CHAPTER 06 / PHYSICAL ASSEMBLY
What physical equipment does AI run on?
Chips and computing systems
AI computation runs on processors supported by memory, storage and networking. These parts form devices, servers and sometimes large clusters. GPUs are one important kind of accelerator, not a requirement for every AI task.
A division of work
CPUs handle general-purpose computation and coordination. GPUs and other accelerators can perform many suitable operations in parallel. Memory holds active values, storage retains data and networking moves information between systems. Performance depends on how these parts work together.
From chip to system
A processor is packaged and mounted in a system with connections for data and power. Boards, cooling and power supplies support operation. A server combines these parts; a cluster connects multiple machines. AI can also run on a workstation, phone or embedded device.
Explore an assembly
The hardware explorer moves from a rack to a server and an illustrative GPU assembly. Separating physical parts answers “what is inside?” Tracing manufacturing, electricity or software dependencies answers “what enables this?” Those are different relationships.
Keep this connection in mind.
A chip is part of a system. Follow computation, information movement, electricity and heat removal together.
- Computing software runs on Computing hardware
- Computing systems can be housed in Data-center facilities
- Computing equipment powered by Electricity
- Computing equipment produced through Manufacturing processes
Sources for this introduction
- Google · Machine Learning Glossary ↗
Models, parameters, training, inference, datasets and evaluation.
- U.S. Department of Energy · Data center design guide (2024) ↗
IT equipment, electrical distribution, cooling and facility efficiency.
Draft teaching text, prepared with AI assistance. Larry’s editorial review is pending. Read the editorial approach.
EXPLORE RELATED AREAS