Thermal Monitoring (About)

[Reloadron]

Thermal Monitoring

Recent years within the computer enthusiast community have developed a new generation of enthusiast with a desire to closely monitor the parameters of Voltages, Fan Speeds and Temperatures. To meet these desires a variety of software has been written and made available to the dedicated enthusiast. Maybe we should take a look at some of these power toys and how they work so we can better understand the results they provide us with.

Long before the computer enthusiast had a need to measure critical parameters, the system itself had the need. Systems needed to be able to measure their own temperatures, voltages and fan speeds to ensure normal operation. To enable this process Intel Corporation developed what is known as the SMBus (System management Bus) in 1995. The SMBus is a simple two wire communications bus allowing communication between various devices on a motherboard, with a focus on power related devices (chips). Newer operating systems support SMBus such as Free BSD, Linux, Windows 2000, Windows XP and Windows Vista, older operating systems such as Windows ’98 and earlier do not support the SMBus. Remember at this point the development and early use of the SMBus was strictly for the benefit of the system and allowed for no human interface. The SMBus was a tool for the system to use and not the user.

One of the first software programs I recall that actually allowed a user interface with the SMBus was likely Speed Fan. Upon its initial release Speed Fan was about as cool of a power toy any enthusiast could want. Through the SMBus Speed Fan allowed a user to actually view information (Data) from various hardware monitoring chips located on the motherboard. Many of these chips have evolved into sort of processors all their own. Early on there were but a small handful of such chips being used by computer motherboard manufacturers, however as the years passed these chips grew in mass numbers to where they are today. The early chips could each measure a few basic parameters such as a few fan speeds and temperatures as well as a few voltages. Today chips like the Winbond 83791D and 83791G are capable of measuring:

• Ten voltage monitoring inputs.
• Five fan speed monitoring inputs.
• Three temperature monitoring inputs from remote sensors.
• Case open detection monitoring input.
• WATCGDOG comparisons of all monitored values.
• Programmable Hysteresis and setting points (alarm thresholds) for all monitored values.

Hardware monitoring has come a long way since those early hardware monitoring chips. The mentioned chip has a voltage monitoring accuracy of +/- 1% and a temperature monitoring accuracy of +/- 3 Degrees C. Really not bad for an 8 bit ADC (Analog to Digital Conversion) device.

So why is it so difficult to obtain reasonably good, accurate and reliable temperature readings? The software needs to first query the SMBus to see what hardware monitoring chips are out there. Remember motherboard manufacturers use a variety of chips similar to the chip I mentioned made by a variety of manufacturers. These chips will identify themselves across the bus by part number. However, once the software has correctly identified the chip the software also needs the necessary code written within it to read from the chip and convert the data to real world numbers we the users can understand. The software needs to be able to correctly interrogate and communicate with the chip and then convert the data. This includes voltage analog to digital conversion for voltages as well as temperatures. The software actually needs a mini library of sorts defining the names of dozens of hardware monitoring chips and what functions to employ to address each chip. If at any point the software fails to correctly identify the chip being used or the data it provides the data and results of the conversions will be erroneous.

A typical example for the chip I mentioned is the temperature range it measures. The chip measures between -40 Deg. C. and 120 Deg. C. It does this on 3 channels dedicated to measuring temperature from a thermal sensor. Should the software fail to get things right it would not be unusual to see a displayed temperature of -40 or 120 Degrees C. displayed. Additionally the allowable error of +/- 3 Degrees C. is for the chip assuming a good and accurate input from the sensor. The employed sensor can further contribute to errors. Next, not all motherboard manufacturers use all channels. When a channel does not see any input device the output will frequently default to a High or Low on all bits. Thus even if the software is correctly reading the chip, it is quite possible to still see erroneous readings based on the motherboard design. There are no shortages of places for things to run amuck and the user to see erroneous data.

We need to keep in mind that the “Power Toys” developed to read the SMBus will not always provide optimum results and the SMBus was not developed originally as a user tool but a system tool to be read and used by the system.

There is another new tool on the block the computer enthusiast is using to measure specifically CPU temperatures and to take it further the actual temperatures in the actual core of the CPU. A program called CoreTemp takes full advantage of a new technology employed in newer CPUs manufactured by both Intel and AMD. CoreTemp takes advantage of DTS (Digital Thermal Sensors) placed (embedded) directly in the CPU at manufacture. This method eliminates the use of the SMBus and reads directly from the CPU. The following is taken directly from the CoreTemp website:

http://www.alcpu.com/CoreTemp/howitworks.html

How It Works

Core Temp lets you monitor Intel "Core Duo", "Core Solo" (Yonah), "Core 2 Duo", "Core 2 Extreme", "Core 2 Quad", " Pentium E2000" series, "Celeron 400500" series (Allendale, Conroe, Merom, Kentsfield, Conroe-L respectively), "Xeon 3000/3200/5100/5300" series (Woodcrest, Clovertown respectively) and all AMD K8 (AMD64) and K10 (Phenom, Opteron) series die temperature.
The temperature readings are very accurate as the data is collected from a Digital Thermal Sensor (or DTS) which is located in each individual processing core, near the hottest part. This sensor is digital, which means it doesn't rely on an external circuit located on the motherboard to report temperature, its value is stored in a special register in the processor so any software can access and read it. This eliminates any inaccuracy that can be caused by external motherboard circuits and sensors and then different types of programs trying to read those sensors.
This is how the program works:
Intel defines a certain Tjunction temperature for the processor. In the case of Yonah it is 85C° or 100C°. First of all the program reads from a Model Specific Register (or MSR), and detects the Tjunction temperature. A different MSR contains the temperature data, this data is represented as Delta in C° between current temperature and Tjunction.
So the actual temperature is calculated like this 'Core Temp = Tjunction - Delta'
The size of the data field is 7 bits. This means a Delta of 0 - 127C° can be reported in theory. But from preliminary tests, the reported temperature doesn't go below 0C°, no matter what kind of cooling was used.
AMD chips report the temperature by a special register in the CPU's NB. Core Temp reads that register and uses a formula provided by AMD to calculate the current temperature.
The formula for the K8 is: 'Core Temp = Value - 49'.
The formula for the K10* is: 'CPU Temp** = Value / 8'.
The sensor in AMD CPUs can report temperatures between -49C and 206C.
*K10 = Phenom (Agena), Opteron (Barcelona). The K10 reports a temperature value that is relative to a certain predefined value, it doesn't report the actual processor temperature! So take that into consideration.
**CPU Temp is because the PhenomOpteron (K10) have only one sensor per package, meaning there is only one reading per processor.

Overall

We need to constantly remember that all programs have room for some degree of error. We need to as enthusiast take the readings and data afforded us with a grain of salt. These programs provide us with basic information and their data should not be taken as exact. They are overall pretty accurate when they correctly read the hardware monitoring chips or directly read from the CPU.

Ron

[halutzparilla]

nice little article too many word though i'm a lazy reader i only read the first two paragraph

but one question which i ask already noone answered me here it is:

Whats the difference of RTemp from CTemp again??? which is better? i kinda like Core Temp...

[polobunny]

Originally Posted by halutzparilla

nice little article too many word though i'm a lazy reader i only read the first two paragraph

but one question which i ask already noone answered me here it is:

Whats the difference of RTemp from CTemp again??? which is better? i kinda like Core Temp...

CoreTemp has been found to be based on many "assumed" values, but the biggest problem is that it takes DTS as a linear value.
The discovery that DTS was not likely a linear value was when many users found out their processor cores to be under the ambient temps, which is obviously not possible unless you are running some form of extreme cooling (TEC, Ln2, phase change etc)
Additionally, you might end up with too high readings because of this too.

Edit: RealTemp is a program looking to palliate this "error", and I frankly believe more RealTemp than CoreTemp for my current system. However, as Ron said in it's overall paragraph, take everything with a grain of salt.

[Reloadron]

Originally Posted by halutzparilla

nice little article too many word though i'm a lazy reader i only read the first two paragraph

but one question which i ask already noone answered me here it is:

Whats the difference of RTemp from CTemp again??? which is better? i kinda like Core Temp...

First I agree it is boring as hell. I had to drink Coke to write it as with beer I would have fallen asleep. This is not something I figured I could shove pictures into. Unless it was off topic pics to help keep it interesting.

Not sure what you mean by Rtemp and Ctemp? I can see Ctemp being maybe the core temperature of a CPU core taken at the hottest place within the core.

<EDIT> Thank You Polo! I wasn't sure what the refrence of Ctemp and Rtemp was. </EDIT>

Ron

[halutzparilla]

Ron = talking about Real Temp vs. Core Temp ....

[Reloadron]

Originally Posted by halutzparilla

Ron = talking about Real Temp vs. Core Temp ....

Me thinks Polo nailed it perfectly.

Ron