For a while, I managed a fleet of about 6500 iPod touch devices, and we would have random failures on devices where a simple re-image would solve the problem. I did a back of the napkin calculation, and our rate of those failures was almost exactly the expected cosmic radiation bitflip chance for the total surface area of all the storage chips in our devices.
Indeed, but smaller structures are less likely to get hit in the first place, it’s a bit of a trade off.
Also, error correcting memory and storage are things, but at the end of the day, that is just more/larger structures, which is a roundabout way of saying you’re correct.
Bitflips are fun.
For a while, I managed a fleet of about 6500 iPod touch devices, and we would have random failures on devices where a simple re-image would solve the problem. I did a back of the napkin calculation, and our rate of those failures was almost exactly the expected cosmic radiation bitflip chance for the total surface area of all the storage chips in our devices.
larger structure sizes can help against spontaneous bit flips, i think. at the cost of consuming more energy for each calculation step.
Indeed, but smaller structures are less likely to get hit in the first place, it’s a bit of a trade off.
Also, error correcting memory and storage are things, but at the end of the day, that is just more/larger structures, which is a roundabout way of saying you’re correct.