It has been a long time coming, but the pieces are slowly falling into place. Intel has now released the official Tjunction Max value for all 65nm and 45nm Core 2 processors at the Intel Developer Forum in Taipei.
In a previous report on the August IDF presentation in San Francisco, Tom’s Hardware discussed how the information provided by Intel had very little real-world value. The reason Tjunction Max (the temperature where thermal protection is engaged) is not the silver bullet for 45nm Core 2 processors is because the sensors suffer from extremely high amounts of "slope error," that is, they become less accurate as the real temperature moves further from Tjunction Max.
An Intel document (PDF) describes the error on Atom processors, which use the same or similar DTS as those on 45nm Core 2 processors:
"The digital thermal sensor (DTS) accuracy is in the order of -5°C +10°C around 90°C; it deteriorates to ±10°C at 50°C. The DTS temperature reading saturates at some temperature below 50°C. Any DTS reading below 50°C should be considered to indicate only a temperature below 50°C and not a specific temperature. External thermal sensor with “BJT” model is required to read thermal diode temperature."
According to Intel, if the actual temperature is below 50°C the temperature can’t be trusted at all. With calibration, the slope error can be offset to an extent, but the reported temperatures will never be as accurate as those which are reported by the DTS on 65nm processors. Furthermore, the sensors can sometimes "stick," particularly at lower temperatures, and the worst of these sensors can’t be calibrated properly.
Unlike their 45nm counterparts, the DTS for 65nm is much less affected by slope error, so that even a temperature readout that has not been calibrated can give a reasonably close representation of the actual temperature. The only major factor which will affect readings is Tjunction Max. Since the 65nm CPUs were released, enthusiasts and developers of temperature monitoring software have debated over the Tjunction Max. It was hoped that Intel would disclose these details at the August IDF; however, we were to be disappointed. A few months later and Intel has finally decided to disclose the Tjunction Max for every processor in the Core 2 line, both 45nm and 65nm, as well as Xeon server CPUs. The Tjunction Max values for all Core 2 processors are as follows:
Model | Target Tj (B2/B3/L2) | Target Tj (G0/M0) |
---|---|---|
E6000 and E4000 series | 70°C | 80°C |
X6800 | 75°C | 85°C |
Q6000 series | 80°C | 90°C |
QX6000 series | 80°C | 90°C |
QX68xx series | 80°C | 80°C |
E1000 series | 75°C | 85°C |
Model | Target Tj (Stepping unspecified) |
---|---|
E8000 and E7000 series | 100°C |
Q9000 and Q8000 series | 100°C |
QX9650 | 95°C |
QX977x | 85°C |
Model | Target Tj (Stepping unspecified) |
---|---|
E7220, E7210 | 80°C |
7100 series | 100°C |
Model | Target Tj (B2) | Target Tj (G0) |
---|---|---|
5080, 5060 | 80°C | 90°C |
5063, 5050, 5030 | 80°C | 90°C |
5160 | N/A | 80°C |
5150, 5140, 5130, 5120, 5110 | N/A | 80°C |
5148 | N/A | 80°C |
L5138 | N/A | 100°C |
3000 series | 80°C | 90°C |
Model | Target Tj (Stepping unspecified) |
---|---|
X7350 | 90°C |
E7340, E7330, E7320, E7310 | 80°C |
L7345 | 80°C |
X5000 series | 90/95°C |
E5000 series | 80°C |
L5000 series | 70°C |
L5318 | 95°C |
X3230, X3220, X3210 | 90°C |
XE | 90°C |
XEE | 80°C |
Model | Target Tj (Stepping unspecified) |
---|---|
L3104 | 95°C |
Model | Target Tj (Stepping unspecified) |
---|---|
X52xx series | 90°C |
E5240 | 90°C |
E5220, E5205 | 70/90°C |
L5240 | 70°C |
L5238, L5215 | 95°C |
E3120, E3113, E3110 | 95°C |
L3110 | 95°C |
Model | Target Tj (Stepping unspecified) |
---|---|
E7440, E7430, E7420 | 90°C |
L7445 | 80°C |
X54xx series | 85°C |
E54xx series | 85°C |
L5408 | 95°C |
L5430, L5420, L5410 | 70°C |
X33xx | 95°C |
L3360 | 90°C |
X33xx series | 95°C |
L3360 | 90°C |
Model | Target Tj (Stepping unspecified) |
---|---|
X7460 | 85°C |
E7455 | 85°C |
L7455 | 85°C |
The above data will mean that for some users the temperatures currently being reported for their cores are actually quite a fair amount off; assuming we take Intel’s word for these maximum temperature values. For example, most temperature reporting programs would use 85°C as the Tjunction Max for a B2 E6600. According to Intel, the "official" Tjunction Max for this processor is only 70°C. That means that for a program which does not take into account any slope error, the temperature would be reported 15°C too high.
Of further interest is that many enthusiasts have done extensive testing to approximate Tjunction Max where it had not been specified. For the 45nm processors, Intel’s data generally aligns with the results of community testing. However, for 65nm the values provided by Intel are significantly different for some processors than what testing has shown. It may very well be that Intel has caused even more confusion than before.
You can expect most software developers to update their programs in the near future to reflect the information provided by Intel.
Update: Unfortunately, the data provided by Intel is not actually what was expected. The developer of Core Temp received an email last month from the presenter of the August IDF presentation on the DTS specifications which said that Tjunction Max would be disclosed for all processors in Taipei. In reality, Intel disclosed a value known as "Target Tjunction," which should not be used as Tjunction Max as it will result in unrealistic temperature readings. Taking the earlier example of the B2 E6600, if you had a low ambient temperature, changing Tjunction Max so that it aligned with the Target Tjunction may result in temperature readings close to or below 0°C when at idle. Temperatures below ambient aren’t possible on normal air cooling.
It seems that yet again Intel has avoided providing any valuable information at IDF. Intel hasn’t even stated exactly what Target Tjunction actually is. It is not recommended to change Tjunction Max to align with these values.