Another issue is thermal cycling operating at high temperatures actually introduces failure mode Nvidia failed to notice in the design of the BGA packaging for their GPUs. A lot more failures occurred in laptops because they are frequently powered up and down with elevated temperatures. These still happened in desktops though as well but thermal reliability is a tricky subject to design for.

http://www.theinquirer.net/inquirer/news/1004378/why-nvidia-chips-defective

No BS there its all in the solid mechanics, materials engineering.

Physical defects from a run of Nvidia chips in 2008 are outside the scope of this discussion. A defective manufacturing process can certainly have an impact on many aspects of a chips capabilities [and even class].

It really is relevant the metals used to solder things down have a finite fatigue life to them there is no unlimited life limit for non-ferrous metals (On such a small scale any small difference in the material expansion can induce large stresses), Nvidia just made a very bad mechanical design for their BGA packages and is an example of extremely short lifespans. Electro-chemical devices have temperature dependent limits even the heat pipes have a upper an ideal operating region for best performance.

It was not a manufacturing defect it was a engineering error that they denied and got into deep trouble for trying to cover it up for a while. They could have over designed the package but instead under designed it in error or negligence.

It is very simple, although thermal design is complex and there is not a direct relationship temperature matters in engineering (One constant uniform invariant ideal temperature would be ideal but that is difficult to achieve in practice). If it didn't then all processors would run at 100C because it would be cheaper to make the computers and the limiter makes it safe there too but it won't last very long and the supporting parts won't either.

While interesting, it really is only relevant to you and your desire to further an argument that is outside the scope of the thread.

Ok thermal engineering knowledge is apparently outside the scope of thermal reliability of microprocessors.

but that doesn't change the underlying physics, the growth of intermetallic compounds in the solder joints is dependent on temperature, higher the temps the faster they grow. Thick intermetallic layers are bad because of their brittle nature, making the joint more prone to crack nucleation and consecutive propagation of the cracks either in the IMC layer itself or on the solder/IMC boundary. And the driving forces of cracks are caused by temperature cycling with differencies in the thermal expansion coefficients of the materials used and the local temperature gradients in the 'system'. Higher the temp delta, higher the stresses caused. So operating for long periods of time at near the (short time) temperature limit of cpu is silly... (though they can take quite a lot of beating to actually fail, I wouldn't risk it. Some will fail sooner than others, it is a statistical phenomenon after all.)

Materials science!, Crack propagation, you can't stop it once it's started (It is always assumed to be started) its only a matter of time (Well you can compress it back to close but in cyclic loading you cant). I was never very good at all the material science courses but I can appreciate how complicated it gets and how there are many factors involved in designing/selecting appropriate materials for the expected life and operating conditions. But in simple terms infinite life in engineering is about 1 million cycles in some courses, which isn't really unlimited in the sense that it's not infinite. (Thermodynamics already doomed the universe one way or another)

Sorry if my single sentence was overly complex. Let me clarify. You'r point is contextually irrelevant. The values herein stipulated by other members as being 'too high' or 'very hot' in the context of 'overheating' on a chip that is no way comparable to a run of defective chips from 3 years ago IS the scope of the OP's issue and subsequent discussion. Your inclusion of this defective run of chips from 2008 is both irreverent and outside the scope of the original discussion which had nothing to do with defective chips or the manufacturing process.

If you want to talk about thermal engineering and the manufacturing process then by all means, go for it - just don't take the fact that I'm ignoring this thread now as a sign of anything other than lack of interest in the new subject matter. The main points have been made and the OP is likely well informed at this stage [and might even chime in to discuss this new direction].

thats the case for something like vibrating machinery parts or something like that iirc, with fine electronics and thermal cycling it comes down to (tens of) thousands or roughly there abouts

Thats probably going to be the case as I doubt glass can survive thermal cycling a million times it was just to illustrate that there is a designed lifespan. 10k cycles just makes it potentially even shorter.

But it still remains the fact that ideally you want to be as close to whatever the optimum is for a system to last the longest. (Near thermal max limits is normally not optimal)

The i5 starts to lower speed and disables turbo mode at around 70-80 degrees I believe.