemp

[VVJ]

papatsonis, Please, read carefully! Each core of the dual-core CPU has two sensors. I mean Windsor and Brisbane.

[AgentGOD]

That's pretty cool. Now we can see exactly how calibrated the sensors are.

[papatsonis]

i messed around with the two sensors/core and only thing i observed was the temperatures from 0C/10-11C were reported reversely , 11C/0C , may try again later (not this pc..)

[PetNorth]

it doesnt work with vista x64, message "Could not install system driver"

Can you solve it?

[uwackme]

Fired it up on a new Windsor 5600+. Sensor 0 is fine on either core, sensor one comes up with a (-) number. The sensor 0 result matches coretemp.

[Lightman]

For my single core Lima (rev. G) it's working fine.
Sensor '0' reports temps around 6.5-7.5C and sensor 1 reports temps around 7-8.5C. Looks like both sensors are broken.

PS. Thanks for sharing this nice util VVJ!!

[The Coolest]

it doesnt work with vista x64, message "Could not install system driver"

Can you solve it?

It's not as easy as you may think, Vista x64 requires the author to digitally sign their driver, which is rather expensive.
There is a workaroung. Your can read about it here:
http://www.alcpu.com/forums/viewtopic.php?t=211

[GD5015]

Interesting development. When I switch between core 1 sensors, it switches the reading in speedfan, also. At idle shows 8 c or 12 c. Core 0 sensor -5 c or 6 c.

[PetNorth]

It's not as easy as you may think, Vista x64 requires the author to digitally sign their driver, which is rather expensive.
There is a workaroung. Your can read about it here:
http://www.alcpu.com/forums/viewtopic.php?t=211

This is not the problem. I can install permanently unsigned drivers in my Vista x64, with this trick (it must be executed in command line with admin rights):

C:\windows\system32\bcdedit -set loadoptions DDISABLE_INTEGRITY_CHECKS

With this, you can ever install unsigned drivers in vista x64. It's not necessary the F8 each time you boot.

The problem with VVJ driver is different. It's not a signed or unsigned issue. It simply refuses to install.

[The Coolest]

I just ran the app on my X2 3600+ Brisbane and got same reading with both sensors
around 7-8C on core 1
around 16-17C on core 2

[AgentGOD]

I got identical readings with any sensor too.

[funnyperson1]

Pretty much identical readings on my side as well.

[VVJ]

Looks like both sensors are broken.

Howdy!
I’ve got the latest Revision Guide for AMD NPT Family 0Fh Processors from AMD, but I didn’t find any mention about broken thermal sensors in Revision DH-G1. I doubt that the sensors are “broken” as users suppose.

[Lightman]

Howdy!
I’ve got the latest Revision Guide for AMD NPT Family 0Fh Processors from AMD, but I didn’t find any mention about broken thermal sensors in Revision DH-G1. I doubt that the sensors are “broken” as users suppose.

Then we can say 'wrongly calibrated' because it is unusual to get around 7-9C inside CPU when ambient temperature is around 22C.
You might be right saying they are not broken, because temperature readings are increasing under load about right amount... (another 5-7C)

[The Coolest]

Then we can say 'wrongly calibrated' because it is unusual to get around 7-9C inside CPU when ambient temperature is around 22C.
You might be right saying they are not broken, because temperature readings are increasing under load about right amount... (another 5-7C)

I agree

[roadie]

Here is a quick comparison I made between CoreSensor and Speedfan (please see link). The programme seems to work as both it and Speedfan report the same kind of temperature rise from idle to load (about 20 degrees). However, the sensors within the chip seem to be broken as at idle CoreSensor gives a temperature in the mid to high teens with an ambient of around 21 degrees.

[Starscream]

got a x2 3600 wich is atm running at 2400mhz default voltage.

speedfan says: 20 degrees
AMD64CT says: 11.5 and 12 on sensor 0
and Sensor 1 says 7 and 2 degrees

i watercool my CPu but them 12 degrees are to low..

[Okda]

it is working on my Opty 165

temps are not the same like what coretemp reports

and it can detect 2 sensors for every core but the temps are the same

[Cooper]

Thermal sensors aren't broken. It's just software that cannot read the data correctly.
Now you can only monitor CPU temperature using existing apps like Sandra, Everest and SpeedFan. S&M scales very good but has wrong zero point.

[The Coolest]

Thermal sensors aren't broken. It's just software that cannot read the data correctly.
Now you can only monitor CPU temperature using existing apps like Sandra, Everest and SpeedFan. S&M scales very good but has wrong zero point.

Why can't the software read the sensors properly? Can you please elaborate? From what I saw programs like Sandra and Everest, etc. gave the same kind of readings as Core Temp and CoreSensor/NextSensor. Or are you reffering to the old fashioned way of reading CPU temperature using the motherboard diode?
I'm getting the information like explained in the K8 BIOS designer's guide. We all use the same kind of method to read CPU temps.

it is working on my Opty 165

temps are not the same like what coretemp reports

and it can detect 2 sensors for every core but the temps are the same

This thread is about the new Brisbane cores, they're the only CPUs which give weird readings. Other CPUs give correct readings in most of the cases.

[Okda]

ah sorry didn't notice that

[2good4you]

Very interesting, should be great to find a way to read coretemp. Both sensors can't be broken.

[DoubleZero]

The right temp there is 61º.