From: Will Honea    
      
   Doug Bissett wrote:   
      
   > No, that isn't right. A quick test simply writes all zeros, and reads   
   > it, then it writes all ones, and reads it, end of test, but at that   
   > point everything should be in good parity (the zero and one tests may   
   > be the other way around). An extended test does those two, then does   
   > half a dozen pattern writes, and reads before it is finished. Of   
   > course, different BIOS implementations might do it a little different.   
   > One thing, for sure, is that by the time an operating system gets   
   > booted, all of memory should be in good parity. If it is not, there is   
   > something wrong.   
      
   Generally, I agree with what you say.  There are an additional condition    
   that may be in play here, though.  The two most common ones are marginal    
   bits which may fail upon temperature rise/fall and pattern sensitivities - a    
   sort of noise thing - which typically occur at random times and in random    
   cells.  About the only way, short of a temperature chamber, to catch these    
   is with something like memtest run for several hours.  This will log the    
   intermittent stuff and what is described here should pop up in short order    
   if it is a memory problem.   
      
   --    
   Will Honea   
      
      
   --- Internet Rex 2.31   
    * Origin: TeraNews.com (1:261/20.999)