Dynamic Random Access Memory (or DRAM) stores information using one transistor and one capacitor for each bit of information. This leads to reduced cost and improved density, compared to Static RAM, since each bit of data can be stored with less hardware.
A problem with using capacitors to store data is that the electrical charge they store fades with time. To keep from losing information, the bits need to be refreshed periodically. The term “dynamic” is used to differentiate this type of RAM from Static Random Access Memory (SRAM), which will retain information (without the need for capacitor refresh) for as long as power is applied. (There is an implementation of DRAM that includes built-in refresh circuitry to imitate the simplicity of Static RAM, which is called PSRAM).
DRAM is also available in ECC form, which has extra Error Correcting Code (ECC) for detecting when hardware errors have corrupted data. When possible, the ECC DRAM unit will repair the corrupted data automatically. ECC DRAM is mainly used in server systems where losing a single bit of data could be disastrous (banking or medical systems, for example). Hardware memory errors are caused by high-energy radiation from outer space, and can generally be expected to affect one bit of data per month, per gigabyte of DRAM. Compared to the amount of data loss from viruses, worms, and shoddy software, the average PC user probably won’t notice the damage from a stray cosmic ray every now and then. If your life depends on it, however, it’s nice to know there’s a way to detect and repair these errors.