Advanced Architecture Topics
Modern computer architecture extends beyond traditional CPUs to include specialised processors, reconfigurable hardware, and embedded systems. Understanding these technologies is essential as computing becomes increasingly diverse and application-specific.
GPU Architecture
A GPU (Graphics Processing Unit) has thousands of small cores optimised for parallel data processing. Originally designed for graphics rendering, GPUs now handle general-purpose computing (GPGPU) using frameworks like CUDA and OpenCL. GPUs excel at matrix operations, machine learning, and scientific computing.
FPGA
A Field-Programmable Gate Array is a chip whose logic can be reconfigured after manufacturing. FPGAs contain programmable logic blocks, interconnects, and I/O blocks. They offer hardware-speed processing with software-like flexibility. Applications include prototyping, signal processing, and network acceleration.
Embedded Systems
An embedded system is a computer designed for a specific function within a larger system. Characteristics include real-time constraints, low power consumption, and limited resources. Microcontrollers (Arduino, PIC, STM32) are common embedded processors. Examples: automotive systems, medical devices, IoT sensors.
RISC-V
RISC-V is an open-source ISA based on RISC principles. Its modular design allows custom extensions. Being open-source, it avoids licensing fees of ARM/x86. RISC-V is gaining adoption in embedded systems, IoT, education, and increasingly in general-purpose computing.
Memory Technologies
Emerging memory technologies include 3D NAND (stacked flash layers for higher density), HBM (High Bandwidth Memory for GPUs), NVMe SSDs (fast storage via PCIe), and research into MRAM, ReRAM, and phase-change memory. These aim to close the processor-memory speed gap.
Quantum Computing Basics
Quantum computing uses qubits that can be in superposition of 0 and 1 simultaneously. Entanglement and interference enable quantum algorithms. Shor's algorithm factors integers exponentially faster; Grover's algorithm searches unsorted data quadratically faster. Quantum computers complement rather than replace classical ones.
Performance and Power
Modern design balances performance, power, and area (PPA). Dynamic Voltage and Frequency Scaling (DVFS) adjusts power consumption. Dark silicon limits how much of a chip can be active simultaneously. Heterogeneous computing (CPU + GPU + accelerators) optimises different workloads.
Summary
Advanced architecture encompasses GPUs, FPGAs, embedded systems, open ISAs like RISC-V, emerging memory technologies, and quantum computing. These diverse technologies address the growing variety of computing workloads and constraints.