EVOPEDIA

← all systems · see it in The System

Computer hardware

The physical parts of a computer: processor (CPU), memory (RAM), motherboard, storage, graphics and sound cards, case, power supply, and the external devices (monitor, mouse, keyboard, speakers) (src:wp:computer-hardware); 'hard or rigid with respect to changes', in contrast to software, which it runs and by which it is 'typically directed'. Built from integrated circuits that 'typically contain billions of tiny' MOSFETs, made in semiconductor fabrication plants. Where the line between hardware and software falls (firmware, virtual hardware) is contested (C-1404).

Properties published scienceour proposal

Sourced fields, and our own readings.

born at1936-1941: Zuse's mechanical binary Z1 (1936-1938) and relay Z3 (1941); relay machines of Stibitz and Aiken; vacuum-tube machines of the 1940s (which was first: C-568)
born precisionyears
boundaryInside: the physical parts of the machine, from transistors and wires through logic circuits, chips (CPU, memory, storage, GPU, controllers) and the motherboard's buses, the power supply, the cooling and the case, up to the instruction set architecture and the firmware that wakes the machine at power-on. Outside: the software that directs it (programs, operating system), the people who use it, the factories that make its chips, the power grid, the air its heat goes into, the networks it joins, and stray particles from space and from its own packaging. On the line: firmware (BIOS/UEFI in ROM or flash), microcode, and virtual hardware ('software that mimics the function of hardware'); contested as C-1404.
nested levelA part of the computer (sys:computer = hardware plus the software it runs); built from integrated circuits made of transistors on silicon; carries the software stack, the operating system and the programs (they run on its L7, booked from their side).
scaletechnological

Measures (35) published science

susceptible DRAM cells (worst module)1 in 1,700
soft error rate per DRAM chip at sea level (one experiment)5,950 FIT (failures per billion hours)
PSU output voltages12, 5, 3.3 (from 120-277 V AC)
accesses to induce a disturbance error139,000 (as few as) accesses
fabrication time, advanced nodeup to 15 (average 11-13) weeks
global computer hardware revenue (2023)705.17 billion USD
fan hysteresis (one setup)on 75, off 65 °C
USB port power5 V, 500 mA (2.5 W)
sea-level neutron flux (New York City)about 14 neutrons/cm2/hour
water use of a large fab (40,000 wafers a month)up to 4.8 million gallons per day
internal bus lines50-100 lines
Moore's law doubling time (1975 revision)2 (41% a year) years
neutron flux increase with altitudeabout 2.2 per 1000 m factor
chip-making tool cost doubling time4 years
upset rate in aircraft relative to sea level>300 times
throughput gain from active coolingup to about 90 %
DRAM modules with disturbance errors110 of 129 (836 of 972 chips)
water per 300-mm waferabout 2,200 (1,500 ultrapure) gallons
characteristic clock: processor clock frequency range (CPU makers after 2004)1-6 GHz
DRAM retention window (refresh clock)64 ms
Z1 machine cycleabout 4 (reconstruction run at no more than 1) Hz
light-speed reach in one 5 GHz cycleabout one chip width in 200 ps
Landauer limit per erased bit at room temperature2.85e-21 J
IC power dissipation0.1-100 W
share of chip energy dissipated in wires and repeaters30-60 %
share of gates that are repeaters50-80 %
on-chip power-network efficiency80-90 (soon) %
supply voltage0.5-2 (floor about 0.2) V
hardware performance growth 1986-2003>50 % per year
carry-lookahead addition timeproportional to log n, area proportional to n
peak FLOPS vs DRAM vs interconnect bandwidth growth (20 years)3.0x / 1.6x / 1.4x per 2 years
US electricity used by data centres (2011)2.2 %
manufacturing share of iPhone life-cycle carbon49 (3GS) to 86 (iPhone 11) %
thermal throttling limit and range (Raspberry Pi 4B)80 °C; 1.5 GHz down to 600 MHz
IBM 1996 estimate of desktop soft errors1 per month per 256 MiB

Open contentions (5) open contention

Where sources or readings disagree. Both stay on the map.

Partners and parts published science

Built from / builds (13)

Sits inside / holds (5)

Trades with (valence) (29)

expressed by (13)

floor filled by (36)

represents (1)

Kinds (13)

wired (5)

Notes published science

Layers

The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.

computer hardware as computer engineers describe it: the physical world at the machine up to the computing system software directs (reader's view)

  1. L1: The physical world at the machine: mains electricity, heat and the air or liquid that carries it away, the materials the parts are made of (silicon, copper, metals), people's fingers and eyes at the devices, radio and cable signals, stray particles (cosmic-ray neutrons, alpha particles from packaging)
  2. L2: Devices: switching elements and storage cells. Today MOSFETs and FinFETs on silicon (billions per chip); earlier steel plates (Z1), relays, vacuum tubes; memory cells as capacitors (DRAM), cross-coupled flip-flops (SRAM), flash cells, magnetic domains
  3. L3: Logic and circuits: gates built on Boolean algebra, adders (ripple or carry-lookahead), flip-flops and latches paced by the clock; wires and repeaters (interconnect) joining them
  4. L4: Chips and components on the motherboard, joined by buses: the CPU (ALU, registers, control unit, caches), DRAM, storage, GPU, network and USB controllers, voltage regulators, firmware ROM
  5. L5: Organisation (microarchitecture): the instruction cycle, pipelining and speculative execution, the memory hierarchy (registers, caches, main memory, storage), buses and interconnect, power and thermal management
  6. L6: The instruction set architecture (CISC, RISC, vector) and the firmware (BIOS, UEFI) that wakes the hardware: the interface software sees
  7. L7: The computing system at work: hardware executing the commands software gives it, as a personal computer, phone, mainframe, supercomputer, warehouse-scale computer or embedded system, its results shown, sent, stored or used to control machines

the boardroom (shared floors)

  1. L4: Chips and components on the motherboard, joined by the internal bus (address, data and control lines): CPU, DRAM, storage and network controllers, GPU, voltage regulators
  2. L5: Organisation (microarchitecture): the instruction cycle, pipelining and speculation, the memory hierarchy and caches, power and thermal management
  3. L6: The instruction set architecture and the firmware: what software sees of the hardware; the BIOS or UEFI that initialises it at power-on
  4. L7: The computing system at work, directed by software: personal computer, phone, mainframe, supercomputer, warehouse-scale computer, embedded system

way in: Electric power (mains, battery, USB)

  1. L1: Mains electricity from the grid (data centres drew 2.2% of US electricity in 2011)
  2. L2: The power supply unit converts 120-277 V AC to 12, 5 or 3.3 V DC
  3. L3: Voltage regulator modules of power MOSFETs set each component's voltage
  4. L4: The chip's own power-ground network delivers 0.5-2 V to its transistors; power rises with the square of voltage

way in: Heat (temperature)

  1. L1: Processors make heat; ambient air and a user's hand add to it
  2. L2: The junction temperature inside the semiconductor is what matters
  3. L3: Thermal sensors in the CPU, motherboard, chipset or GPU
  4. L4: A set limit is crossed (80 °C on a Raspberry Pi 4B)

way in: Neighbouring cells (DRAM disturbance, row hammer)

  1. L1: A row is opened again and again; its wordline toggles on and off
  2. L2: Coupling drains charge from cells in nearby rows faster than refresh restores it

way in: Network signals (cable and radio)

  1. L1: Radio at antennas (phones) and signals on cables
  2. L2: The network interface controller
  3. L3: Frames moved into main memory without the CPU (DMA)

way in: Sensors (camera, microphone, GPS; phones)

  1. L1: Light, sound and satellite signals at a phone's sensors
  2. L2: Sensors and analog-to-digital converters: in von Neumann designs sensing, memory and computing sit apart
  3. L3: Data moved from sensor to memory to processor; in-sensor computing does some of the work where it is sensed, after the retina

way in: Stored data coming back (disks, SSDs, flash, ROM)

  1. L1: Bits held in non-volatile media: flash, ROM and its kin, magnetic disks
  2. L2: The drive reads them: larger and cheaper than memory, slower to reach
  3. L3: DMA moves the blocks into main memory

way in: Stray particles (cosmic-ray neutrons, alpha particles)

  1. L1: Secondary particles from cosmic-ray air showers (about 14 neutrons per cm2 per hour at sea level in New York, more with altitude, less with sunspots), and alpha particles from radioactive decay in the package
  2. L2: A particle lands on the silicon die and frees charge in a cell
  3. L3: The stored bit flips (single-event upset)
  4. L4: Error-correcting codes or parity catch it, where present; in many systems it goes unseen

way in: The clock (time)

  1. L1: An electronic oscillator, the clock generator
  2. L2: The clock signal: a square wave, most often 50% duty cycle, between high and low at constant frequency
  3. L3: Distributed by a metal grid to every flip-flop and latch, which all change state at once
  4. L4: The clock period is set above the slowest (worst-case) signal path in the chip

way in: The making: patterns written into silicon at the fab

  1. L1: Layers deposited on a silicon wafer and patterned with light, all components at once
  2. L2: Repeated photolithography, deposition, implantation and etching build the circuit on the wafer; the chip is printed as a unit
  3. L3: Clean mini-environments keep particles off; yield counts the working dies
  4. L4: Dies cut from the wafer, packaged and mounted

way in: The stored program (instructions and data from memory)

  1. L1: Instructions and data held by one mechanism in memory
  2. L2: Brought up the memory hierarchy, from DRAM through caches to registers
  3. L3: The program counter names the next instruction to fetch
  4. L4: Fetched, decoded and executed: the instruction cycle

way in: User input (keyboard, mouse, touchpad, touchscreen)

  1. L1: A person's fingers press, move and touch
  2. L2: Keyboards, mice, touchpads, pointing sticks and touchscreens turn movement into signals
  3. L3: The external bus (most often USB), run by a bus controller at its own speed

way out: Display (light)

  1. L1: The monitor's light reaches the user's eyes
  2. L4: The GPU (on-die or on a card) draws the image
  3. L7: Software sends what is to be shown

way out: End of life: e-waste and recycling

  1. L1: Recycled, or broken up and burned informally, harming workers
  2. L2: Hazardous and valuable materials in the parts
  3. L7: Machines replaced every few years (warehouse-scale computers) and discarded

way out: Heat out (heatsinks, fans, liquid)

  1. L1: Heat into room air or data-centre coolant
  2. L2: Every erased bit pays at least kT ln 2; real chips pay far more, much of it in wires
  3. L3: Heatsinks, fans, liquid coolers; the case directs the air
  4. L5: Fans switched by temperature (on at 75 °C, off at 65 °C in one setup)

way out: Network out

  1. L1: Signals on the cable or radio
  2. L4: The network interface controller sends

way out: Power out to devices (USB)

  1. L1: 5 V up to 500 mA on a standard port

way out: Recording out (disks, SSDs, flash)

  1. L1: Bits held without power
  2. L4: DMA moves blocks from memory to the drive

way out: Sound (speakers)

  1. L1: Speakers move the air
  2. L4: The sound card or on-board audio

way out: Throttle or shut down (frequency, voltage, clock gating)

  1. L1: Less heat made (and less work done); at the limit, shutdown
  2. L2: Lower frequency and voltage, or features switched off
  3. L5: Power management: clock and power gating, per-core voltage scaling, a core dedicated to power scheduling

way out: Writing into memory cells (stores, refresh, error correction)

  1. L1: The cells hold full or empty charge again
  2. L2: The row is opened: the row-buffer reads the cells and writes their charge back
  3. L4: The memory controller issues writes and refresh commands; ECC rewrites corrected data

Who fills each layer

Sources (42) published science

Computer hardware (Wikipedia, via Wikimedia Enterprise snapshot)
Memory hierarchy (Wikipedia, WME snapshot) (behind its links)
Wikipedia: Central processing unit (behind its links)
Kocher et al. 2018, Spectre Attacks: Exploiting Speculative Execution (arXiv:1801.01203) (behind its links)
Logic gate (Wikipedia, WME snapshot) (behind its links)
Von Neumann architecture (Wikipedia, rev 1381757639, 2026-10-24, via Wikimedia Enterprise) (behind its links)
Carry-lookahead adder (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Rojas R. The Z1: architecture and algorithms of Konrad Zuse's first computer. arXiv:1406.1886 (v3 2025) (behind its links)
Computer memory (Wikipedia, WME snapshot) (behind its links)
Integrated circuit (Wikipedia) (behind its links)
Transistor (Wikipedia) (behind its links)
Kim, Daly, Kim, Fallin, Lee, Lee, Wilkerson, Lai and Mutlu, Flipping Bits in Memory Without Accessing Them: An Experimental Study of DRAM Disturbance Errors, ISCA 2014 (doi:10.1145/2678373.2665726) (behind its links)
Wikipedia: Soft error (behind its links)
Semiconductor device fabrication (Wikipedia, WME snapshot) (behind its links)
Benoit-Cattin T, Velasco-Montero D, Fernandez-Berni J. Impact of thermal throttling on long-term visual inference in a CPU-based edge device. arXiv:2010.06291 (2020) (behind its links)
Sandhu, Zumeit et al. Semiconductor manufacturing wastewater challenges and the potential solutions via printed electronics. iScience (2025). doi:10.1016/j.isci.2025.113576 (PMC12510042) (behind its links)
Moore's law (Wikipedia, WME snapshot) (behind its links)
Markov IL (2014) Limits on fundamental limits to computation. Nature 512:147-154; arXiv:1408.3821 (behind its links)
Brent RP, Kung HT. A regular layout for parallel adders. IEEE Trans. Computers C-31:260-264 (1982); author's retyped copy. doi:10.1109/TC.1982.1675982 (behind its links)
Gholami A, Yao Z, Kim S, Hooper C, Mahoney MW, Keutzer K. AI and memory wall. arXiv:2403.14123 (2024); IEEE Micro (behind its links)
Gupta U et al. Chasing carbon: the elusive environmental footprint of computing. arXiv:2011.02839 (2020); HPCA 2021 (behind its links)
Lipp et al. 2018, Meltdown, arXiv:1801.01207 (behind its links)
Clock signal (Wikipedia, WME snapshot) (behind its links)
Fan X, Weber W-D, Barroso LA (2007). Power Provisioning for a Warehouse-sized Computer. ISCA '07. doi:10.1145/1250662.1250665 (behind its links)
Mell P, Grance T (2011). The NIST Definition of Cloud Computing. NIST Special Publication 800-145. doi:10.6028/NIST.SP.800-145 (behind its links)
Computer cooling (Wikipedia, WME snapshot) (behind its links)
Issah I, Arko-Mensah J et al. Health risks associated with informal electronic waste recycling in Africa: a systematic review. Int J Environ Res Public Health 19:14278 (2022). doi:10.3390/ijerph192114278 (PMC9655142) (behind its links)
Wikipedia: Operating system (behind its links)
Programming language (Wikipedia via Wikimedia Enterprise, cached research/showcase/cache/wme) (behind its links)
Leiserson CE, Thompson NC, Emer JS, Kuszmaul BC, Lampson BW, Sanchez D, Schardl TB (2020) There's plenty of room at the Top: What will drive computer performance after Moore's law? Science 368:eaam9744. doi:10.1126/science.aam9744 (author's final manuscript, MIT Open Access, CC BY-NC-SA 4.0) (behind its links)
Malware (Wikipedia, via Wikimedia Enterprise snapshot) (behind its links)
Anderson R (2008) Security Engineering, 2nd edition, chapter 4 Access Control (author's free copy) (behind its links)
Draper, Kutin, Rains and Svore, A Logarithmic-Depth Quantum Carry-Lookahead Adder, arXiv:quant-ph/0406142 (preprint) (behind its links)
Gray J (1981) The Transaction Concept: Virtues and Limitations. Tandem Technical Report TR 81.3 (VLDB 1981) (behind its links)
Armbrust M, Fox A, Griffith R, et al. (2009). Above the Clouds: A Berkeley View of Cloud Computing. UC Berkeley EECS-2009-28 (behind its links)
Katte, Fernandez 2023. A Survey Report on Hardware Trojan Detection by Multiple-Parameter Side-Channel Analysis. arXiv:2307.02012 (behind its links)
Hadi HJ, Ullah I, Harris S (2023) SSD Forensic: Evidence Generation And Forensic Research On Solid State Drives Using Trim Analysis. arXiv:2307.10192 (behind its links)
Mutlu O, Ghose S, Gomez-Luna J, Ausavarungnirun R, Sadrosadati M, Oliveira GF. A Modern Primer on Processing in Memory. arXiv:2012.03112 (behind its links)
Wikipedia: Direct memory access (behind its links)
Tang H, Yu et al. In-sensor-memory computing for post-von Neumann intelligence: a perspective. Nano-Micro Letters (2026). doi:10.1007/s40820-026-02191-y (PMC13092463) (behind its links)
Park S-Y, Jeong EG, Jo S-H (2025) Soft-Error-Resilient Static Random Access Memory with Enhanced Write Ability for Radiation Environments. Micromachines 16:1212. doi:10.3390/mi16111212 (CC BY) (behind its links)
Bérut, Petrosyan & Ciliberto, Information and thermodynamics: Experimental verification of Landauer's erasure principle, arXiv:1503.06537 (J. Stat. Mech. 2015) (behind its links)