Intel Warning over Tcase Temperature

bulldog147 · 21 Jun 2011 at 14:11

I just went across to Intel forum and read all about i5 2500K and i7 2600K and I was not surprise that Intel had reply to someone on the forum to warn others not to exceed "Maximum Tcase" temperature of 72.6 degree on all cores for 24/7 in use. One person doing 80C on all cores and the cpu had died within a few months because the chip is 32nm not 45nm. The maximum vcore by Intel is not to exceed 1.35v. (even thought their intel spec say VID range within between 0.2500v and 1.5200v on their note should be ingorned).

So, keep your i5 and i7 below 1.35vcore and below 72.6C on all cores for 24/7 in use for longer lifespan.

Corasik · 21 Jun 2011 at 14:27

tCase is almost impossible to measure though, you have to embed a thermal probe into the heatsink, without breaking thermal conductivity!.

Most of us use tJunct, using the internal probes of the processors.

tJunct is a inverse measurement, with zero being the throttle point of the processor, and incrementing by 1 degree for each degree below the throttle point the processor is.

Most applications like coretemp estamate the cpu temperature by guessing (or using intel leaked documents) to work out what the throttle temperature is, and then subtracting the tJunct from that.

On the whole its generally considered that tJunct is around 15 degrees hotter than tCase!.

tCase is also for the entire cpu, not "per core".

32nm i5/i7's should never exceed 1.35v if you want them to live for any reasonable time.

RJC · 21 Jun 2011 at 15:06

I thought the T case was the sensor under the heat spreader and not the actual core temps.

bulldog147 · 21 Jun 2011 at 15:18

RJC said:
I thought the T case was the sensor under the heat spreader and not the actual core temps.

I thought that one as well, but I just spoken to Intel via chat support and he say 72.6C on all 4 cores temperatures using RealTemp, CoreTemp or HWMonitor or manufacturing board temperature sensor.

Jokester · 21 Jun 2011 at 15:19

I've been banging on about the VID values in the datasheet ever since it came out. It is not anything to do with safe Vcore but just the range of values that the VID register on the CPU may be interpreted as. Often see the 1.52V value quoted as being the maximum safe value, the datasheet doesn't actually give one.

bulldog147 · 21 Jun 2011 at 15:24

Intel gave me this spec data sheets here: http://download.intel.com/design/processor/datashts/324641.pdf

mmj_uk · 21 Jun 2011 at 15:25

http://forums.pureoverclock.com/showthread.php?threadid=12386

What is Tcase?

Tcase is the temperature measure from the center point of the surface of heatspreader. Some motherboards use calculation by coding algorithms in BIOS, others may use surface mounted diode sensors. These methods of measurement is not the same as what and how the processor maker have prescribed in obtaining Tcase temperature as stated in figure 6.2. The only way to measure processor temp accurately or at least close enough to Tcase temp listed in the processor specification is to use a calibrated thermometer and place a temperature probe at the center of heatspreader. This is ann impossible method for most of us but there is an alternative method is to place a temp probe close to the center as possible then add 10°C to the reading to provide a temperature equivalent to Tcase specified.

72C on the heatspreader would probably take far higher temperature on the cores.

bulldog147 · 21 Jun 2011 at 15:29

If all cores temp is 72.6C on core 0,1,2 and 3, it would mean the CPU temperature will be lots higher than 72.6C!

mmj_uk · 21 Jun 2011 at 15:35

bulldog147 said:
If all cores temp is 72.6C on core 0,1,2 and 3, it would mean the CPU temperature will be lots higher than 72.6C!

Only if the transfer of heat from the cores to the heatspreader was 100 efficient, I'm no expert but imo the cores are always going to run hotter than the heatspreader - it's the reason people used to remove the heatspreader so they could get direct contact with the core, they got lower temps by taking the heatspreader out of the equation because the heat was being transferred from the cores to the heatsink more effectively.

as905 · 21 Jun 2011 at 17:15

here are the temp reading for my cpu,can someone please explains what is cpu package,cpu IA cores & cpu GT cores ? also which one is the Tcase temp ?

CPU 28c
CPU Package 38c
CPU IA Cores 38c
CPU GT Cores 33c
CPU1 37c
CPU2 38c
CPU3 33c
CPU4 37c

Redmint · 21 Jun 2011 at 17:32

bulldog147 said:
If all cores temp is 72.6C on core 0,1,2 and 3, it would mean the CPU temperature will be lots higher than 72.6C!

Not this debate again.

Lets go back and have a look at the i7 bloomfield series shall we?

From the Intel spec sheet:

TCASE 67.9°C

That's right, 68C.

Now what temp were people happy to run them up to for stress testing, 80C? Some people were pushing 85C core temps on hot days? They didn't actually throttle until 100C I believe. As others have said the safe Tcase and Tcore temps are different by about 10-15C.

Justintime · 21 Jun 2011 at 18:16

bulldog147 said:
I thought that one as well, but I just spoken to Intel via chat support and he say 72.6C on all 4 cores temperatures using RealTemp, CoreTemp or HWMonitor or manufacturing board temperature sensor.

Yes, 72.6C is the max.. look at all of those dead Q6600, dead 920s etc.. that exceeded their so called tcase max.. oh wait.
Dude would you realistically take advice from the purple shirt bridgade or actually take DIY tips from the average salesperson at B&Q? The chat blokes probably have a script to follow and possibly don't even have much technical experience. A coretemp or realtemp reading of say 75C might well be more like 65C Tcase. I tend to stay about 15-20C away from the TJmax countdown to 0, i don't even look at temps much because software just guesses that, if for instance it guessed that a chips Tjmax was 100C and the distance to Tjmax was 15c, then it would read 85C, however if it was actually 90c it would really be 75C.. just an example.

Redmint · 21 Jun 2011 at 18:23

As above really, 'some guy on chat support' doesn't really know any more than a quick search on google will tell you most of the time. I recently had someone from T-mobile try and tell me that my download speed doesn't affect my speed of browsing the internet...

Mr Krugga · 22 Jun 2011 at 00:48

Justintime said:
Yes, 72.6C is the max.. look at all of those dead Q6600, dead 920s etc.. that exceeded their so called tcase max.. oh wait.
Dude would you realistically take advice from the purple shirt bridgade or actually take DIY tips from the average salesperson at B&Q? The chat blokes probably have a script to follow and possibly don't even have much technical experience. A coretemp or realtemp reading of say 75C might well be more like 65C Tcase. I tend to stay about 15-20C away from the TJmax countdown to 0, i don't even look at temps much because software just guesses that, if for instance it guessed that a chips Tjmax was 100C and the distance to Tjmax was 15c, then it would read 85C, however if it was actually 90c it would really be 75C.. just an example.

Plus the sensors are rarely calibrated for anything but thermal throttling.

IPAddis · 22 Jun 2011 at 08:43

A BT engineer once told me that our work router had reset itself to defaults due to "cosmic rays". Riiiiight.

Working for a large corporation does not instantly make you a genuis or product specialist.

Gripen90 · 22 Jun 2011 at 08:59

IPAddis said:
Working for a large corporation does not instantly make you a genuis or product specialist.

Nope but certainly increase your chances to get away with a bad excuse :rolleyes:

Martini1991 · 22 Jun 2011 at 09:13

bulldog147 said:
If all cores temp is 72.6C on core 0,1,2 and 3, it would mean the CPU temperature will be lots higher than 72.6C!

Lunawolfy · 22 Jun 2011 at 09:44

Surely the real experts are all of us, the enthusiasts, who push their chips as far as they go, then after some time a few brick and we build a knowledge of safe temps/volts

ShortWarning · 22 Jun 2011 at 09:58

IPAddis said:
A BT engineer once told me that our work router had reset itself to defaults due to "cosmic rays". Riiiiight.

Amazing!

rikky · 22 Jun 2011 at 17:55

Lunawolfy said:
Surely the real experts are all of us, the enthusiasts, who push their chips as far as they go, then after some time a few brick and we build a knowledge of safe temps/volts

So if that is the case what do the "experts" say is the maximum temp to run a 2500k@24/7??