Parity Errors

 Subject: What hardware parity errors are, and some suggestions on
          finding and correcting the problem.

 Q. What is a parity error?

 A. The memory controller chip on your PC reports a parity error
    when it reads a byte of data and the 9 bits it used to encode
    the byte do not add up to 1 (odd parity).  Parity errors are
    always hardware-related.  Software applications cannot cause
    parity errors, although an application may cause one to be
    detected.

    In the digital world, all information is represented by the
    binary numbers 0 and 1.  The binary digit, or bit, is the
    fundamental building block of digital information in a
    computer, and it stores information in two states:  off or on
    (0 or 1, respectively).  One bit can make a big difference.
    Here's why:

    The binary number for the letter U is:

    01010101

    If you change just the fourth bit over from the left, from one
    state to the other, the binary number becomes the letter E.

    01000101

    Now while there are 8 bits in a byte, your memory controller
    handles information 9 bits at a time.  This extra bit is called
    a "parity bit", and is the computer's way to verify the
    integrity of your data. Whenever you write data to memory, the
    memory controller adds up the number of 1's in each byte of
    information, and then sets the ninth bit to make the sum of all
    nine bytes odd. IBM, the original designers of the PC, could
    have chosen to make the sum of the nine bytes even (even
    parity), but they chose to store data in memory with odd parity
    and every other PC manufacturer followed suit.

    In the example above, the letter U has the binary value of
    01010101, which has 4 1s in it, and the letter E is 01000101,
    which has 3 1s in it.  When your PC reads each byte of data, it
    sums the 9 bits to make sure the number of 1s in the byte is
    still odd.  If the state of a single bit gets changed from 1 to
    0, or 0 to 1, the parity of the nine bits becomes even and the
    memory controller asserts the NMI (Non-Maskable Interrupt).
    This signal is put directly on a pin of the CPU, then the code
    pointed at by Interrupt 2 posts a Parity Error message, which
    warns you that there is a problem with your RAM.


 Q. How do I determine which piece of hardware is causing the
    problem?

 A. First, check is the RAM in your system.  An easy test is to
    disable everything that uses EMS and XMS memory so you can
    create a RAMDRIVE the size of all your system memory.  (Refer
    to your DOS manual for information on creating a RAMDRIVE.)

    Then:

    a) Run CHKDSK on the RAMDRIVE, or
    b) Copy files to the RAMDRIVE until it is full.

    Either way, if you have bad memory on your system, eventually
    you will get a parity error or a General Drive Failure on the
    RAMDRIVE.

    The first thing you can do to try to remedy this problem is to
    make sure that the RAM chips are seated properly in their
    sockets.  If they are DRAMs or SIPPs, make sure the pins aren't
    broken off or bent.  If they are SIMMs or the memory is on a
    card, you may just need to clean the contacts.  If the chips
    physically check out ok, the chip speeds could be mismatched
    with memory that is too slow for the CPU/memory bus, or a
    controller chip could be bad.  At this point the only sure way
    to test this is to swap out the chips for ones that you know
    are good.

    Parity errors may also be caused by the presence of an
    autoswitching video card or one that is using 16-bit ROM
    access. Your motherboard could be assigning parity to the
    address space where your EMS page frame is located.  Also there
    may be some special features of the computer in the CMOS Setup
    that could be causing problems.  Try disabling the computer's
    shadowing of BIOS or video ROM or turning off memory caching or
    other features to see if one of them is involved.  This may
    allow you to pinpoint the cause of the problem. In all these
    cases you should refer to the documentation that came with your
    hardware product to disable a particular feature.

 Q. I ran a hardware diagnostic program on my machine, and it
    didn't report bad memory.  Why not?

 A. While there are several diagnostic programs on the market that
    will test your memory for errors, they may not duplicate
    conditions that would cause marginal memory to fail.  Most are
    not even designed to be run with a memory manager.  When parity
    errors are encountered, it is time to have the hardware
    components of the machine examined.

 Q. Is there any software I can use to get around this problem?

 A. No.  Note that all of the parity operations are performed
    directly by your computer's hardware, regardless of which
    operating system (DOS, OS/2, UNIX) you use and regardless of
    which utility programs or application software you run.  One
    exception is Macintosh computers, which use 8 bit SIMM chips
    that do not have parity. When errors occur, the system just
    malfunctions from the invalid data.  Also remember that parity
    checking will only detect if one bit in a byte gets changed. If
    two bits in the same byte get changed it will accurately
    reflect that the sum is still odd and errors will not be
    detected.