RAM (Random Access Memory), Dynamic Memory

RAM (Random Access Memory)

In older machines, RAM existed as individual chips socketed to the systemboard in banks or rows of nine chips. Each bank held one byte by storing one bit in each chip with the ninth chip holding a parity bit as shown in the figure below. On older PCs, the parity chip was slightly separated from the other eight chips. Parity refers to an error-checking procedure whereby either every byte has an even number of ones, or every byte has an odd number of ones. The use of a parity bit means that every byte occupies nine rather than eight bits.

Dynamic Memory

There are two types of RAM: dynamic RAM (DRAM) and static RAM (SRAM).

Dynamic RAM chips hold data for a very short time; static RAM chips hold data until the power is turned off. Because DRAM is much less expensive than SRAM, most of the RAM on the systemboard is DRAM. DRAM comes in three types: parity, non-parity, or an altogether new method of error checking called ECC (Error Correcting Code) that cannot only detect an error but also correct it.

Parity is a method of testing the integrity of the bits stored in RAM or some secondary medium, or testing the integrity of bits sent over a communications device.

When data is written to RAM, the computer calculates how many ON bits (binary 1) there are in the 8 bits of a byte. If the computer uses odd parity, it makes the ninth or parity bit either a 1 or a 0 to make the number of 1s in the 9 bits odd.

Using even parity, the computer makes the parity bit a 1 or 0 to make the number of 1s in the 9 bits even. Later, when the byte is read back, the computer checks the odd or even state. If the number of bits is not an odd number for odd parity or an even number for even parity, a parity error occurs.

A parity error always causes the system to halt. On the screen, you see the error message “Parity Error 1” or “Parity Error 2” or a similar error message about parity.

Parity Error 1 is a parity error on the systemboard; Parity Error 2 is a parity error on a memory expansion board.

Parity errors can be caused by RAM chips that have become undependable and that are unable to hold data reliably. Sometimes this happens when the chips overheat or power falters.

Computers were made to hold RAM on a group of chips stored in a single physical unit called a SIMM (Single Inline Memory Module).

A SIMM is a miniboard that stores an entire bank or banks of RAM. A SIMM can have several chips with 30 or 72 pins on the edge connector of the tiny board.

RAM is then upgraded or changed by unplugging and plugging in SIMMs, which are much easier to work with than single chips.

RAM chips or SIMMs are located either on the systemboard or on memory expansion cards.

SIMMs hold from 8 MB to 64 MB of RAM on one board. All new systemboards today use DIMMs (Dual Inline Memory Module), which have 168 pins on the edge connector of the board.

A DIMM can hold from 8 MB to 256 MB of RAM on a single board. The figure below shows the two kinds of SIMMs and a DIMM module. Most systemboards today use 168-pin DIMMs.

Memory can be managed using several technologies that involve how memory is accessed, how timing the access is managed, and how the systemboard and the CPU relate to the memory modules. The more prevalent memory technologies (and some variations of each) used by the industry are listed in the following table.

Regardless of the type, dynamic RAM chips do not hold their data very long and must be refreshed about every 3.86 milliseconds.

To refresh RAM means that the computer must rewrite the data to the chip. Refreshing RAM is done by the DMA (Dynamic Memory Access) chip or sometimes by circuitry on the systemboard other than the DMA chip.