Data General originally made a name for itself with a simple minicomputer called a "Nova."
The big innovations of the Nova were not technical, so much as packaging. It saved money and increased reliability by packaging as much as possible of the computer on a single huge 15 inch by 15 inch printed circuit board.
The backplane was wirewrapped. The peripherals were literally daisy-chained to establish the interrupt priorities.
The Nova was a 16-bit machine. Each instruction was one word. Each word had a number of subfields:
jump (single bit), opcode, source, destination, shift, skip (2 bits) [corrections would be appreciated- I lost my manual years ago!]
Basically, if the jump bit was set, the rest of the word was the address. Since the Nova addressed memory as 16-bit words, this gave it a 64K-byte address space, which was adequate.
Otherwise, the opcode, source, and destination would be executed, then the shift performed on the destination, and finally, the conditional skip would depend on the result of the condition codes set by the operation.
The op-code was, I recall, a 4 bit field. The register fields were each 3 bits. One of the selections was an immediate value, and the other was a "register indirect" fetch. The Nova had a unique feature in that if it fetched a value whose most significant bit was 1, it sould treat the rest of the word as an address, and fetch from that address. It could get address-dereferencing chains that were hundreds of fetches long. (Not quite unique. Some older DEC computers had this silly feature. De Castro and some other Data General engineers previously worked at DEC :-)
The shift would describe the type of shift and direction to perform on the destination.
The skip would tell whether to skip the next instruction, based on the result of the operation. It could skip on zero, carry, negative, or don't skip. Usually, the thing to skip was a jump instruction.
This was all pretty crude, but it was quite effective and very fast for its day, at least at this low-cost end of the market.