|Reference||STATE OF THE ART||Stan Augarten|
|The Microprocessor Decentralized|
A Bit-Slice Processor, the 2901
ADVANCED MICRO DEVICES
Microprocessors are invariably built out of metal-oxide-semiconductor (MOS) transistors, which were invented by engineers at RCA in 1962 (p. 12). Compared to the bipolar, or junction, transistor, the other basic form (pp. viii-ix), the MOS transistor takes up relatively little space and consumes much less power. Today's highly sophisticated microprocessors, containing hundreds of thousands of elements on the same chip, are made possible in part by the miniscule dimensions of the MOS transistor.
But MOS transistors are rather slow, which makes them far from ideal for high-speed computers. Bipolar transistors, on the other hand, are as fast a greased lightning. A typical 4-bit microprocessor might require ten millionths of a second to perform a single operation, whereas a 4-bit bipolar chip like the 3901 shown on the right needs a mere hundred billionths of a second, making it some one thousand times faster.
But it's uneconomical to fashion a microprocessor out of bipolar transistors. Because bipolar elements require a lot of power, they give off a great deal of heat, thereby necessitating larger chips, and production yields of working ICs plummet as chip dimensions rise. So engineers, looking for a way around the impasse, developed bit-slice processors, composed of bipolar transistors, each of which can carry out a portion of a microprocessor's chores. Grouped, or cascaded, together, several bit-slice ICs can be made to function like a single microprocessor, but working much faster and generating little heat because they have relatively few components.
A bit-slice system is like a water tower with several descending pipes. The amount of water dispatched through the pipes is controlled by a central valve; each pipe is the same size and leads to the same giant vat at the tower's base. In a bit-slice system, it's electricity, not water, that's shunted about, but the principle is the same. A separate control chip functions in lieu of the water tower's valve, routing electrical signals to the bit-slice chips, which act as pipes endowed with mathematical processing ability; the control chip keeps track of the results.
Bit-slice ICs typically come in 4-bit denominations. Hence, functioning under the supervision of a control chip, two bit-slice chips may carry out operations on 8-bit numbers. Likewise, three bit-slices can manipulate 12-bit numbers, and so on, to numbers as large as users desire. Bit-slice chips are most frequently used in large and fast computers, where they are often strung together to make 32-bit operating systems; large computers are usually 32-bit machines.
|The 2901 is pictured here without its upper layer of aluminum interconnections, which would have given it an entirely different appearance. The 2901 is the most widely used bit-slice processor. Actual size: 0.124 x 0.130 inches|
|STATE OF THE ART
©Copyright Stan Augarten
|This book is provided for general reference. The National Museum of American History and the Smithsonian Institution make no claims as to the accuracy or completeness of this work.|
|page:||Index 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76|