The original machine stood 9 feet by 12 feet by 5 feet, but most of this was just to house the display [and impress the crowds...?]. Electronics took up less than 2% of that volume. It ran at a healthy six kilowatts, though four of these were for the display lamps.
It housed 480 valves, all 12AT7 double-triodes. Only 350 of them actually took part in the action; the rest were spares being "burned in". 120 relays drove the displays. A few germanium diodes were used as 'OR' gates. Most of the connections were directly soldered — rather than using pluggable connectors — for reliability. (This, however, made for difficult transportation!)
Unlike most digital computers, then and now, NIMROD used directly-coupled voltage levels rather than clocked pulses, so it could run as slowly as the demonstrator wished. When run at full speed, it was driven by 10kc/s steps [10kHz in modern parlance].
Designing a processor was no doubt rather simpler in those days, but work on the design was begun on 1 December 1950, and it was delivered on 12 April 1951. The Science Exhibition opened on the 4th of May, and ran through September. Unlike transistors, valves burn out, so there were a few failures, but it seems these were easily traced, amd things would be running again in an hour or so. There was one original wiring mistake that early on caused a couple of errors a day, but once this was found it ran flawlessly.
In a piece on "Digital Computers Applied to Games"†, Alan Turing recounts a rather more off-beat happening during its stay in South Kensington. Some folks from the Society for Psychical Research set up in a nearby room, to see if they could influence — or be influenced by — the progress of the machine's game. Sadly, they had no success either way.
In October Nimrod went on to Berlin for three weeks, to thrill crowds in the Industrial Show there, and finally on to Toronto for a showing to the Society of Engineers. Apparently it was lost track of after that, somewhere in N. America.
† Chapter 25 in "Faster than Thought", B. V. Bowden ed. [Pitman 1953]
A comprehensive review of the state of computing at that time. Turing's chapter — as well as a review of chess and draughts playing programs — describes the logic of multiple NIM, where a player may take from more than one heap in a turn.