As for the technical issues of space applications, there are more than just radiation.
The largest one is not really technical but legal - chip qualification. The agency has to verify that the chip is robust enough to be used in a space flight - the easiest way to do it is to use the same chip that was used before. Manufacturers don't do this job easier by disclaiming any responsibility for non-standard applications (medical, military, space etc).
Technical issues are generally related to reliability over extended periods of time. On one hand new processes are inherently less reliable (more common manufacturing defects, doping profiles diffuse away faster, metal connections are more prone to electromigration, transistors - to hot carrier effects), on the other - these issues receive great deal of attention nowadays, with (semi-)automatic verification, testing etc being applied as a part of standard development flows. However, typical modern devices usually are verified (using models extracted from forced aging) for ~10years of continuous operations. Automotive, medical and other special applications go a bit further but they are usually lagging 4 or more process nodes behind (that's the time it takes to refine the process, characterize it from reliability point of view and to develop necessary tools and libraries for designing such chips).
Another issue is complexity. If our application requires a powerful CPU, millions of lines of code and we have no time for testing several chips, we simply shouldn't expect much.
On top of that comes radiation that indeed may cause spurious computation errors or trigger latch-up. It's not an unsolvable problem but one that is quite hard to test and affects so small numbers of applications that it gets very little attention from chip manufacturers. There are circuit and system level techniques to beat it but still we have to solve other issues.
↧