-------- Forwarded Message -------- > From: Thayne Harbaugh> Reply-To: ecc@ya... > To: ecc@ya... > Subject: Re: [ecc] Can anyone help with testing tools? > Date: 30 Oct 2003 07:24:15 -0700 > On Thu, 2003-10-30 at 05:50, sailor1357 wrote: > > Is there any tool available for testing ECC? Can we find a tool that > > can emulate ECC hardware error? Just wrote something for chipset > > e8870. I am wondering how testing is done for our existing chipsets. > > There are a few good ways to check functionality. > > 1) Gently apply heat to the DIMMs with a heat lamp or heat gun. > > 2) Find some bad memory, label it, and hang on to it for testing > purposes. > > 3) Stress the memory with something like Linpack (do this in conjunction > with the other methods). > > 4) Cut a thin strip off a silver, anti-static component bag.(ps. adhesive > tape is also useful here.) Powerdown, pull up DIMM,insert it in the socket > with the DIMM so that some of the pins don"t make contact. Then power on. > It"s best to go to http://jedec.org and lookup the DIMM > specification to know which pins to mask. I believe pins 12 and 13 are > data pins and are convenient to count to and mark. Obviously you only > want to mask one pin if the machine only has SECDED capability. > > Here"s some more information on how to do that: > > +------------------------------- > | \ > | -------+ > | Masking Strip | > | -------+ > | / > +------------------------------- > > The narrow part ends up being ~2-3mm wide, the wide part can be whatever > is convenient. This way the strip is wide enough to hold on to and > maneuver, yet is narrow enough to mask pins. > > Here"s an ugly representation of a DIMM: > > +---------------------------------------+ > | | > | DIMM Front | > | | > | +-+ | > |00000000000000000000000| |0000000000000| > +-----------------------+ +-------------+ > ^ ^ > | | > +-- Pin #1 Pin #92 --+ > > Pin 12 is DQ8 > Pin 13 is DQ9 > > I use these because it"s the first pair of adjacent pins. If I"m trying > to mask one pin then it gives me some margin for error. If I"m trying > to mask two pins then they are adjacent and it can be done with a single > strip. > > The nice thing about this method is that you can instantly have a bad > DIMM and you can control where the bits are bad which can be verified > with the syndrome and row decoding. The drawback is that it means that > there"s *always* an error and it isn"t very random. > -- Thayne Harbaugh