Do gamma rays from the sun really flip bits every once in a while?

Question

Possible Duplicate:
Cosmic Rays: what is the probability they will affect a program?

Is this just a tongue in cheek expression or is this really true, and if so, what precautions should we take in software (or these precautions hardware only)?

Solution 1

Well, I did dig up this paper, which claims that your RAM will get bit alterations from "Atmospheric Neutrons" (aka: Cosmic Rays) at a rate of about 1.3*10^-12 /bit/hour.

An article by Berke Durak uses that to calculate that your probablilty of having at least one bit error in 4 gigabytes of memory at sea level on planet Earth in 72 hours is over 95%. Of course that assumes you are using non-error-correcting memory (non-ECC). With ECC, he figured, you can wait 2.7 million years before you get an uncorrectable bit error at a probability of 96%.

Solution 2

Alpha particles, not gamma rays, but yes.

Solution 3

Gamma rays from space can't get through the atmosphere -- that's why gamma ray astronomy has to be done using satellites. You're probably thinking of cosmic rays -- high energy charged particles, rather than photons -- and yes, they can cause bit flips and other such effects due to the ionization that can occur when they interact with the atoms in semiconductors.

Comments

[Paul McMillan]

Yep. This is one of the reasons we pay for ECC in servers.

[Peter Mortensen]

According to that source, those alpha particles come from the packaging material of the integrated circuits. So isn't the answer to the OPs question no, then?

[zwol]

lambda-diode.com/opinion/ecc-memory-2 also has a lot of good information and links to studies. Regarding just what sort of cosmic radiation causes problems: Alpha particles from space won't get through the computer's case (they can be stopped by a sheet of paper, even) but neutrons can and do. But alpha particles from radioactive contaminants in the chips cause more bit-flips than neutrons from space, which is why some DIMMs are much more prone to these errors than others (they got unlucky and have more radioactive contaminants)

[Russell Borogove]

@Peter - in the early days of ICs, I think certain ceramic packages generated their own alpha particles (trace amounts of unstable rare earth elements?) but now atmospheric cosmic rays are the problem. I guess the sun isn't the origin for most of the high energy alphas after all; I was misremembering. en.wikipedia.org/wiki/Cosmic_rays

[Admin]

Actually, I didn't know the difference between Gamma and Cosmic Rays. Thanks for clarifying.

[Peter Mortensen]

Atmospheric neutrons are not the same as cosmic rays. Cosmic rays are high energy charged particles (mostly protons). The atmospheric neutrons are some indirect consequence of those particles hitting the atmosphere (I would be interested to know what the mechanisms are for producing neutrons from the protons or helium nuclei). It is an interesting article you have dug up.

[Slater Victoroff]

Thank you for pointing out that there is a difference, but I am still sad to see that your explanation of the difference between the two is lacking. Photons are particles, and the effects have nothing to do with ionization and cosmic rays are usually protons which happen to be charged particles, but that has nothing to do with being a cosmic rays, and the interactions are at a lower level than entire atoms, and also cosmic rays only exist at impact with the top of the atmosphere and we only ever encounter the daughter particles down here, and countless other things, but still thank you.

[user984260]

@WinMain There are many software techniques for quantifying and removing soft-errors, see my survey paper academia.edu/12046032/…

[lucaswxp]

@PaulMcMillan What does "ECC" stand for? Is that standard?