PDA

View Full Version : Heating problems explained


albatros
03-13-2006, 01:22 AM
I know that in a computer excessive heat degrades performance.
But how that happens technical speaking.
I mean what are the physical processes involved.

For ex.does heat affects the flow of electrons?

Thank you

Mini-Me
03-13-2006, 01:46 AM
Generally speaking, electrons don't care that much about heat per-se'.
But the substrate or die on which the chips are formed have a so-called critical temperature, and when this is reached and/or exceeded, then the chip crystal or die begins to break-down chemically. Once this happens, the microscopic pathways and etchings on the die become breached, and the chip electrically fails. This is a catastrophic failure, and not repairable.
Also, it is worth noting that the microscopic crystals which form the individual transistor junctions on the die, also have their own critical temperature, and if this is exceeded, you may not have a failure of the die, but of the components etched on the die - either way, it's curtains for the chip...

As my mentor taught me: "It is the smoke that keeps the device working. Once the smoke gets out, the device won't work, and it's very hard to get the smoke back in!"
;)

Interestingly, most descrete electronics such as transistors and regulators can be quite happy working at horribly high temperatures approaching 100'C, yet most computer chips, especially CPU die's, get very upset with you, if allowed to run hotter then about 75'C...

Lets say that the critical die temperature of a certain CPU is 70'C, and you run it at 75'C, then it could either fail totally, or not actually fail, but slowly "Melt" for want of a better description, like an ice-cream on a hot day, and this in effect permantly damages the CPU die, and makes the CPU useless, or extremely unstable in use - even if you let it cool, or run it cooler from then on. Once the critical die temperature is exceeded, then the device is generally damaged in some fashion.
:p

I used to see quite a lot of catistrophic failures in car-stereos, as they have very limited heatsinking, and it's all passive(no fan).
Generally, you'd pull the stereo apart, and find that there was a nice black hole physically blown through the case of the power-amplifier IC.

This is generally caused by a breakdown of the IC die, and if the die melts in the way mentioned above, it can easily short out the main power-supply feeding it, causing a direct short across the supply voltage. This literally blows the chip package apart due to a mixture of current and excessive heat.

...looks impressive, though!!!
:D


MM.

Sylvander
03-13-2006, 07:03 AM
There is very commonly a lack of knowledge and understanding regarding the term "heat".
I too misunderstood until I took an Engineering course that included Physics and Thermodynamics.

What's really important here is not "heat" but "temperature".

HEAT...
Is the name given to thermal energy whilst it is in the process of TRANSFER [from one region to another; normally involving matter, therefore from one region material to another] DUE TO A DIFFERENCE IN TEMPERATURE.
It's the temperature that provides the motive force for transfer; rather like electricity flows as a result of an electro-motive force [difference in EMF].
It is only given the name "Heat" whilst on the move; rather like water which is on the move [falling] from a cloud to the earth beneath is named "rain".

TEMPERATURE...
Think of it as an indirect measure of the vigour with which the molecules of a material are oscillating back and forth.
On a microscopic scale, molecules would only be completely at rest at "Absolute Zero" temperature.
As the temperature rises above absolute zero, the molecules begin to oscillate and if the temperature is measured that fact is shown indirectly [the oscillations would be transferred to the mercury in a thermometer and the mercury therefore occupies more volume per unit mass, therefore the free surface creeps up the fine glass internal bore].
The changing vigour of the oscillation of the molecules of a material produce lots of different and interesting changes in the material.
It might change "phase" between solid, liquid, and gas, or even be a mixture of 2 or 3 phases.
Generally...
Increasing oscillations result in more rapid interactions [chemical processes happen more rapidly], and more chaotic or unpredictable behaviour [resistance to the flow of electricity (electrons moving from atom to atom) increases].
Now, I have no idea what happens to the crystals of a silicon chip as the TEMPERATURE increases [molecular oscillations increase], but I'd guess it isn't good.
I can imagine the crystalline structure changing, and not for the better [my education in metallurgy comes to mind here].

jcnoernberg
03-13-2006, 09:01 AM
check out http://en.wikipedia.org/wiki/Electrical_resistivity#Temperature_dependence

the temp. has a lot of effect on the resistance of the metal wiring. resistance is directly related to the current as well.

saphalline
03-13-2006, 01:27 PM
The mathematical and scientific explanations for why CPU's and other chips and IC's fail at high temps is incredibly complex. In addition to what has been mentioned, quantum physics also plays a big role at the microscopic scale of current lithographic processes. Further complications arise from various transistor-level enhancements such as SOI and extra layers. Also keep in mind that a 3GHz CPU has individual transistors that operate much faster than just 3 billion state changes per second, since multiple transistors comprise each state in the pipeline.

In simple terms, think of a CPU structure as a chain. If any one part melts due to a high die temp, it fails. Just like the old saying "a chain is only as strong as its weakest link". Once a single transistor or pathway melts and breaks, the whole CPU is done for.