Post #1

kmarc — Fri, 04 Apr 2008 23:08:57 +0800

I've noticed a lot of forumers asking about CPU core temperature even though there are lots of information on the net. As such, I have decided to create a guide that explains the very basic of CPU core temperature.

This guide concentrates mainly on Intel CPU as that is what I'm using now.

Disclaimer:
1) This is only a simple guide to provide information in as simple a format as I can make it. Any loss of CPU is NOT MY FAULT!
2) I'm not in the computer field so correct me if any facts are wrong
3) All the facts are compiled from the internet and easily available if you search for it
4) Any advise/comments/corrections are most welcomed!
5) The guide is long-winded but in the hopes of making you understand what CPU temperature is all about

References for this guide are as below:
1) Core 2 Quad and Duo Temperature guide - the best guide ever (by Computronix)!
2) Good info on thermal management (Intel)
3) Intel Thermal & Mechanical design guidelines - for C2D/penryn (Intel)
4) Intel Thermal & Mechanical design guidelines - for quads (Intel)
5) Accurate temperature reading - the problems (Anandtech)
6) The truth about processor degradation (Anandtech)
7) Overclocking's impact of CPU's life (by Joe Citarella)

Useful info/help from our forumers:
1) cks2k2

Some programs that you can use to monitor your CPU temperature (you can google for the latest version):
1) Core Temp
2) Real Temp
3) Speedfan
4) Everest
5) Sisoft Sandra
6) Your own motherboard's monitoring software e.g. Easytune5, ITE smartguardian.....

Index
1) Basics of a CPU (the brains of your computer)
2) CPU temperature sensors
3) CPU temperature - which value is important?
4) CPU temperature - which one should you follow? TcaseMax or Tjunction?
4a) Newest update (August 2008) : Published Tjunction for 45nm procs by Intel!!!
5) Is there such a thing as safe/optimum temperature?
6) How to reduce my CPU temperature?
7) Summary

------------ 1) Basics of a CPU (the brains of your computer) ---------------

Note : The term CPU in this guide means the processor itself and NOT your whole computer

Understand the parts of a CPU:
1) Whole CPU
2) IHS (Integrated Heatspreader)
3) Core Die
4) Cores

This is a whole CPU : Intel E2160 (left), AMD64 X2 4200+ (right)

[attachmentid=434108]

Each CPU has an IHS on top of it. The IHS is the flat metal piece that has all the wordings written on it.

The picture below shows a pentium 3 without the IHS (left picture) and with the IHS (right picture).

[attachmentid=434111]

Function of IHS:
1) To evenly distribute the heat generated from the die/cores
2) Provide mechanical strength for the substrate (the substrate is quite thin and mechanically fragile)
3) Protect the die - from heavy heatsink that requires a lot of force to hold (Imagine a BIG HSF sitting on a naked chip... )

Note : Some extreme overclockers actually remove the IHS so that their heatsink can come directly in contact with the chip underneath. This usually results in a lower temperature which is the main aim.

Inside each CPU, is the die (or chip which contains all the silicon integrated circuit) - can be seen on the above left picture
The die lies on a "substrate", which makes up the whole chip (above picture, green color material)

See this picture of a dual-core for a clearer understanding.

[attachmentid=434123]

From the above picture, you can also see 2 cores (Core0 and Core1) and an "advanced smart cache" region. These all make up the die of the CPU.

Even wonder why people like to show this picture of a silicon wafer?

[attachmentid=434136]

Well, each small box region is actually a die/chip! And the diameter of the wafer is only 200mm (20cm) - usually used for chipsets. See how many chips can fit into one 200mm wafer..... You should see a 45nm process die with a 300mm wafer.... (Intel CPUs are from these)

Understand so far?

----------------- 2) CPU temperature sensors-------------------------

Current Intel CPUs has 2 thermal sensors on its CPU:

1) A on-die thermal diode - located within the CPU die between the cores
--- measures Tdiode (which is CPU temperature)
--- the converted digital values of Tdiode is calibrated in BIOS and displayed by temperature software
--- so, BIOS can affect the accuracy of Tdiode (so now you know why BIOS update causes different CPU temp readings!!)

2) A DTS (Digital Thermal Sensor) - located within each core
--- measures Tjunction (Tj) (which is core temperature)
--- the converted digital values are factory calibrated and displayed by temperature software
--- As such, BIOS does not affect the accuracy of Tjunction

Quote from Intel info : "In order to improve temperature reading, multiple sense points monitor different hot spots on the die and report the maximum temperature of the die. An independent temperature reading from each core is available, with optional reporting of the maximum temperature of the entire die, e.g., the highest temperature of both cores"

From what I can understand from the above, the Tdiode is measured from different places and not a single point. The maximum temperature of the die is then shown as the CPU temperature. No mention of the location of the DTS except that it is in each core......

Important :
Intel C2D, Pentium dual-core & Celeron dual-core all have both on-die thermal diode and DTS
Intel Quad (e.g. Q6600) and penryn CPUs e.g. (E8400, E8500) do not have an on-die thermal diode any more. They have been replaced by the DTS.

Example:
C2D Quad CPUs - has 4 Tjunction sensors
C2D Duo CPUs - has 1 Tcase and 2 Tjunction sensors

------------- 3) CPU temperature - which value is important? ---------------

So, in order to understand which value is important in a CPU, you have to understand a few variables that are set by Intel.

TcaseMax (TcMax)- This value stands for the maximum temperature measured at the geometric center on the topside of the processor IHS (see picture below)
--- It is termed as "Thermal specification" in Intel's CPU spec sheet
--- It is NOT indicated by the on-die thermal sensor on the die
--- The temperature is designated as Tc

TjunctionMax - This value stands for the maximum temperature at the junction between the processor die and the PCB it sits on, this is usually much higher than the TCaseMax value.
--- It is the temperature of a CPU when throttling occurs
--- Throttling reduces the speed/frequency ™ or both the speed/frequency and voltage (TM2) of the CPU
--- As such, when throttling is activated, the CPU temperature should come down......
--- Important : As such, it is important to enable the "CPU Thermal Monitor 2 (TM2)" in BIOS if you are overclocking, just in case something happens and you forgot to look at the CPu temperature. It is a CPU overheating protection function.

Intel's definition of "thermal specification" - The thermal specification shown is the maximum case temperature at the maximum Thermal Design Power (TDP) value for that processor.

Left picture : The place where Tc is measured
Right picture : Intel's recommendation on how to measure Tc (you can refer to Intel Thermal & Mechanical design guidelines for further info)
[attachmentid=434858]

TCaseMax/Thermal specification
Why is this variable important? That's because Intel do published the maximum temperature a CPU can operate based on the TCaseMax.

See this example below:
[attachmentid=434873]

This data shows two steppings of Intel Q6600 - an older "B3" stepping and a newer "G0" stepping
If you look at their thermal specification:
B3 - 62.2'c
G0 - 71'c

For non-overclockers, they are very not important.
However, for overclockers, apparently, the higher the thermal specification, the better it is for overclocking.

These are a few advantages of having a higher thermal specification (which may include other factors like newer steppings/fixed bugs)
1) Uses less power and dissipate less heat
--- at idle (as it dissipate less heat at C1E state)
--- at load
--- when overclocked
2) Runs cooler due to the above
3) Better overclocking capabilities (in general) - as heat is the number 1 enemy for overclockers

You can read more about different stepping/thermal specification here:
Anandtech : Q6600 G0 vs B3

So now you know why everybody is scrambling for a Q6600 "G0" stepping....

Note : You can easily search for your own Intel's CPU information by googling this way "Q6600 sspec". The first result is usually the data sheet. So, if you have a E2160 L2 stepping, and you want to compare with E2160 M0 stepping, just google separately for "E2160 sspec L2" and "E2160 sspec M0".....

Tjunction/TjunctionMax
There is one important fact that needs understanding.
The DTS in each core:
- DO NOT give an actual temperature value e.g. 50'c (that means, it doesn't read your core temperature as 50'c)
- Gives a value relative to the Tjunction e.g. 35'c from Tjunction
- so basically what it does is "count down the margin to the core's maximum thermal limit. When this value reaches zero, the core temperature has reached its Tjunction set point" (Anandtech).
- so, when DTS gives a value of "zero", the core temperature has reached its TjunctionMax, resulting in CPU throttling...

For example, a DTS in an Intel Q6600 would give a value of 35 (NOT in Celcius/Fahrenheit, just a value) . So, if you know that the Tjunction of the Q6600, then you can easily calculate it's core temperature

- DTS reading : 35
- Tjunction of Q6600 : 100'c
- Actual core temperature : 100-35 = 65'c

Understand so far?

Now, the problem is, Intel DOES NOT supply the Tjunction value for their desktop CPUs!!!!!!!

QUOTE

Intel has never published Tjunction values for any CPU other than the mobile models. The reason for this is simple. Since mobile processors lack an integrated heat spreader (IHS), it is not possible to establish a thermal specification with respect to its maximum case temperature ("Tcase"), normally measured from an embedded probe located top, dead-center in the IHS (Anandtech)

QUOTE

More than a few programs have been released over the last few years, each claiming to accurately report these DTS values in real-time. The truth is that none can be fully trusted as the Tjunction values utilized in these transformations may not always be correct. Moreover, Intel representatives have informed us that these as-of-yet unpublished Tjunction values may actually vary from model to model - sometimes even between different steppings - and that the temperature response curves may not be entirely accurate across the whole reporting range. Since all of today's monitoring programs have come to incorrectly assume that Tjunction values are a function of the processor family/stepping only, we have no choice but to call everything we thought we had come to know into question. Until Intel decides to publish these values on a per-model basis, the best these DTS readings can do for us is give a relative indication of each core's remaining thermal margin, whatever that may be. (Anandtech)

The Tjunction value of each Intel processor is produced NOT by Intel, but an arbitrary value based on observations and measurements made by computer enthusiast.
These are a few estimated Tjunctions for some Intel CPUs:
- Q6600 G0 stepping - 100'c
- Q6600 B3 stepping - 85'c
- E6550, E6750, E6850 (All G0 steppings) - 100'c
- E2160 - 85'c

However, if you search on the net, there are always uncertainty regarding the Tjunction of a certain CPU, again, mainly because Intel doesn't provide that value for desktop CPUs.

Here's a good example of different software giving different Tjunctions for my Q6600. See the picture below

[attachmentid=437593]

Realtemp's Q6600 Tjunction is 95'c
Coretemp's Q6600 Tjunction is 100'c

Because of that, the temperature readings of realtemp is 5'c lower then coretemp's readings.

You can read more about it here : Accurate temperature reading - the problems (Anandtech)

Tjunction - is it correctly used in desktop CPUs?
For completeness sake, let me explain a bit regarding the term "Tjunction" which is incorrectly used. Below are extra info regarding Intel's definition of "Thermal specification":

QUOTE

Definition of thermal specification : The thermal specification shown is the maximum case temperature at the maximum Thermal Design Power (TDP) value for that processor (note - that means the TcaseMax). It is measured at the geometric center on the topside of the processor integrated heat spreader. For processors without integrated heat spreaders such as mobile processors, the thermal specification is referred to as the junction temperature (Tj).

So, in summary, thermal specification for desktop CPU (with IHS) and mobile CPU (without IHS) is different.
- Desktop CPU : Thermal specification refers to the TCaseMax
- Mobile CPU : Thermal specification refers to the Tjunction

As such, stricly speaking, desktop CPUs core's maximum thermal limit is incorrectly termed as Tjunction, which is supposed to be used only in mobile CPUs WITHOUT the IHS.

However, since almost all users and software programmers uses the term for desktop CPUs, we will just have to follow......

-- 4) CPU temp - which one should you follow? TcaseMax or Tjunction? --

Not only can you use both, you actually have 3 options:
1) TCaseMax
2) Tjunction
3) Delta to Tjunction

However, you need to calibrate the TCaseMax and Tjunction so that it more accurately reflects the CPU temperature.
Remember that:
- the most accurate measurement of TCaseMax is putting a probe in the dead-centre top of the IHS (which we average users cannot do)
- the Tjunction is NOT an actual temperature value but a value to indicate how far it is from a specified maximum core temperature (Tjunction)
- calibration of TCaseMax and Tjunction using temperature softwares e.g. speedfan attempts to make it more accurate but by no means reflects true CPU core temperature
- "Delta to Tjunction" - Delta is a mathematical term to mean differences between 2 values.
--- For example, the delta of 2 temperatures e.g. 50'c and 58'c is 8'c.
--- So, "Delta to Tjunction" means how far it is from the maximum core temperature (Tjunction)

Understand so far?

TCaseMax calibrations
The method of calibration are summarized from this excellent guide : Core 2 Quad and Duo Temperature guide - the best guide ever (by Computronix)!

Note : I will summarized this part later on. In the meantime, please refer to the above link

Tjunction calibrations
The method of calibration are summarized from this excellent guide : Core 2 Quad and Duo Temperature guide - the best guide ever (by Computronix)!

Note : I will summarized this part later on. In the meantime, please refer to the above link.

Delta to Tjunction
This is the easiest. Just use Core Temp, go to Options>Settings, then check the "Show Delta to Tjunction Max temp." You will get readings like the picture below (my Q6600 at load).

Left picture showing the new settings, right picture without the settings.
[attachmentid=436016]

The readings are directly from the DTS, so it is the most accurate reading you can get.

So, what is an acceptable value for "Delta to Tjunction". Well, nobody knows!!! From observations of "safe" temperature and recommendations by certains enthusiast, the acceptable value seems to be ABOVE 20-25'c (Note the word ABOVE, NOT below 20-25'c)

So, if you look at my Q6600 "Delta to Tjunction", I'm quite far from my maximum core temperature, so I'm quite "safe".

Note : Alternatively, you can use realtemp, which shows you the "Distance to TJMax" which has the same functions.

4a) Newest update (August 2008) : Published Tjunction for 45nm procs by Intel!!!
Looks like Intel finally published the Tjunction for desktop processors, although it is only for 45nm procs.

Here's the published Tjunctions:

The Tjunctions for the 45nm procs are out!

45nm Desktop Dual-Core Processors
Intel Core 2 Duo processor E8000 and E7000 series - 100°C

45 nm Desktop Quad-Core Processors
Intel Core 2 Quad processor Q9000 and Q8000 series - 100°C
Intel Core 2 Extreme processor QX9650 - 95°C
Intel Core 2 Extreme processor QX9770 - 85°C

Link : http://intel.wingateweb.com/US08/published...MTS001_100r.pdf

From the datasheet, it looks like Intel is planning to design a temperature sensor that outputs the temperature in degrees celcius! This would really help to reduce headaches in monitoring the core temperatures!

--------- 5) Is there such a thing as safe/optimum temperature? ----------

Note : If you're not an overclocker, then this question is not really important.

Most people will ask "What is a safe temperature to run my overclocked CPU?"

Well, there is actually no straight forward answer to this.

The "answer" depends on various factors such as:
1) Type of CPU
2) Maximum temperature
3) Duration of high temperature @ How you run your rig/CPU e.g. a few hours a day @ never switched off.....

Type of CPU
This is obvious as different CPU would have different thermal specification.

For example, take a Q6600's thermal specification (TcaseMax):
Q6600 B3 stepping: 62.2'c
Q6600 GO stepping: 71'c

Obviously, if both processors were running at a TcaseMax of 60'c, definitely the B3 will be the more "tortured" of the two CPUs!

And I guess it would be also true for mobile & server CPUs which are more reliable and can take higher temperatures....

Maximum temperature
Unfortunately, there is no absolute value for a safe temperature.
--- some would recommend 60-65'c
--- some would say 50-55'c
--- and some would run their procs at 70'c without any short-term problems

Just remember this : Heat is the number one enemy for your overclocked CPU (voltages comes in 2nd.....)

The higher the core temperature:
- the higher the risk of damage
- the shorter the life expectancy

Note : Remember that the lifespan of a CPU can be very long..... usually >10 years? >50 years?

In general, for current Intel processors, the recommendation is around 60-65'c

Very good in-depth explanation here :
The truth about processor degradation (Anandtech)
Overclocking's impact of CPU's life - by Joe Citarella

Duration of high temperature
Running your CPU at a high temperature for a few hours would definitely be better than running it 24/7

------------------ 6) How to reduce my CPU temperature? -------------------

Can be divided into different components:
1) Casing
2) Cooling (e.g. air, water, more extreme)
3) Air cooling - the fan setup
4) Thermal paste
5) CPU
6) Ambient temperature
7) Power saving methods
8) Overclocking/overvolting

Casing
> Have a good airflow in your casing
--- there are many guides on how to do this, which includes cable management, strategic placements of intake and outlet fans, no dust build-up in the casing or on the heatsink

> Open your side casing panel
--- also be wary of dust buildup with this setup

> Alternatively, you can run naked (no, not you la.... your computer... )
--- if you run your rig without a casing, there won't be any heat build-up that normally occurs in a closed-case system
--- be wary of dust buildup with this setup

Cooling
> Don't use stock CPU Heatsink fan (HSF)
--- use a good 3rd party coolers (e.g. air-cooling, water-cooling or more extreme)

> Ensure your HSF is seated properly
--- sometimes, your HSF is not in proper contact with the CPU (this can happen if you use push-pins to secure your HSF to your CPU)
--- if you have a 3rd party HSF that has push-pins and also screw kit, use the screw kit, as it ensures better contact

Air cooling - the fan setup
Single fan - for maximum cooling, you can:
- Set the CPU fan speed to run at maximum.
- Change to a more powerful fan (usually measured in "cfm", the higher the cfm, the better it is)
- Use a larger size fan e.g. from 8cm to 12cm

Multiple fans - for certain HSF, you can actually put 2 fans, or even 3 fans.
- one fan to blow on the HSF, one fan to suck out the air from the HSF

Thermal paste
Don't use the normal thermal paste that comes with the stock HSF (they are good but not as good as well-known ones)
Use good thermal paste e.g. Arctic Cooling MX-1 or MX-2, Tuniq TX-2
Currently, the best is Shin-Etsu, which is unfortunately not available here

CPU
> Lap your CPU or HSF
--- this will increase the surface contact between the CPU and HSF
--- lapping the CPU will void the CPU warranty.....

> Remove the IHS - not for the average users, more for extreme overclockers (as removing the IHS will void the warranty)
--- By removing the IHS, the CPU heatsink and come in direct contact with the CPU die/chip. This will cause a significant drop in core temperature....

Ambient temperature
Note : Ambient temperature means the general temperature at your location e.g. room, office, etc....
Lower your ambient temperature e.g. in air-con room

Power saving methods
Activate power management functions
--- for Intel CPUs, enable C1E and EIST function in BIOS
--- for AMD CPUs, enable Cool&Quiet function in BIOS

Overclocking/overvolting
As you know, the higher you overclock/overvolt, the more the heat is generated. Hence, the higher the CPU temperature.

Lower overvolting
--- remember the phrase "The point of diminishing returns"
--- This means that when you overclock and overvolt your CPU, there comes a point where you need to increase your vcore by a huge amount but only gain a few mhz in the process
--- for example, this is the results of an overclocked Q6600
----- vcore ----- overclocked speed
----- 1.30v ----- 3.0 Ghz
----- 1.35v ----- 3.3 Ghz
----- 1.40v ----- 3.4 Ghz
----- 1.45v ----- 3.43 Ghz
----- 1.50v ----- 3.44 Ghz
--- if you look at the example above, the point of diminishing return is around 1.40-1.45v, where increasing your vcore from 1.40v to 1.45v only gain you 30mhz (3.43Ghz - 3.40Ghz = 30mhz). In addition, increasing from 1.45v to 1.50v only gain you 10mhz (3.44Ghz - 3.43Ghz = 10mhz)
--- So, instead of running 1.50v at 3.44Ghz, you could run your CPU at 1.40v at 3.4Ghz, which would help to reduce your core temperature by a significant amount.....

---------------- 7) Summary -------------------

1) Heat is the number 1 enemy for your CPU. The cooler your CPU runs, the better it is

2) Software monitoring of the actual CPU temperature is not entirely correct.
--- This applies to ANY software available including popular softwares e.g. core temp, speedfan, etc

3) Delta to TjunctionMax (Distance to maximum core temperature)
--- Is the most accurate indication of your CPU's relative temperature
--- It is accurate as it obtains the readings directly from the DTS
--- It is NOT an absolute temperature reading (rather, it indicates how far away the temperature is, from maximum core temperature

4) TjunctionMax
--- Nobody knows the actual value of TjunctionMax (as Intel does not release the value)
--- Intel only releases the "Thermal specification" to indicate TcaseMax (not easy to measure)

5) There is no such thing as a safe temperature for your CPU
--- As it depends on various factors (as stated above)
--- Again, the cooler it is, the better.
--- In general, most enthusiast will recommend not exceeding 60-65'c

6) If you really really REALLY want the most accurate temperature reading, then follow the guide on how to place a temperature sensor on top of your CPU
--- To measure Tcase
--- Intel has a guide to this but it is not easy (involves making a groove in your IHS)

Lowyat.NET: Latest posts in CPU core temperature guide

Post #1