Hi folks. Has anyone run into single point failures where only one byte in a RAM stick fails, and only then under very specific conditions?
It might be radioactive encapsulant!
There was an article not so long ago in Electronics World (tm) about certain manufacturers running into issues with a change in material supplier, as RAM chips are particularly sensitive in this way. Now if memory serves the issue was somewhat resolved using advanced screening and better encapsulant.
Its emerged again with GDDR6, as the feature size is so much smaller even a few electrons out of place can be very bad.
To some extent the more expen$ive cards mitigate by doing at least basic inline correction and/or running at a lower clock rate.
In this case the risk to the end user is near zero, the sort of radioactivity that can affect these is below the threshold of most sensors (α, β) and in fact barely detectable. My sensors only ever saw one case of this and it was 60Co contaminated steel on a USB port, independently verified by a third party with better equipment.