Christian Schulz wrote: >On Tue, Jun 24, 2008 at 07:22:34AM +0000, Brian Beesley wrote: > > > >>Lots of them. Consumer hardware is not very reliable at all, and it >>only takes one bit flip during the execution of the test to ruin the >>result. The miracle is that the failure rate is so small. >> >> > >Suppose there is a double-check on an exponent yielding a difference >w.r.t. the "original" result - is there a feature that notifies one >(in case the account is still valid)? I'd be interesting in getting to >know wether "my" results prove reliable, but cross-checking by hand >seems annoying... > >Christian > > By far the biggest factor upping the error rate is the 'overclocking' of computers. Overclocking should be avoided at all costs when doing calculations such as LL testing. See the enormous peak in errors in the 33,000,000 to 34,000,000 range in another thread. Virtually all of them caused by overclockers.
If you want to know the reliability of your machine you could check the error codes in the results file. That is a direct and good indication of the reliability of your machine. Yours on the Net, Henk. _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
