Memory Failures

Because of the massive number of memory chips that hardware manufacturers produce and the high standards customers have for reliability, memory testing is a highly refined science. Different memory tests are designed to detect specific types of common failures, including the following:

• A bit may always return 1, even if set to 0. Similarly, a bit may always return 0, even if set to 1. This is known as a Stuck-At Fault (SAF).
• The wrong bit is addressed when attempting to read or write a specific bit. This is known as an Address Decoder Fault (AF).
• A section of memory may not allow values to change. This is known as a Transition Fault (TF).
• A section of memory changes when being read. This is called a Read Disturb Fault (RDF).
• One or more bits lose their contents after a period of time. This is known as a Retention Fault (RF) and can be one of the more challenging types of failures to detect.
• A change to one bit affects another bit. This is known as a Coupling Fault (CF) if the faulty bit changes to the same value as the modified bit, an Inversion Coupling Fault (CFin) if the faulty bit changes to the opposite value as the modified bit, or an Idempotent Coupling Fault (CFid) if the faulty bit always becomes a certain value (1 or 0) after any transition in the modified bit. This behavior can also occur because of a short between two cells, known as a Bridging Fault (BF).

Given these types of failures, it's clear that no single test could properly diagnose all the problems. For example, a test that wrote all 1s to memory and then verified that the memory returned all 1s would properly diagnose an SAF fault in which memory is stuck at 0. However, it would fail to diagnose an SAF fault in which memory is stuck at 1, and it would not be complex enough to find many BFs or CFs. Therefore, to properly diagnose all types of memory failures, Windows Memory Diagnostics provides several different types of tests.