The genious part was to simplify the problem
I mean that's what distinguishes him from modern day computing company managers. A C64 probably took more engineering effort to design.
The Cray 1 didn't use any custom silicon. It used generic ECL gates which you could buy in bulk in any store. It used careful design to get speed out of it. For example every board was designed so the propagation delay was constant. Every line between 2 components and particularly between 2 boards was a well run propagation line. Every long line between 2 boards had the same length. All of this suddenly makes the problem much easier as you could count on certain universal preconditions. For example your signal would arrive 1 clock cycle later at the other board because of wire delays. It would _always_ do that and you could count on that.
ECL also has the nice effect of taking a constant amount of power. That way there are no current transients on your boards which are a huge problem today. That's why, under most CPUs in modern PCs you will find a whole battery of capacitors to satisfy the current demands. The Cray just took a constant current. This also simplified the power supply. It was a simple 6 phase rectifier with a bit of capacitors after it. The regulation was done externally with an electromechanical converter converting both the line power to 400 Hz 3 phase as well as regulating the output voltage for slow variations of the supply voltage.
There are 2 talks by him on Youtube. They are worth watching, even if you are not into engineering. He's a rather good speaker:
https://www.youtube.com/watch?v=vtOA1vuoDgQ
https://www.youtube.com/watch?v=xW7j2ipE2Ck
BTW Steve Jobs was a salesperson, Seymour Cray actually designed most of the logic in boolean equations. So comparing those is kinda offensive to engineers.
Biting the hand that feeds IT
