Thermal information about Intel Skylake CPUs

sbab123

Reputable
Mar 21, 2016
119
0
4,680
Hi all,

I bought an Intel Core i5-6500 CPU and I was planning to do some thermal test with Prime95 to check how effective is my CPU cooler in cooling the processor.

So, I was looking at the thermal information here:
https://ark.intel.com/products/88184/Intel-Core-i5-6500-Processor-6M-Cache-up-to-3_60-GHz

I was looking at TCASE(maximum temperature allowed at the processor's Integrated Heat Spreader) in these details and the value is 71°C.

What will happen if I perform the test under 100% load and the TCASE temperature rises to around 80°C? Will my CPU not get damaged during the test since the maximum CPU temperature allowed is 71°C?

Also, why does Intel not give more thermal information about the CPU, such as throttle temperature, maximum junction temperature etc?

Please let me know your thoughts.

Regards
 
Solution
@ Ecky; Agreed.



@ sbab123; Your question opens a huge can of worms. There's no short direct answer, except to say that for your CPU Generation, it varies. It's a complicated discussion that involves the "gradient" issue, which I've shared with Ecky extensively.

Here's the simplified basic version:

If the CPU is Sandy Bridge (2nd Gen) or earlier, the Integrated Heat Spreader (IHS) is soldered to the Die, so Tcase (CPU temperature) is reliably 5C lower than Tjunction (Core temperature).

If the CPU is Ivy Bridge (3rd Gen) or later, the IHS is TIM'd on the Die with a Dow-Corning thermal compound to reduce costs. An ideal CPU...
Most intel processors have been able to manage their own temperature for a while now. They won't get hot enough to fail even if you remove the heatsink. (Please don't try that...)
 


I understand that there is no way to read TCASE because there's no sensor there. But say I perform the thermal test and measure the Core temperature, which turns out to be something like 79°C. I can assume that TCASE will be higher than 71°C since it is only around 5°C less than the Core temperature.

Is this a good thing in this case?
 
How can you know it's about 5c less than core temp?

When I removed the heatspreader from my CPU and ran it bare die, the core temperature went down by about 25c, which suggested to me that the thermal gradient was about 25c through the heatspreader. YMMV.
 
We have. As far as I'm concerned, Tcase is not terribly useful, especially now that Intel has quit listing it for their CPUs - Intel has removed Tcase entirely as of Kaby Lake, and Kaby Lake is not significantly different from Skylake. The only published temperature specification on Ark.Intel for desktop Kaby is Tjunction, which is 100c, same as mobile CPUs going back to Haswell.

https://ark.intel.com/products/97123/Intel-Core-i5-7500-Processor-6M-Cache-up-to-3_80-GHz

 
Yes, I agree. Except for developers, Tcase never has been useful, except to serve as a point of confusion for end users. I made the appropriate changes in the Temp Guide on Kaby Lake release Day 1. Intel's switch from Tcase to Tjunction is reflected in Sections 1, 3, 5, 7 and 8.


You will find that just as in prior Generations, the Tcase specification is still included in the Datasheet:

• 7th Generation Intel® Processor Datasheet for S-Platforms, Volume 1 - http://www.intel.com/content/www/us/en/processors/core/7th-gen-core-family-desktop-s-processor-lines-datasheet-vol-1.html

As you know, out of respect for the hard-earned property of others, I can not and do not, in good conscience, suggest, encourage or condone the operation of Intel CPU's near Throttle temperature (TjMax), even though they might be capable. In my professional career, academic credentials not withstanding, if I've learned anything about electronics, I know that heat over time damages electronics.

Although it might seem a bit conservative to the few knowledgeable users such as yourself, after devoting myself to studying this topic for over 5,000 hours spanning more than 10 years, including much painstaking hands-on testing, it is my considered opinion that sustained Core temperatures of 80C is a responsible recommendation for a practical real-world Core temperature limit.

Let's keep in mind that overclocked or not, some CPU's become unstable over 80C regardless of Vcore, while other have TjMax values as low as 80C. Intel even implemented Configurable Thermal Design Power (cTDP) and Scenario Design Power (SDP) in 6th and 7th Gen processors where throttling can occur at just 80C.

Among numerous other reasons, with so many advanced and inexpensive cooling solutions currently available, I see no need to allow Core temperatures to run above 80C.

CT :sol:
 



How am I supposed to know Tjunction for Core i5-6500, Ecky. This information is not available on Intel's website for Core i5-6500?
 
Tjunction max for all recent Intel CPUs is 100c. Tjunction = core temperature.

EDIT: And while I do feel CT's advice is conservative, it's definitely backed by a lot of research and experience. Keeping your core temp at or below 80c during typical usage seems a very reasonable recommendation.
 

sbab123,

See page 93, Table 5-4: Junction Temperature Specifications

6th Generation Intel® Processor Datasheet for S-Platforms, Volume 1 - http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/desktop-6th-gen-core-family-datasheet-vol-1.pdf


Ecky,

As a blanket statement, that doesn't quite work. Tjunction Max for all 7th Gen i5 and i7 "T" variants (35 Watts) is only 80C (cTDP Down).

Here's an example: Intel® Core™ i5-7500T Processor - http://ark.intel.com/products/97121/Intel-Core-i5-7500T-Processor-6M-Cache-up-to-3_30-GHz

The i3, Pentium and Celeron "T" variants (35 Watts) are 92C, while the i3 "TE" variant (also 35 Watts) is 88C.

I haven't cross referenced the 6th Gen Datasheet, but they're most likely the same as 7th Gen.

CT :sol:
 
OK Cool.

So you guys have convinced me that the Core temperature should not be above 80C....As long as I monitor this, my CPU will not be damaged by heat...Good!!

But out of curiosity, when Core temperature is 80C, what temperature is the Tcase?
 
Thanks CT.

@OP, The only way to know for sure is to measure it. I don't personally know of anyone who has machined out a groove in either heatsink or IHS to put a sensor there. It's not something that you can measure in any practical way.
 
@ Ecky; Agreed.



@ sbab123; Your question opens a huge can of worms. There's no short direct answer, except to say that for your CPU Generation, it varies. It's a complicated discussion that involves the "gradient" issue, which I've shared with Ecky extensively.

Here's the simplified basic version:

If the CPU is Sandy Bridge (2nd Gen) or earlier, the Integrated Heat Spreader (IHS) is soldered to the Die, so Tcase (CPU temperature) is reliably 5C lower than Tjunction (Core temperature).

If the CPU is Ivy Bridge (3rd Gen) or later, the IHS is TIM'd on the Die with a Dow-Corning thermal compound to reduce costs. An ideal CPU will have a gradient of 7C, but a "defective" CPU may have a gradient as high as 15 to 25C, which indicates poor thermal conductivity and will have comparatively high Core temperatures. The tighter the gradient, the better the thermal conductivity and cooling, and vice-versa. This is essentially why owners began "delidding" their CPU's.

From Silicon Lottery, where they offer de-lidding service: https://siliconlottery.com/collections/frontpage/products/delid

"Temperature improvements vary depending on processor sample and architecture. Temperatures under an overclocked load (1.3V-1.4V) will decrease anywhere from 5° to 25°C. Typical temperature improvements are listed below:

Ivy Bridge: 10°C to 25°C
Haswell: 10°C to 25°C
Devil's Canyon: 7°C to 15°C
Broadwell: 8°C to 18°C
Skylake: 8°C to 18°C
Kaby Lake: 12° to 25°C"

Regardless of your CPU's actual Die to IHS "gradient", the Tjunction Max specification (Throttle temperature) remains constant, so Tcase values are more-or-less just intangible values on 3rd Gen and later CPU's.

If you read Section 8 in the Temp Guide ...

"Although Intel measures Tcase on the surface of the Integrated Heat Spreader (IHS), they also calculate Tcase Specifications based on a combination of processor TDP and stock cooler TDP, which is expressed in Watts. Different cooler models with different TDP values are packaged with different TDP processors. Several Generations of Quad Core processors were packaged with a universal 95 Watt TDP cooler, but the i7 7700K / 6700K and i5 7600K / 6600K cooler is 130 Watts TDP and is sold separately: Intel’s Skylake Cooler - http://vr-zone.com/articles/this-is-what-intels-first-cpu-cooler-for-skylake-looks-like/97189.html.

Compared below are three Intel processor / cooler combinations with respect to TDP and Tcase Specifications:

Example 1: i7 2700K 95 Watts TDP / Cooler 95 Watts TDP / Difference 0 Watts / Tcase 72C.
Example 2: i7 3770K 77 Watts TDP / Cooler 95 Watts TDP / Difference 18 Watts / Tcase 67C.
Example 3: i7 6700K 91 Watts TDP / Cooler 130 Watts TDP / Difference 39 Watts / Tcase 64C.

When the cooler TDP is higher than the processor TDP, Tcase Specifications are lower, just as when the stock cooler is upgraded to a higher TDP aftermarket cooler, Core temperatures are lower. Tcase Specifications are based on a combination of processor TDP and stock cooler TDP. This is the primary reason why there’s so much variation in Tcase Specifications, and reveals a bigger picture beyond Tcase numbers."

For simplicity purposes, I discuss only a 5C gradient in the Temp Guide, but as I've said, don't get yourself stuck on Tcase numbers. Tcase is a very misleading specification. It's Core temperature (Tjunction) which should be your only concern. This is the point I've tried to emphasize toward the end of Section 8.

CT :sol:
 
Solution
Thanks guys, specially Ecky and CompuTronix, for your contribution.

Now I am convinced that Tcase is of no use. Only Tjunction is important and the maximum is 100C but even then, Tjunction should not be more than 80C!

Regards