For a long time, there has been speculation about actual error rates, especially for DRAMs. One of the most common assumptions was that errors in DRAM were relatively uncommon. A couple research studies involving DRAM modules, or DIMMs, have been conducted and published that seem to help explain what actually is going on. In a study of Google servers, failure rates of DIMMs occurred surprisingly often. It was found that there was an average of nearly 4000 correctable errors per DIMM per year, and that more than 8 % of all the DIMMs had correctable errors. At a server farm at Oak Ridge National Laboratories, there was a DRAM fault every six to seven hours. 70 % of the errors were recurring faults. 47 % of the errors were single bit errors and 22.5% of the errors are from a single row or column. Again, both these studies are of servers, which implement many DRAMs in a module form factor. For high-quality, embedded applications that don’t use modules, rather have a small number of DRAMs directly on the PCB, error rates are not clear, but an engineer might expect fewer errors. Yet, consequences of errors in these applications is often of a high concern.