CISC vs RISC: How the tables turned

Imagine its the year 800AD, you left your village to go hunting, and when you come back your food is gone, and your family is missing. Your village has been raided!

You need information, who did it, when, and where did they go? You can ask two survivors in the village, one old man who is extremely slow in speaking, but very detailed in his langauge. He will explain everything in a few words that you need to know, but pull up a chair. There is also a young girl in the village, who speaks very quickly, but has a very simple vocabulary. She will explain the same information but in far more words, which may or may not take longer in the end.

The old man is CISC architecture, and the young girl is RISC.

The war between Complex Instruction Set Computers (CISC) (like Intel X86) and Reduced Instruction Set Computers (RISC) (like ARM) peaked in the 90s. RISC architectures like PowerPC and MIPS were theoretically better. They executed simple instructions extremely fast, but the same software needed more instructions and used more RAM. But Intel's x86 architecture had the software ecosystem and worked out faster for most tasks. Intel just threw billions of dollars and increasingly massive transistor counts at the problem. They made CISC fast enough to blur the lines, and high RAM prices did the rest.

Then mobile happened, the iPhone, iPad and others running on tiny, thin batteries. The smaller the better.

You can't negotiate with a phone battery. When Apple and Google needed chips for smartphones, x86 was simply too power-hungry. They went with ARM, a RISC architecture. For years, ARM was viewed as a lightweight, great for saving power, but not suitable for RAM heavy desktop software.

The turning point wasn't gradual. Apple dropped the M1 chip and suddenly a fanless MacBook was outperforming Intel's flagship desktop processors while sipping a fraction of the power. It proved that RISC wasn't just more efficient. At modern manufacturing scales, it was fundamentally faster.

Now, AWS is filling data centers with custom Graviton processors. The cloud runs on electricity, and electricity is expensive. x86 is losing its grip on the server market for the exact same reason it lost mobile.

The tables didn't just turn. The CISC approach finally hit the wall of physics, and RISC won. For now.