Comment by ttul

Google TPUs are built around a 128×128 systolic array of multiply-accumulate (MAC) units. Trainium 1, Trainium 2, and Inferentia 2 also feature a 128x128 systolic array.

You learn something every day. Today, it was the term "systolic array": A systolic array is a specialized grid of simple, interconnected processing units designed to execute parallel data operations—like matrix multiplication—by rhythmically passing data directly from cell to neighboring cell without writing intermediate results back to main memory.

The term comes from the biological word systole (the contraction of the heart pumping blood through the body). In a systolic array, data "pulses" through a network of processing elements on every clock cycle, driven by a global clock beat.