How limited would computers be if silicon transistors were never invented?

Question

In a world where silicon transistors were never invented, how many limitations would be incurred if we had to continue to use germanium transistors? In this world, silicon transistors were never invented, no one thought of them. Don't question why, I don't have an answer. Instead, the primary transistor material remained Germanium.

I have heard the primary limitations of germanium transistors were smaller band gap, worse thermal conductivity, & primarily a higher cost. How much would this limit computing capabilities? Especially with germanium transistors being more expensive than vacuum tubes (afaik) meaning cheaper computers would be stuck with those.

Note that I'm assuming germanium transistors because from what I can find that was the primary alternative material for transistors before silicon. If other things can be used it would be helpful to know how other varieties of transistors affect things.

What limitations would this scenario create for computing with the lack of silicon transistors?

Answer 1

A frame challenge in the form of a little history.

TL;DR: Nobody invented the silicon transistor. It was always known that where germanium worked, silicon would work too. What needed to be perfected was how to manufacture sufficiently pure silicon crystals, and from the very beginning the race was on.

Long version:

Silica was of course always known. Stone Age people made tools out of flint, of which the Latin name is silex. (That's silec-s, stem silic-.) By the end of the 18th century, chemists began to suspect very strongly that silica was the oxide of an unknown element, for which Sir Humphry Davy FRS proposed the name silic-ium, with -ium because he believed that the still not yet isolated element was a metal. Silicium was first isolated in pure-ish form by Berzelius, in 1823, and it was not a metal; hence, the name silic-on, like bor-on and carb-on.

By the middle of the 19th century, sufficiently many elements were known that chemists began searching for some sort of order in the chaos. After quite a few good but incomplete attempts, the prize went to the Russian chemist Dmitri Mendeleev, who had the idea of arranging the elements in a two-dimensional table in increasing order of their atomic weights. (How atomic weights were discovered and measured is a fascinating story; the names of the English chemist John Dalton FRS and of the Swedish chemist Jacob Berzelius feature prominently.) Mendeleev's table showed that the elements which fell in the same column had similar chemical and physical properties, with the columns showing definite trends as one went top to bottom.

The table was published in 1871. But in order for his table to show those beautiful periodic trends, Mendeleev had to leave empty boxes in the table, to be occupied by yet undiscovered elements. This is one of the best examples of scientific prediction. Three of the elements prophesized by Mendeleev were eka-boron (later discovered by Lars Fredrik Nilson and named scandium), eka-aluminium (later discovered by Lecoq de Boisbaudran and named gallium), and eka-silicon, discovered by Clemens Winkler in 1886 and named germanium. (Gallium and germanium because Lecoq de Boisbaudran was French and Clemens Winkler was German...)

^{Eka- means one in Sanskrit. Mendeleev was friends with the notable Sanskritist Otto von Böhtlingk, so he named his predicted elements using Sanskrit eka- and dvi- prepended to the name of the element in the table which fell in the same column above the empty boxes.}

The point is that germanium was always known to be similar to silicon, fifteen years before germanium was discovered. There was no point in time when germanium was known and it was not known that germanium and silicon had similar properties.

When William Shockley and his colleagues at Bell Labs invented the transistor in 1947 using germanium crystals, they already knew that silicon would work too. Herbert Mataré had already developed germanium and silicon rectifiers for use with German radar during the Second World War. The problem was only manufacturing pure silicon crystals, and several competing teams were already working on the problem. In the end, the first silicon transistor was demonstrated at Bell Labs by Morris Tannenbaum in January 1954 as a one-off wonder; but, unknown to him, Texas Instruments' Gordon Teal was much more advanced in the quest for a practicable industrial process, and by May 1954 TI's silicon transistors came on the market.

Answer 2

There are other chemistries appropriate for transistors, and more importantly, other technologies that were tried as well. Mechanical calculators were common in the early twentieth century and could be scaled up with precision industrial manufacturing. Vacuum tube computers were the first large-scale electric computers and were successful, just really hot and prone to failure. Punch cards were replaced by core memory, which were replaced by various kinds of silicon memory.

Many technologies were tried, many failed, many were replaced. In general, performance improved, efficiency improved, reliability improved, connectivity improved, power usage and heat production improved. People tried out different ideas and went with what worked better.

In other words, progress happens in respond to need, and not necessarily in a straightforward fashion. Maybe in your world there's relatively little pure silicon, or maybe nobody figured out how to run an A/C electric motor so large-scale electric infrastructure never got built, or some other fundamental thing we take for granted never got going. Aluminum refining, for instance, was a huge win and would absolutely affect your world.