Computer Architecture: Measures for Design & Analysis

Traces forty-two years of computer performance growth, driven by transistor scaling and architectural innovations such as risk architecture and caches, and notes the slowdown.

Explore Denel scaling limits and why microprocessor growth slowed due to power, and how architects use parallelism and domain-specific cores to sustain performance gains in devices like smartphones.

Explore classes of computers, categorized by application requirements and computing technology. Learn how transistor scaling, miniaturization of ICs, and domain-specific architectures drive mobile devices, servers, IoT, and machine learning hardware.

Embedded systems integrate a processor, memory, and input/output to perform specific tasks, then connect to the internet as IoT for monitoring, analysis, and control with low power edge processing.

Explore how designers balance cost, energy, and performance in battery-powered mobile devices by optimizing memory, power dissipation, and responsiveness under embedded and real-time constraints.

Examine how desktop and laptop designs balance price, energy constraints, and graphics performance, from $300 to $2500, with tablet-mode 2-in-1s and always-on connectivity for work, study, and online prescriptions.

Explore how servers provide storage and computing services across organizations and travel experiences, serving as the backbone of enterprise computing. Analyze availability, scalability, and throughput as key design factors.

Explore warehouse scale computers, where tens of thousands of computers form a single system over a LAN, balancing cost, availability, and cooling with software layers.

Explore parallel architecture by combining data-level and task-level parallelism, including data-level parallelism, instruction-level parallelism, and vector architecture, with pipelines and speculative execution.

Explore Flynn's classification of parallelism, covering SISD, SIMD, MISD, and MIMD architectures, and explain how instruction and data stream configurations influence design choices and future power and energy performance calculations.

Define the job of a computer architect as designing systems to maximize performance and energy efficiency under cost, power, and availability constraints, addressing instruction set design and implementation under EDA.

Explore instruction set architecture as the contract between hardware and software, covering registers, addressing modes, load/store operations, and encoding across RISC and x86 families.

Explain genuine computer architecture as a blend of instruction set architecture, organization (micro architecture), and hardware, focusing on memory systems and interconnect design of the processor to meet functional requirements.

Assess trends in technology and instruction set architecture to guide designs that survive rapid shifts in transistor density, DRAM, flash, magnetic storage, and network performance.

Understand how bandwidth outpaces latency across microprocessors, memory modules, and networks, with throughput in MB/s and latency in ms; bandwidth growth is at least the square of latency improvement.

Shrinks feature size from micrometers to nanometers, boosting transistor density and lowering voltage, enabling multicore on a chip while wire delays and impedance rise challenge performance.

Investigate how power distribution, heat dissipation, and thermal design power constrain modern microprocessors, and why energy best measures performance for given workloads.

Explore energy and power in modern microprocessors, from dynamic energy per switch and capacitance to clock rate, dvfs, power gating, leakage power, and domain-specific processors for higher performance per watt.

Explore how cost drives design decisions through the learning curve, time, yield, and manufacturing volume, and how commoditization and vendor competition push lower costs while maintaining quality.

Understand how ic cost derives from wafer yield, packaging, and testing. See how process complexity and defects reduce yield and raise cost, and how redundancy improves yield.

Explore how cost, price, and margins drive computer industry decisions, balancing capex and opex with manufacturing costs, power, cooling, and r&d for market success.

Explore dependability in computer architecture by examining availability, reliability, and maintainability, and how service level agreements and objectives ensure uninterrupted service.

Explore how execution time, wall clock time, and response time shape computer performance, highlighting throughput, I/O, and OS overhead, while benchmarks compare CPUs.

Explore how benchmarks measure processor performance using standard tests and benchmark suites, examine bench marketing risks, and compare real-world throughputs across apps like AR, ML, and 3D rendering.

Ensure reproducibility by detailing benchmarks, configurations, and baseline and optimized results for specked benchmarks, including spec ratio, aspect ratio, and geometric mean.