I love ECC as much as the next guy, but the reality is that the entry overstates the importance of bit-flips by assuming the bits always flip in something that

I love ECC as much as the next guy, but the reality is that the entry overstates the importance of bit-flips by assuming the bits always flip in something that matters. The simple fact is that most bit flips occur in portions of memory no one cares about. If the error even manifests a lot of the time it'll just manifest as one pixel in some image somewhere changing color by one bit. With the author's estimate of 1,000 bit flips over the lifetime of a computer, maybe 10 of them cause crashes. Most of those crashes are likely to be a web browser on most people's desktops anyways, so if you just imagine you have the previous version of the flash player, you can simulate an increased SEU rate pretty nicely. ECC is standard on servers because we assume the data they carry matters the large majority of the time. We assume that server's typically have a larger portion of their memory devoted to "important" things. (I.E. not images, video or stuff a javascript interpreter forgot to free().) On desktops, it is still probably reasonable to purchase non-ECC hardware for the time being. I agree with the author though, that this is only getting worse. The trends are all in directions where this is going to start affecting consumer-level stuff at some point, but I'm not sure we're there yet. As always, it's a matter of your workload combined with good risk analysis.