In some respects this is vastly better than a once-a-month fault. AMD have a better chance of reproducing it even if, clock speeds being what they are these days, the are a tonne of instructions churning between events.