Work

A Comparative Analysis of Central Processing Unit and Neural Processing Unit Architectures

Year: 2026

Type: preprint

Abstract: The increasing computational demands of machine learning have driven the development of specialized processor architectures. This paper presents a comparative analysis of Central Processing Unit (CPU) and Neural Processing Unit (NPU) architectures, focusing on computational execution, memory organization, performance, and power efficiency. The study compares CPU general-purpose instruction execution and hierarchical memory systems with NPU parallel execution, multiply-accumulate (MAC) operations, and reduced-precision arithmetic. The analysis highlights the architectural trade-off between CPU flexibility and NPU specialization for different computational workloads.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.23012099

Author: John Benedict, Esteva

Language: English

Topic: Parallel Computing and Optimization Techniques

Subfield: Hardware and Architecture

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7214647652

Cites:

Cited by:

Related to: 10