(First posted on comp.hardware, but I received no response.)
I have a pile of IDE hard disks (1 to 4GB) at work that I need to run through some reliability testing to make reasonably sure they won't die or show intermittent problems while running. These hdds will go into computers primarily used for front-office banking, hence I am very concerned about reliability. (Added to this are the embarrassment and waste factors, in showing up at a banking center with a replacement that either dies immediately or very soon.)
I downloaded COREHDT for the testing, and it ran read/write tests on one particular 1.08GB drive. The test showed 65 errors noted in 1e+12 bits. (The errors were all in the "Read-MultiSector" and "Read-64KB" test modes.)
1. How many bits of testing are the minimum for measuring reliability in this way? I recall reading specs that commonly give error rates in the "x per 1e+12" (1 trillion bits) range. For running COREHDT on a 486, it takes about 24hr to run through 1e+12 bits.
2. How much error is tolerable? For instance, 4 errors per 1e+12 bits?
I've tested half a dozen drives this way, and others have 0 errors.
I have a pile of IDE hard disks (1 to 4GB) at work that I need to run through some reliability testing to make reasonably sure they won't die or show intermittent problems while running. These hdds will go into computers primarily used for front-office banking, hence I am very concerned about reliability. (Added to this are the embarrassment and waste factors, in showing up at a banking center with a replacement that either dies immediately or very soon.)
I downloaded COREHDT for the testing, and it ran read/write tests on one particular 1.08GB drive. The test showed 65 errors noted in 1e+12 bits. (The errors were all in the "Read-MultiSector" and "Read-64KB" test modes.)
1. How many bits of testing are the minimum for measuring reliability in this way? I recall reading specs that commonly give error rates in the "x per 1e+12" (1 trillion bits) range. For running COREHDT on a 486, it takes about 24hr to run through 1e+12 bits.
2. How much error is tolerable? For instance, 4 errors per 1e+12 bits?
I've tested half a dozen drives this way, and others have 0 errors.