Tired of reading one conflicting forum post about temperatures after another? This guide will help you understand some of the details behind temperature readings for current Intel processors. It also has a section for FAQs about temperatures. Some parts are a little technical but I have tried to keep it as understandable.
This guide only applies to members of the Core, Core 2 and Core i7 families of processors (including Pentium Dual Core and Core 2-based Celerons), not the older Pentium 4′s or Pentium D’s. This is because the older processors do not have Digital Thermal Sensors on the cores.
Table of contents:
1. Definitions for commonly misused and misunderstood terms.
2. More detail and answers to common questions
3. Recommendations and additional reading.
1. Definitions for commonly misused and misunderstood terms.
Digital Thermal Sensor (DTS): Each core has its own DTS and this is where the Tjunction or core temperature reading is taken from. The DTS reports a relative distance to a fixed temperature (Tjunction Max) which must then be translated to an absolute temperature by software using the formula shown later in this post. This is usually called Distance to Tjunction Max or Delta to Tjunction Max. The primary function of the DTS is to regulate thermal protection mechanisms, not retrieve core temperatures.
The DTS is more accurate at temperatures closer to Tjunction Max and decreases in accuracy as temperature decreases.
Tcase: Technically this refers to the temperature that you can measure using a thermocouple embedded in the centre of the heat spreader, but this is only done in the factory by Intel or by users willing to drill their heat spreaders open to insert a thermocouple. Therefore Intel provided a diode between and below the cores with a reading calibrated by the BIOS which can be used instead. This reading can vary greatly when the BIOS version is changed, but will not necessarily change if the BIOS calibrations were not altered between versions.
Software supplied by motherboard manufacturers typically reads only this sensor, not the DTS. It is often labelled “CPU.” SpeedFan usually labels this as “CPU” as well.
Tcase Max and Thermal Specification: Can be found on the Intel Processor Spec Finder. This is often confused by users as the maximum temperature for the cores, but in actual fact it refers to the maximum temperature that should be reached by the Tcase sensor. This specification is rounded to a relatively precise 0.1 C.
The only exceptions are mobile CPUs which do not have a heat spreader, where Thermal Specification refers to Tjunction Max instead and is rounded to the nearest 5 degrees.
Tjunction: Synonymous with core temperatures and calculated based on the output from the DTS using the formula shown further down in this post.
Tjunction Max: This is the maximum temperature the cores can reach before the PROCHOT# signal is sent and thermal throttling is activated, although throttling typically occurs a degree or two below this. In this post Tjunction Max refers to either the actual throttling point for a core/CPU or what is used by a temperature monitoring program.
As you will read in the definition below, Intel released a list of values for use by temperature monitoring software which are rounded to the nearest 5 degrees. Remember that the DTS is only meant for controlling thermal protection, so precise rounding is not required.
Tjunction Target: According to Intel Tjunction Target is synonymous with Tjunction Max and this is their official term for it. This is true in the sense that the value for Tjunction Max that most enthusiasts will use in temperature utilities will be an estimate or average just like Tjunction Target. However, as discussed later on, Tjunction Max is not a fixed value.
Intel provided a small range of Tjunction Targets at the August IDF and for all CPUs at the October IDF in 2008. Many debated how correct some of those values were as they caused strange readings when used (below ambient on air cooling). After a bit of persuasion from certain individuals, Intel subsequently updated several of the Tjunction Targets to more realistic values, claiming a “mistake.” Unfortunately this update went largely unnoticed except by temperature utility developers, and many guides still recommend the old values.
In this post Tjunction Target will be used when referring to Intel’s “specifications” for Tjunction Max.
“Stuck” sensors: All sensors have a point at which they no longer respond to decreases in temperature. In many cases, particularly with the sensors used on 45nm Core 2 CPUs, these sensors stop responding at temperatures well above ambient and make idle temperature appear unrealistically high. This is often referred to as “sticking” by users and “bottoming out” by Intel. While the temperature is above the “sticking” point the sensors usually act normally.
Contrary to popular belief, this is not a fault and is not a justification for an RMA.
2. More detail and answers to common questions
How is Tjunction / Core temperature calculated?
Without calibration the temperature is calculated using a very simple formula:
Tjunction = (Tjunction Max – DTS output)
Recall that the DTS output is a relative Distance to Tjunction Max. So for an example, let’s assume that the Tjunction Max being used by a particular program is 100 C and the DTS output is 50.
Tjunction = (100 – 50)
Tjunction = 50 C
The DTS has a 7-bit output with a range of 0 to 127. The hotter the CPU, the lower the DTS output and the closer to Tjunction Max. When Tjunction Max is reached, the DTS output will be 0. An anomaly occurs when the CPU gets hotter than Tjunction Max. Because the DTS can’t report a negative Delta to Tjunction Max, it “wraps around” to 127 again, resulting in the following:
Tjunction = (100 – 127)
Tjunction = -27 C
To my knowledge, only Real Temp correctly offsets this anomaly and reports the temperature as 101 C. Therefore if you see negative temperatures without using extreme cooling, you should be worried.
What Tjunction Max should I be using?
Many users stress over what value they should be using for Tjunction Max. Despite what most people think, Tjunction Max is not a fixed value for a whole range of CPUs like Intel makes it appear with Tjunction Target. It is subject to calibration error in the factory and has an undisclosed margin of variation (see “More information on Tjunction Max and DTS Calibration”). Tjunction Max can vary between individual CPUs and even between individual cores of the same CPU, particularly in Core 2 Quad CPUs which are Multi-Chip Modules, but also to a lesser extent in Core 2 Duo and Core i7.
Without thorough testing it is impossible to know exactly what Tjunction Max is for each core, nor is it particularly important. As I’ve said multiple times, the function of the DTS is to control thermal protection mechanisms. These trigger based on the DTS output (Distance to Tj Max), not the temperature of the CPU. Absolute temperature is completely irrelevant to proper functioning of thermal protection mechanisms.
My temperatures are different between the cores, is something wrong?
As mentioned before, Tjunction Max is not a fixed value. At full load and a high temperature most of the cores in a Core 2 Duo and Core i7 should report fairly similar temperatures. If one or two cores are a couple of degrees out and you are using a utility that allows you to alter the currently used Tjunction Max then you can change it to align the readings more closely. The important thing is that you only do this under full load on all cores. Do NOT change Tjunction Max if only your idle temperatures aren’t the same.
There is another factor which affects differences in readings particularly at idle and low load. This is known as Slope Error and refers to the decreasing accuracy of a sensor as the Distance to Tjunction Max increases (and temperature decreases). It is important to know that:
1 Degree change in actual core temperature does not mean the DTS output will change by 1.
Therefore the reported temperature may not reflect the actual temperature. Some sensors are so poor at idle that they can be 10 degrees or more higher or lower than the actual temperature, but won’t always be. The 45nm Core 2 CPUs are the worst for poor idle accuracy. Real Temp has a function for inputting an idle calibration factor which can align idle temperatures more closely. Note that if one sensor “sticks” above another you can not calibrate it properly and should ignore it until you reach a temperature above the sticking point.
On a final note, Core i7 stores Tjunction Target in a software-readable Model-Specific Register so that if Intel changes it at any time (it sometimes changes with a new stepping) software will not have any problems adjusting to the change. Additionally, the DTS used by Core i7 is vastly superior to that of Core 2, and the real Tjunction Max is quite consistent with Tjunction Target, so if you have an i7 you don’t even need to think about Tjunction Max.
More information on Tjunction Max and DTS calibration
One of Intel’s requirements for thermal throttling is that ideally it doesn’t happen below Tcase Max. Intel calibrates each DTS so that regardless of where a DTS calibration point falls inside the margin of variation it will not cause the CPU to throttle below Tcase Max.
As you can see from the diagrams, the margin of variation is not disclosed by Intel. The DTS calibration point is X above Tjunction Target. The margin is ±X around the DTS calibration point. At the lowest end of the margin (-X) we have Tcase Max, and as the third diagram shows, this is the lowest point where throttling can occur (DTS = 0 or Tjunction Max).
So what does mean if we give X a hypothetical value? If Tjunction Target is 100 C and X is 5 C, then the DTS calibration point will be 105 C. Therefore the margin of variation will have a range of 2X above Tjunction Target and the actual temperature of a particular chip will be somewhere between 100 and 110 C.
3. Recommendations and additional reading
Recommended software:
Core Temp: Useful if you don’t want to do any calibrations at idle. It works on AMD setups as well.
Real Temp: Allows calibration of idle and load temperatures, can sound an alarm at a specified temperature and also shut down the computer at a specified temperature. Perhaps the most useful feature is the Sensor Test which allows you to check how your sensors respond over a full load range, making stuck sensors obvious.
SpeedFan: Advanced users may prefer to use SpeedFan to calibrate both their Tcase and Tjunction temperatures, as well as view temperatures for other system devices. If you intend to use SpeedFan you should refer to the last link in the “Additional Reading” section below.
Additional Reading:
Real Temp documentation
Real Temp thread at XtremeSystems
Core 2 and Core i7 Temperature Guide at Tom’s Hardware







6 Comments
[...] Отговор: Intel Core i5 3570K 3.4/3.8 GHz Ето ти дефиниция и обяснение на Tcase: Guide to Understanding Intel Temperatures | TechREACTION [...]
[...] “Junction temperature” or “Distance from TJ Max” (wikipedia basic info / Guide to Understanding Intel Temperatures – very informative, helpful, in-depth [...]
[...] Guide to Understanding Intel Temperatures | TechREACTION Intel Core i7-2600K (CM8062300833908 / BX80623I72600K / BXC80623I72600K) Intel® Core Note: i [...]
I am using a Core2 Quad Q9650 and have the strange situation where the average of the core temperatures is about 60C with a TCase temperature of 57C and then, sometime during the night, the TCase temperature shoots up to 68C with no change in the core temperatures. When I reduce the CPU usage from close to 100% to 5% the core temperatures drop to an average of about 38C but the TCase temperature does not change. This has happened about 5 times in the last fortnight, always during the night. When I switch the computer off for an hour or so and then back on again the TCase temperature is back to normal. I hope this does not mean that the CPU is failing.
[...] what MANY others have said on this forum. Stay below 85C on the Tjunction and you are fine. http://www.techreaction.net/2009/10/…-temperatures/ [...]
[...] CookieSayWhat Alright, I just noticed a 10C difference and was like that's huge. Here's why: http://www.techreaction.net/2009/10/…-temperatures/ [...]