Memory error
14 Jun 2007 20:14:29 UTC
Topic 192847
(moderation:
I've found one of the two 512 MB memory modules in my computer to be defective (memtest reports the error after running for just a few seconds). I suspect this error having been there for about a month (as since then I got strange results on my selfmade programs now and then); however, the E@H results of that computer were all reported as being verified successfully (at least those I remember having seen).
Does verification find such errors caused by hardware defects or is there a realistic chance of errors not being caught?
Memory error
)
You don't have to worry, the validator is very strict. It works in two passes: first the resultfile is checked for "syntax" and ranges for the individual items in the file. The more crucial test happens in a second stage when your result is compared to that of another host. the validator allows for some minor "jitter" in the (floating point) values computed by the two hosts, but any grave error that would be caused by a memory fault would be detected.
Einstein@Home doesn't use that much memory after all, and the bulk of the memory accesses will happen in the CPU's cache anyway most of the time. So it's not that surprising that E@H wasn't affected by that memory problem.
CU
BRM