Shneiky :
DeathAndPain do you have any computer science, electronics, electrical engineering or related education of a bachelor or a higher level? Or basic knowledge of physics? Anyway, are you even of a legal drinking age?
Well, let's see... I am 40+ years of age... both my private and business activities have been centered around computers since 5th grade in school... I have studied computer science at the Technical University of Berlin... I used to be a network administrator for several years... I am a software specialist right now... yes, I think I do qualify.
Shneiky :
1 - Higher temperature makes the possibility of micro defects appearing in the structure in the chip a lot higher. Any disruption of the integrity of the chip itself could lead to electromigration. Google it. As well as Joule heating.
So far the theory, yes. Estimated lifetime of a CPU is ~12 years of constant operation. If this is reduce to 8 or even 6 years (which would be 50%), the owner will hardly care.
Shneiky :
2 - Higher temperature leads to higher consumption. Higher temperatures increase resistance(Physics 101), so the chip compensates by using more power.
A pretty weird logics, seeing that higher resistance usually means lower and not higher currents, and power consumption is defined as the product of voltage and current.
Either way, there is an upper limit to this, which is the throttling temperature, at which the corresponding mechanisms of the CPU will intervene and make sure things do not go the wrong way.
Shneiky :
3 - There is no magic silicon that runs forever. Ask any electronics/hardware specialist, or even here at Tom's.
See my response to 1.
Shneiky :
4 - Speaking from practice. The memory controller is the most easy to damage part in modern Intel CPUs. It is a lot less resistant to heat or voltage than the actual CPU cores. Why do I say this - in the past 2 and a bit years I had colleagues and family members with failing CPUs due to prolonged exposure to high temperature (but below throttling point). 5 Ivys and 3 Haswells died prematurely
Wow, you must really have a lot of computer fanatics among your colleagues and family members... Haswell is on the market for barely 6 months, and you not only know three persons who own one, but even three whose Haswell died prematurely. How many other colleagues and family members do you have that own a Haswell? 100? Or would you say that Haswell CPUs die like the flies, indicating a major mis-design?
I believe in something different here. It is known as major bs.
Shneiky :
5 - Current Intel CPUs are not "designed" to run at 100C. That is their TjMAX. It is maximum temperature of operation. Any higher temperature will result in inevitable damage to the chip.
That is what you keep repeating like a mantra. That does not make it any more true though.
As a side remark, Intel offers warranty for their products. In some countries, they are even forced to provide some warranty. This warranty exceeds the 6 months of Haswell market presence by a large margin. If it was not in accordance to the chip design to run a Haswell at its throttling temperature for an extended amount of time, don't you think Intel would include a corresponding rule in their specifications to protect themselves from countless warranty claims? In fact every single of the 8 CPUs in your vicinity that allegedly died from high temperature would be a valid case of warranty, because if your words are to be believed, these CPUs were always operated within their specifications.
Shneiky :
Anyway Haswell has a TCASE of 72.72°C. That is the maximum allowed temperature on the heatspreader. Any thermal monitoring software reports the temperature only of the cores, and not on any of the other components inside the CPU itself. Your posts were showing, that you make no distinction between CPU temperature and Core temperature.
Perhaps it would be a good idea if you looked into
this document. You may even know it already, seeing that this is the document where your 72.72°C originate from. Beginning on page 75, the thermal self protection mechanism of the Haswells is described. There are two of them, named TM1 and TM2, and they activate in escalating order to keep the CPU within its limits.
If you read the last paragraph of page 76, it reads:
"If TM2 is unable to reduce the processor temperature, then TM1 will be also be
activated. TM1 and TM2 will then work together to reduce power dissipation and
temperature. It is expected that only a catastrophic thermal solution failure would
create a situation where both TM1 and TM2 are active."
So the thermal self-regulation of the CPU is so good that only a catastrophic messup of the cooler, such as the cooler falling off the CPU, could possibly cause these self-protection measures to kick in.
Some time earlier Intel states:
"An underdesigned
thermal solution that is not able to prevent excessive activation of the TCC in
the anticipated ambient environment may cause a noticeable performance loss, and in
some cases may result in a TCASE that exceeds the specified maximum temperature
and may affect the long-term reliability of the processor."
I suppose this is what you base your stuff upon. However, the temperature difference between heat spreader and CPU cores is directly related to the quality of the cooling solution. If you have a good cooler, you can run your cores at a constant 100°C without the heat spreader (TCC) ever exceeding those 72.72°C, because the cooler is dissipating the heat efficiently enough. Of course, if you are employing a crappy cooling solution, then the heat will accumulate in the heat spreader area, causing the ambient to grow too hot in the long term.
I never suggested running the CPU with a crappy or improperly-mounted cooler. But I consider it equally silly to deduct from the heat spreader temp that your core temp is too hot when Intel, employing multiple heat-protection mechanisms, explicitly allows it. Especially when you demand a maximum of 70°C for the cores when even the heat spreader, divided from the cores by Intel's poor thermal grease, is officially fien to be hotter than that.
Shneiky :
6 - There is something called designed failure or at least that is how i call it. Maybe more appropriate is planned obsolescence. Google it.
I am well familiar with the concept, but there has never been any need for it in the CPU area. CPUs grow obsolete so quickly that there is no necessity to make them fail prematurely. All Intel could achieve here is major damage to their reputation. And if Ivys and Haswells failed as frequently as you claim it to be the case among the people that surround you, the forums would be full of complaints and people warning against Intel CPUs for reason of poor reliability.
The truth, however, is that countless people keep overclocking and overvolting their Ivys and, to a somewhat lesser extent, also Haswells with good success and no problems whatsoever, even though power consumption and heat dissipation rises linear with clock speed and even square with voltage, causing even a small amount of overclocking to have a high impact on heat dissipation.
Shneiky :
I got a key holder made out of 20 CPUs ranging from Pentium I up until Sandy I3/I5/I7s which died prematurely, due to heat. Took them from my dad's box of "failed memories".
Then your dad must have been sloppy on properly mounting adequate coolers. As I said before, I have administered a network with hundreds of CPUs (both Intel and AMD) before, many of them in cramped cases with poor airflow, and never saw a single of them fail.
Shneiky :
He does computers for living. He is an electrical engineer with major in electronics, with an experience of over 30 years. He tough me a lot and I am trying to pass down some knowledge if you wish to listen. If not - have it your way.
As you may have noticed by now, my qualification is not unlike that of your father. Perhaps it pays listening to a second opinion now and then rather than taking anything he says for granted. I can see him not willing to admit - perhaps not even to himself - that his work concerning proper CPU cooler installation was blunderish at times.
Facebook
Twitter
Instagram