Do soft errors still occur in today's machines?

[Muhammad Usman]

TL;DR Summary

Do soft errors still occur in today's machines specially laptops, routers and is it the most probable cause behind the reboot-fix?

Hi,

I was reading about the soft errors. I read that the soft errors are the one that basically caused by alpha particles and even some time thermal issues (Too much heat in the machines). Although I search lot of mitigation techniques but I am curious that are these errors still exist and are they the leading cause of issues that are fixed with the reboot. Thanks

 

[Svein]

Most things that are fixed with reboots are software bugs. That is why I banned reset buttons from motherboards that were to be used in critical applications (find the error - do not act as if nothing happened!).

 

[Baluncore]

The majority of crashes are due to software bugs. Very few crashes are due to soft errors.
If it is important to be soft error free, then you must run the program twice and check the outputs are the same.
Soft errors due to contaminated packaging have been greatly reduced over the last 40 years.
https://en.wikipedia.org/wiki/Soft_error#Designing_around_soft_errors

 

[Rive]

If it is important to be soft error free, then you must run the program twice and check the outputs are the same.

In critical applications (where delay is not acceptable) it's not twice but thrice in paralell, and two similar result is the requirement.
TMR
Guess this might count as existing mitigation technique for soft errors.

In everyday computer usage the end result is mostly the same for software- and soft errors - for a singular event you will never know what happened.

 

[trurle]

Summary:: Do soft errors still occur in today's machines specially laptops, routers and is it the most probable cause behind the reboot-fix?

Hi,

I was reading about the soft errors. I read that the soft errors are the one that basically caused by alpha particles and even some time thermal issues (Too much heat in the machines). Although I search lot of mitigation techniques but I am curious that are these errors still exist and are they the leading cause of issues that are fixed with the reboot. Thanks

Soft errors are mostly occurring in DRAM; other parts of computer have a build-in rejection of low-energy events and should not have soft errors if designed properly.

In terrestrial applications, the soft error rates are negligible for most DRAM chips manufactured after 2010. Before 2010 it was higher because some DRAM makers (i cannot say who due to my work contract restrictions) did used a fault-prone variant of deep trench technology. The occurrence of faulty technology was common enough to make popular the ECC and parity-checked DRAM modules.

In space application radiation doses are much higher (0.3-10 Sieverts/year) compared to terrestrial (~0.002 Sieverts/year), therefore soft errors in form of bit-flip and SEU are still common in satellites, despite of less sensitive modern DRAM.