Cosmic Rays a problem for PC reliability in near future?

[Dav333]

Is it to become a major problem as transitors get smaller & RAM amount goes up?

A blog on NS said "This problem is projected to become a major limiter of computer reliability in the next decade." Will it be like moore's law with transistors size being a limitation?

My comp has 4GB RAM & has never crashed on its own, most times its from 2 programs conflicting with each other or similar.

 

[Pattonias]

Sometimes I wonder if cosmic rays are what's causing my computer to malfunction. Surely it isn't my constant upgrading and sometimes poor wire management.

 

[Naty1]

That seems a somewhat exaggerted concern. (My guess, I tried to confirm as explained below.) I would think the grounded metal case surrounding a typical computer mother board would provide pretty good protection...

But increases in sunspot activity, one source of typically lower energy cosmic rays, can cause issues with satellite communications and even electrical power transmission lines.

I'd be interested in any theory underlying semiconductor devices becoming more prone to cosmic rays.

I checked Wikipedia here:
http://en.wikipedia.org/wiki/Noise_(electronics )

and interestingly it did NOT mention cosmic rays as a possibly noise source...

and here: http://en.wikipedia.org/wiki/Solar_cycle#Effects_on_Earth

does NOT mention solar flares (cosmic rays, etc) as a source of individual solid state performance degredation...

 

[lostinxlation]

It definitely causes soft errors on the data, but it won't cause a massive error.
The material's atomic number must be high to interact with photons(THat's the reason why Pb is used for shielding the gamma/X-ray, as well as Pb is radioactively very stable).
The silion's atomic number is only 14 and the chance that a photon can cause multiple errors isn't very high. If it's just one or two bit error per word data, there are many coding theories which can fix the error bits.

 

[D H]

The problem is not noise, and the problem is not caused by protons. Energetic cosmic rays don't reach the Earth (usually). Instead they collide with molecules in the atmosphere, and the resultant particles from those collisions collide again. What reaches the Earth is a rain of neutrons. Those neutrons can be absorbed by the chip or by its surrounds. The neutron itself doesn't do damage. The damage results from the absorbed neutron making the atom into an unstable isotope. The alpha particles from the decay of that unstable isotope can do damage to the chip. It might flip a bit in memory (single event upset, aka soft error), it might cause a short circuit (single event latchup), or it might do something more serious (single event burnout).

People in the avionics industry have worried about single event effects for quite some time. Radiation is a big part of the reason that computers for aircraft and space vehicles are a decade or two behind the state of art -- and the gap is growing. State of the art in rad hardened processors are equivalents of the MIPS R3000 and the SPARC V8 (introduced for terrestrial use in 1988 and 1990).

Intel thinks the problem is serious right here, not just up in space. US patent 7,309,866, assigned to Intel, is for "Cosmic ray detectors for integrated circuit chips". From the background art description of the patent,

The normal background radiation environment on the surface of the Earth has ionizing components that sometimes affects the reliability of semiconductor integrated circuit chips, such as memory chips used in computers. If an intruding particle is near a p-n junction in the chip, it may induce a soft error, or single-event upset which can cause signals to change voltage and, accordingly, bits of data to change voltage value. Excess electron-hole pairs may be generated in the wake of the penetrating particle. The field in the neighborhood of the p-n junction, if sufficiently strong, separates these electrons and holes before they recombine, and sweeps the excess carriers of the appropriate sign to a nearby device contact. A random signal may be registered if this collected charge exceeds a critical threshold value.

Cosmic particles in the form of neutrons or protons can collide randomly with silicon nuclei in the chip and fragment some of them, producing alpha-particles and other secondary particles, including the recoiling nucleus. These can travel in all directions with energies which can be quite high (though of course less than the incoming nucleon energy). Alpha-particle tracks so produced can sometimes extend a hundred microns through the silicon. The track of an ionizing particle may extend a fraction of a micron to many microns through the chip volume of interest, generating in its wake electron-hole pairs at a rate of one pair per 3.6-eV (electronvolts) loss of energy. A typical track might represent a million pairs of holes and electron.

Cosmic ray induced computer crashes have occurred and are expected to increase with frequency as devices (for example, transistors) decrease in size in chips. This problem is projected to become a major limiter of computer reliability in the next decade.​

Source: http://patft.uspto.gov/netacgi/nph-...7,309,866.PN.&OS=PN/7,309,866&RS=PN/7,309,866

 

[mgb_phys]

Typical cosmic ray rates on a CCD at sea level are around 1/cm^2/min, a bit higher at altitude.

CCDs have biased junctions which are more sensitive to cosmic rays than memory (they are after all designed to detect light) but:

A cosmic ray hit might be 100um long on a ccd which would be a 1000 memory cells on modern DRAM.
It's only a function of the area of the memory, so high DRAM densities mean less hits overall, but a hit might affect a larger area.
The CCD is inside a 1" thick wall vacuum dewar, so the thin steel box of a PC isn't going to do anything.
The main source of bad bits in a PC is probably radioactive decay from elements in the material making up the PC.

 

[Naty1]

Post #4:

The material's atomic number must be high to interact with photons

Cosmic rays are not generally photons...but PROTONS...

Cosmic rays are energetic particles originating from outer space that impinge on Earth's atmosphere. Almost 90% of all the incoming cosmic ray particles are simple protons, with nearly 10% being helium nuclei (alpha particles), and slightly under 1% are heavier elements, electrons (beta particles), or gamma ray photons.

http://en.wikipedia.org/wiki/Cosmic_rays

DH interesting post, one thing confuses me:

You say

The problem is not noise,

and also

A random signal may be registered if this collected charge exceeds a critical threshold value.

That sounds like a direct contradiction...can you explain? Thanks

 

[mgb_phys]

That sounds like a direct contradiction...can you explain? Thanks

Servers use error corrected memory.
An extra bit of memory in each block stores the numbers of 1s (or 0s) in that block, when it's read it checks this checksum, if the number of 1s doesn't match then the bad bit can be detected and corrected

 

[waht]

Just to follow up on mgb_phys's CCD camera comment, here is a nice youtube clip on how to make a webcam radiation detector:

Pretty neat.

 

[Pattonias]

Servers use error corrected memory.
An extra bit of memory in each block stores the numbers of 1s (or 0s) in that block, when it's read it checks this checksum, if the number of 1s doesn't match then the bad bit can be detected and corrected

Cool, I don't guess anyone had ever explained that, and I always secretly wondered how it worked.

 

[PaulS1950]

We probably have more to be concerned about the effects of flares and intense cosmic rays than with the actual occurance.
These phenomenon that are a part of space weather can cause huge power surges and outages. For most people something as common as a lightning strike near their home may cause a signifigant spike that could kill most of our electronics - even if we have surge suppressors connecting our equipment to the line. Amajor solar event can cause greater problems over a much wider area.