In gate array designs, the silicon vendor offers a range of chip sizes. Each size of chip has a fixed layout and the location of each transistor, resistor and IO pad is common to every design that uses that size. Gate arrays are configured for a particular design by wiring up the transistors, gates and other components in the desired way. Many cells will be unused. For mask-programmed devices, the wiring up was done with the top two or three layers of metal wiring. Therefore only two or three custom masks were needed be made to make a new design. In FPGAs the programming is purely electronic (RAM cells control pass transistors).
The disadvantage of gate arrays is their intrinsic low density of active silicon.
Standard cell designs use a set of well-proven logic cells on the chip, much in the way that previous generations of standard logic have been used as board-level products, such as Texas Instruments' System 74. A variation on the gate array is to include full-custom macrocells such as processor cores in fixed positions on the die. About 25 to 40 percent of chip sale revenue now comes from field programmable logic devices. These are chips which can be programmed electronically on the user's site to provide the desired function. The Xilinx FPGA parts used in the Part 1B E+A classes are one of the most important examples of field-programmable logic.
Field-programmable devices may be volatile (need programming every time after power up), reprogrammable or one-time programmable. This depends on how the programming information is stored inside the devices, which can be in RAM cells or in any of the ways used for ROM, such as electrostatic charge storage (e.g. FLASH). Except for niche applications FPGAs are now always used instead of masked gate arrays.