A reason for ECC

Why buy ECC RAM?  This is a discussion I’ve seen many times.  I’ve always heard that without the error checking, you can’t tell if a random bit was flipped by a cosmic particle.  That seems like a very remote threat.  Over the last week, I went to Science North in Sudbury, Canada, and saw the Diffusion Cloud Chamber.  I took a photo myself.  Both of those picture represent an instant in time, and each of those squiggles in the chamber in that instant represents some particle zipping through space that miiiiiight scramble your RAM.  That’s… a lot more common than I thought.

4 Replies to “A reason for ECC”

  1. How does HAMMER respond to flipped bits on a system without ECC RAM? Is a flipped bit possible to detected as a change in the file system?

  2. This would be affecting RAM, not disk, so it’s not affecting the file system as much as it’s affecting memory operation. Hammer does apply checksums when moving data, so it could be noticed at that point. I’d expect it to cause errors at a level before it hits any file system, as a vague guess.

  3. Data corrupted in non-ECC RAM that has yet to be pushed to disk (or SSD, w/e) will still be corrupted when it does get to the filesystem.

  4. Faulty memory non-ECC memory can definitely get on the disk (leading to CRC errors later on). I’ve had such a case. Same goes for non-correctable errors when using ECC memory.

Comments are closed.