>I'm curious on how you harden a cpu??? Redundancy inside the
>chip or sumthing?
Can be done in many ways or a combination of them. Now note my limited knowledge on the subject is related to image sensors, and the idea there is not to avoid such elements from causing a glitch (on the contrary, quite often the idea is indeed to detect those elements, and make 'pictures' of them), but rather to ensure the sensor keeps working even when being bombarded by high energy neutrons or alpha particles over many years.
For cpu's here on earth, I doubt the problem is that these chips would get killed by this, the issue is rather avoiding bitflips/errors/glitches when they do get hit by such an element.
I'll try and answer the question though; building in redundancy is indeed the most important thing. This is already done to some extend: cache is nowadays always protected by ECC, mostly for this very reason. (BTW, Suns UltraSparc II did not have ECC protected CPU cache, and did exhibit abnormal high crash rates, again, almost certainly because of this, so its not unreal).
There are also circuit engineering techniques that reduce sensitivity to radiation, but I am unsure how many of these are applicable for CPU's, as some of them would directly affect clockablilty I think. For image sensors, using higher voltages, higher capacitance helps. Selective reworking of ciritical paths by increasing the size of the 'wires'. None of these seem like good idea's for a high end cpu, so I guess its not strange 486's and the like are still used for aerospace applications.
Doing some googling on the subject I also found out that SOI would help (well, certainly NOT for cmos image sensors LOL), certain forms of doping of the silicon, and even special solder with lower radioactive isotopes.
The main thing however, is making your critical paths wider than necessary and having higher voltages (bigger difference between 1 and O). An 840EE or X2 is therefore not likely a spaceworthy cpu
😀
>Still does it have a place in computers meant for us
>ordinary people?
Hard to say. I would guess so. As process nodes shrink further, transistors shrink, voltages keep dropping and frequencies go up and transistor count goes through the roof, the sensitivity of these chips for radiation would also grow exponentially. I have seen a study where they tested Pentiums for their space worthyness, and the Pentium II was already 10x as critical as the Pentium I for gamma particles.
So yes, I think this is becoming a concern although I would worry far more about this happening to RAM modules than to the CPU.
>As you also said It would be nice to know of how much
>influence eg: cosmic particles has on pc's generally. (I
>hope that I can explain 99 / 100 crashes on my pc...)
If you where running zOS or Nonstop kernels, maybe.. but with windows and 200 third party drivers,.. nah, I don't think they account for 99% of the crashes yet
= The views stated herein are my personal views, and not necessarily the views of my wife. =
Facebook
Twitter
Instagram