announced in 1958, this was one of the first computers on the market to be almost entirely solid-state. it was a bi-quinary decimal computer with a word length of 10 digits, 60 words of core in the base model, and a 5,000 word drum. it required a power supply of 14.4 kilowatts, or 10 kilowatts without the printer, per UNIVAC literature.
UNIVAC's literature talked up the transistor, but its primary logic was composed of 3000 "Ferractor" magnetic amplifiers and 700 transistors, with 20 vacuum tubes for power control, including six 4CX250B power tetrode tubes in the clock amplifier, running in push-pull/parallel, from a power supply providing -1.6KV cathode and -800V screen, yielding an output of a kilowatt for a clock speed of 707 kilohertz.
this is the nightmare. the UNIVAC Solid State was the only large-scale commercial machine based primarily on mag amps. magnetic amplifiers are essentially a transformer that can be controlled with an extra winding; applying current pushes the core into saturation, allowing it to act as a switch, or a low-gain low-frequency amplifier. they were much more reliable than contemporary transistors and vacuum tubes, with more tightly controlled specifications than transistors, and possessed certain electrical characteristics that allowed for reduction of component count compared to transistors, saving costs.
from chapter 7 of "Birthing the Computer: From Drums to Cores", the basic Ferractor operational cycle is as follows:
Each Ferractor performed a 1-bit operation. In the first half of a cycle, the Ferractor was magnetized or not depending on the current in a bias coil. In the second half, the Ferractor showed on a secondary coil either high- or low-impedance. Low impedance generated a bias current on the next Ferractor.
the disadvantage is that a magnetic amplifier is a magnetic device; using them in computing applications required the generation of a very high-current clock signal. this could not be generated with transistors. instead, it was read from a "timing band" written to the main drum memory, rotating at 17,670 RPM in helium. this is essentially the equivalent of the clock being based off a very high-speed tape recording. this signal was then amplified to 1,000 watts, distributed throughout the machine as high-voltage RF, and stepped down to 36 volts where it was needed by oil-filled transformers. (this works out to a total 27 amps of clock, split between areas of the computer.)
the computer could be very easily heard on 707 KH (the clock signal) and 1414 KHz (its first harmonic) AM, despite the shielded enclosure for the clock amplifier. i do not know of any other computer constructed that required a high-power clock, because it is a hideous, horrifying idea. this machine was reliable (with at least one unit in the Netherlands kept operational until 1977), it was fast, it was well-received, and it should never have been built
