Read-only memory (ROM) is a type ofnon-volatile memory used incomputers and otherelectronic devices. Data stored in ROM cannot be electronically modified after the manufacture of thememory device. Read-only memory is useful for storingsoftware that is rarely changed during the life of the system, also known asfirmware. Software applications, such asvideo games, for programmable devices can be distributed asplug-in cartridges containing ROM.

Strictly speaking,read-only memory refers to hard-wired memory, such asdiode matrix or amask ROMintegrated circuit (IC), that cannot be electronically changed after manufacture. Although discrete circuits can be altered in principle, through the addition ofbodge wires and the removal or replacement of components, ICs cannot. Correction of errors, or updates to the software, require new devices to be manufactured and to replace the installed device.

Floating-gate ROMsemiconductor memory in the form oferasable programmable read-only memory (EPROM),electrically erasable programmable read-only memory (EEPROM) andflash memory can be erased and re-programmed. But usually, this can only be done at relatively slow speeds, may require special equipment to achieve, and is typically only possible a certain number of times.^[1]

The term "ROM" is sometimes used to refer to a ROM device containing specific software or a file with software to be stored in a writable ROM device. For example, users modifying or replacing theAndroid operating system describe files containing a modified or replacement operating system as "custom ROMs" after the type of storage the file used to be written to, and they may distinguish between ROM (where software and data is stored, usuallyFlash memory) and RAM.

ROM and RAM are essential components of a computer, each serving distinct roles. RAM, or Random Access Memory, is a temporary, volatile storage medium that loses data when the system powers down. In contrast, ROM, being non-volatile, preserves its data even after the computer is switched off.^[2]

IBM used capacitor read-only storage (CROS) andtransformer read-only storage (TROS) to store microcode for the smallerSystem/360 models, the360/85, and the initial twoSystem/370 models (370/155 and370/165). On some models there was also awriteable control store (WCS) for additional diagnostics and emulation support. TheApollo Guidance Computer usedcore rope memory, programmed by threading wires through magnetic cores.

The simplest type ofsolid-state ROM is as old as thesemiconductor technology itself.Combinationallogic gates can be joined manually to mapn-bit address input onto arbitrary values ofm-bit data output (alook-up table). With the invention of theintegrated circuit camemask ROM. Mask ROM consists of a grid ofword lines (the address input) and bit lines (the data output), selectively joined withtransistor switches, and can represent an arbitrary look-up table with a regular physical layout and predictablepropagation delay. Mask ROM is programmed withphotomasks inphotolithography duringsemiconductor manufacturing. The mask defines physical features or structures that will be removed, or added in the ROM chips, and the presence or absence of these features will represent either a 1 or a 0 bit, depending on the ROM design.^[3] Thus by design, any attempts to electronically change the data will fail, since the data is defined by the presence or absence of physical features or structures that cannot be electronically changed. For every software program, even for revisions of the same program, the entire mask must be changed, which can be costly.

In mask ROM, the data is physically encoded in the circuit, so it can only be programmed during fabrication. This leads to a number of serious disadvantages:

• It is only economical to buy mask ROM in large quantities, since users must contract with afoundry to produce a custom design for every piece, or revision of software.
• The turnaround time between completing the design for a mask ROM and receiving the finished product is long, for the same reason.
• Mask ROM is impractical forR&D work since designers frequently need to quickly modify the contents of memory as they refine a design.
• If a product is shipped with faulty mask ROM, the only way to fix it is torecall the product and physically replace the ROM in every unit shipped. This has happened in the real world with a faultycarbon monoxide detector.^[4]

Subsequent developments have addressed these shortcomings.Programmable read-only memory (PROM), invented byWen Tsing Chow in 1956,^[5][6] allowed users to program its contents exactly once by physically altering its structure with the application of high-voltage pulses. This addressed problems 1 and 2 above, since a company can simply order a large batch of fresh PROM chips and program them with the desired contents at its designers' convenience.

The advent of themetal–oxide–semiconductor field-effect transistor (MOSFET), invented atBell Labs in 1959,^[7] enabled the practical use ofmetal–oxide–semiconductor (MOS) transistors asmemory cell storage elements insemiconductor memory, a function previously served bymagnetic cores incomputer memory.^{[citation needed]} In 1967,Dawon Kahng andSimon Sze of Bell Labs proposed that thefloating gate of a MOSsemiconductor device could be used for the cell of a reprogrammable ROM, which led toDov Frohman ofIntel inventingerasable programmable read-only memory (EPROM) in 1971.^[8][9] The 1971 invention of EPROM essentially solved problem 3, since EPROM (unlike PROM) can be repeatedly reset to its unprogrammed state by exposure to strong ultraviolet light.

Electrically erasable programmable read-only memory (EEPROM), developed by Yasuo Tarui, Yutaka Hayashi and Kiyoko Naga at theElectrotechnical Laboratory in 1972,^[10] went a long way to solving problem 4, since an EEPROM can be programmedin-place if the containing device provides a means to receive the program contents from an external source (for example, a personal computer via aserial cable).Flash memory, invented byFujio Masuoka atToshiba in the early 1980s and commercialized in the late 1980s, is a form of EEPROM that makes very efficient use of chip area and can be erased and reprogrammed thousands of times without damage. It permits erasure and programming of only a specific part of the device, instead of the entire device. This can be done at high speed, hence the name "flash".^[11][12]

All of these technologies improved the flexibility of ROM, but at a significant cost-per-chip, so that in large quantities mask ROM would remain an economical choice for many years. (Decreasing cost of reprogrammable devices had almost eliminated the market for mask ROM by the year 2000.) Rewriteable technologies were envisioned as replacements for mask ROM.

The most recent development isNAND flash, also invented at Toshiba. Its designers explicitly broke from past practice, stating plainly that "the aim of NAND flash is to replacehard disks,"^[13] rather than the traditional use of ROM as a form of non-volatileprimary storage. As of 2021^[update], NAND has nearly completely achieved this goal by offering throughput higher than hard disks, lower latency, higher tolerance of physical shock, extreme miniaturization (in the form ofUSB flash drives and tinymicroSDmemory cards, for example), and much lower power consumption.

Manystored-program computers use a form ofnon-volatile storage (that is, storage that retains its data when power is removed) to store the initial program that runs when the computer is powered on or otherwise begins execution (a process known^[a] asbootstrapping, often abbreviated to "booting" or "booting up"). Likewise, every non-trivial computer needs some form of mutable memory to record changes in itsstate as it executes.

Forms of read-only memory were employed as non-volatile storage for programs in most early stored-program computers, such asENIACafter 1948. (Until then it was not a stored-program computer as every program had to be manually wired into the machine, which could take days to weeks.) Read-only memory was simpler to implement since it needed only a mechanism to read stored values, and not to change them in-place, and thus could be implemented with very crude electromechanical devices (seehistorical examples below). With the advent ofintegrated circuits in the 1960s, both ROM and its mutable counterpartstatic RAM were implemented as arrays oftransistors in silicon chips; however, a ROM memory cell could be implemented using fewer transistors than an SRAM memory cell, since the latter needs alatch (comprising 5-20 transistors) to retain its contents, while a ROM cell might consist of the absence (logical 0) or presence (logical 1) of one transistor connecting a bit line to a word line.^[14] Consequently, ROM could be implemented at a lower cost-per-bit than RAM for many years.

Mosthome computers of the 1980s stored aBASIC interpreter oroperating system in ROM as other forms of non-volatile storage such asmagnetic disk drives were too costly. For example, theCommodore 64 included 64KB of RAM and 20 KB of ROM containing a BASIC interpreter and theKERNAL operating system. Later home or office computers such as theIBMPC XT often included magnetic disk drives, and larger amounts of RAM, allowing them to load their operating systems from disk into RAM, with only a minimal hardware initialization core andbootloader remaining in ROM (known as theBIOS inIBM-compatible computers). This arrangement allowed for a more complex and easily upgradeable operating system.

In modern PCs, "ROM" is used to store the basic bootstrappingfirmware for the processor, as well as the variousfirmware needed to internally control self-contained devices such asgraphic cards,hard disk drives,solid-state drives,optical disc drives,TFT screens, etc., in the system. Today, many of these "read-only" memories – especially theBIOS/UEFI – are often replaced withEEPROM orFlash memory (see below), to permit in-place reprogramming should the need for a firmware upgrade arise. However, simple and mature sub-systems (such as the keyboard or some communication controllers in the integrated circuits on the main board, for example) may employ mask ROM orOTP (one-time programmable).

ROM andsuccessor technologies such as flash are prevalent inembedded systems. These are in everything fromindustrial robots tohome appliances andconsumer electronics (MP3 players,set-top boxes, etc.) all of which are designed for specific functions, but are based on general-purposemicroprocessors. With software usually tightly coupled to hardware, program changes are rarely needed in such devices (which typically lack hard disks for reasons of cost, size, or power consumption). As of 2008, most products use Flash rather than mask ROM, and many provide some means for connecting to a PC forfirmware updates; for example, a digital audio player might be updated to support a newfile format. Some hobbyists have taken advantage of this flexibility to reprogram consumer products for new purposes; for example, theiPodLinux andOpenWrt projects have enabled users to run full-featuredLinuxdistributions on their MP3 players and wireless routers, respectively.

ROM is also useful for binary storage ofcryptographic data, as it makes them difficult to replace, which may be desirable in order to enhanceinformation security.

Since ROM (at least in hard-wired mask form) cannot be modified, it is only suitable for storing data which is not expected to need modification for the life of the device. To that end, ROM has been used in many computers to storelook-up tables for the evaluation of mathematical and logical functions (for example, afloating-point unit mighttabulate the sine function in order to facilitate faster computation). This was especially effective whenCPUs were slow and ROM was cheap compared to RAM.

Notably, thedisplay adapters of early personal computers stored tables of bitmapped font characters in ROM. This usually meant that the text displayfont could not be changed interactively. This was the case for both theCGA andMDA adapters available with the IBM PC (type 5150).^[15]

The use of ROM to store such small amounts of data has disappeared almost completely in modern general-purpose computers. However,NAND Flash has taken over a new role as a medium formass storage orsecondary storage of files.

Mask ROM is a read-only memory whose contents are programmed by theintegrated circuit manufacturer (rather than by the user). The desired memory contents are furnished by the customer to the device manufacturer. The desired data is converted into a customphotomask/mask layer for the final metallization of interconnections on the memory chip (hence the name).

Mask ROM can be made in several ways, all of which aim to change the electrical response of a transistor when it is addressed on a grid, such as:

• In a ROM with transistors in a NOR configuration, using a photomask to define only specific areas of a grid with transistors, to fill with metal thus connecting to the grid only part of all the transistors in the ROM chip^[3] thus making a grid where transistors that are connected cause a different electrical response when addressed, than spaces in the grid where the transistors are not connected, a connected transistor may represent a 1 and an unconnected one a 0, or viceversa. This is the least expensive, and fastest way of making mask ROM^[3] as it only needs one mask with data, and has the lowest density of all mask ROM types as it is done at the metallization layer,^[3] whose features can be relatively large in respect to other parts of the ROM. This is known as contact-programmed ROM. In ROM with a NAND configuration, this is known as metal-layer programming and the mask defines where to fill the areas surrounding transistors with metal which short-circuits the transistors instead, a transistor that is not short circuited may represent a 0, and one that is may represent a 1, or viceversa.^[16]
• Using two masks to define two types of ion implantation regions for transistors, to change their electrical properties when addressed in a grid and define two types of transistors.^[3] The type of transistor defines if it represents a 1 or a 0 bit. One mask defines where to deposit one type of ion implantation (the "1" transistors), and another defines where to deposit the other (the "0" transistors). This is known as voltage threshold ROM (VTROM) as the different ion implantation types define different voltage thresholds in the transistors, and it's the voltage threshold on a transistor that defines a 0, or a 1. Can be used with NAND and NOR configurations. This technique offers a high level of resistance against optical reading of the contents as ion-implantation regions are difficult to distinguish optically,^[16] which may be attempted withdecapping of the ROM and a microscope.
• Using two levels of thickness for a gate oxide in transistors,^[3] and using a mask to define where to deposit one thickness of oxide, and another mask to deposit the other. Depending on the thickness a transistor can have different electrical properties and thus represent either a 1 or a 0.
• Using several masks to define the presence or absence of the transistors themselves, on a grid. Addressing a non-existent transistor may be interpreted as a 0, and if a transistor is present it may be interpreted as a 1, or viceversa. This is known as active-layer programming.^[16]

Mask ROM transistors can be arranged in either NOR or NAND configurations and can achieve one of the smallest cell sizes possible as each bit is represented by only one transistor. NAND offers higher storage density than NOR. OR configurations are also possible, but compared to NOR it only connects transistors to V_cc instead of V_ss.^[16] Mask ROMs used to be the most inexpensive, and are the simplest semiconductor memory devices, with only one metal layer and one polysilicon layer, making it the type of semiconductor memory with the highest manufacturing yield^[3] (the highest number of working devices per manufacturing run). ROM can be made using one of several semiconductor device fabrication technologies such asCMOS,nMOS,pMOS, andbipolar transistors.^[17]

It is common practice to use rewritablenon-volatile memory – such as UV-EPROM orEEPROM – for the development phase of a project, and to switch to mask ROM when the code has been finalized. For example,Atmel microcontrollers come in both EEPROM and mask ROM formats.

The main advantage of mask ROM is its cost. Per bit, mask ROM was more compact than any other kind ofsemiconductor memory. Since the cost of anintegrated circuit strongly depends on its size, mask ROM is significantly cheaper than any other kind of semiconductor memory.

However, the one-time masking cost is high and there is a long turn-around time from design to product phase. Design errors are costly: if an error in the data or code is found, the mask ROM is useless and must be replaced in order to change the code or data.^[18]

As of 2003^[update], four companies produce most such mask ROM chips:Samsung Electronics,NEC Corporation,Oki Electric Industry, andMacronix.^{[19][needs update]}

Some integrated circuits contain only mask ROM. Other integrated circuits contain mask ROM as well as a variety of other devices. In particular, manymicroprocessors have mask ROM to store theirmicrocode. Somemicrocontrollers have mask ROM to store thebootloader or all of theirfirmware.

Classic mask-programmed ROM chips are integrated circuits that physically encode the data to be stored, and thus it is impossible to change their contents after fabrication.

It is also possible to write the contents of a Laser ROM by using a laser to alter the electrical properties of only some diodes on the ROM, or by using a laser to cut only some polysilicon links, instead of using a mask.^[20][21][16]

• Programmable read-only memory (PROM), orone-time programmable ROM (OTP), can be written to orprogrammed via a special device called aPROM programmer. Typically, this device uses high voltages to permanently destroy or create internal links (fuses orantifuses) within the chip. Consequently, a PROM can only be programmed once.
• Erasable programmable read-only memory (EPROM) can be erased by exposure to strongultraviolet light (typically for 10 minutes or longer), then rewritten with a process that again needs higher than usual voltage applied. Repeated exposure to UV light will eventually wear out an EPROM, but theendurance of most EPROM chips exceeds 1000 cycles of erasing and reprogramming. EPROM chip packages can often be identified by the prominentquartz "window" which allows UV light to enter. After programming, the window is typically covered with a label to prevent accidental erasure. Some EPROM chips are factory-erased before they are packaged, and include no window; these are effectively PROM.
• Electrically erasable programmable read-only memory (EEPROM) is based on a similar semiconductor structure to EPROM, but allows its entire contents (or selectedbanks) to be electrically erased, then rewritten electrically, so that they need not be removed from the computer (whether general-purpose or an embedded computer in a camera, MP3 player, etc.). Writing orflashing an EEPROM is much slower (milliseconds per bit) than reading from a ROM or writing to a RAM (nanoseconds in both cases).
  • Electrically alterable read-only memory (EAROM) is a type of EEPROM that can be modified one or a fewbits at a time.^[22] Writing is a very slow process and again needs higher voltage (usually around 12V) than is used for read access. EAROMs are intended for applications that require infrequent and only partial rewriting. EAROM may be used asnon-volatile storage for critical system setup information; in many applications, EAROM has been supplanted byCMOSRAM supplied bymains power and backed up with alithium battery.
  • Flash memory (or simplyflash) is a modern type of EEPROM invented in 1984. Flash memory can be erased and rewritten faster than ordinary EEPROM, and newer designs feature very high endurance (exceeding 1,000,000 cycles). ModernNAND flash makes efficient use of silicon chip area, resulting in individual ICs with a capacity as high as 32GB as of 2007^[update]; this feature, along with its endurance and physical durability, has allowed NAND flash to replacemagnetic in some applications (such asUSB flash drives).NOR flash memory is sometimes calledflash ROM orflash EEPROM when used as a replacement for older ROM types, but not in applications that take advantage of its ability to be modified quickly and frequently.

By applyingwrite protection, some types of reprogrammable ROMs may temporarily become read-only memory.

There are other types of non-volatile memory which are not based on solid-state IC technology, including:

• Optical storage media, suchCD-ROM which is read-only (analogous to masked ROM).CD-R isWrite Once Read Many (analogous to PROM), whileCD-RW supports erase-rewrite cycles (analogous to EEPROM); both are designed forbackwards-compatibility with CD-ROM.

• Diode matrix ROM, used in small amounts in many computers in the 1960s as well as electronic deskcalculators and keyboard encoders forterminals. This ROM was programmed by installing discrete semiconductor diodes at selected locations between a matrix ofword line traces andbit line traces on aprinted circuit board.
• Resistor orcapacitor matrix ROM, used in many computers until the 1970s. Like diode matrix ROM, it was programmed by placing components at selected locations between a matrix ofword lines andbit lines.ENIAC's Function Tables were resistor matrix ROM, programmed by manually setting rotary switches. Various models of theIBMSystem/360 and complex peripheral devices stored theirmicrocode in a capacitor matrix, in variants calledBCROS forbalanced capacitor read-only storage on the360/50 and360/65, orCCROS forcard capacitor read-only storage on the360/30.
• Transformer matrix ROM achieves higher density storage than diode, resistor, or capacitor matris ROMs, by using each matrix element to store multiple bits.
  • Dimond Ring Translator, named after Bell Labs inventor Thomas L. Dimond, in which wires are threaded through a sequence of large ferrite rings that function as transformers, coupling drive pulses to sense windings.^[23][24] Invented in the early 1940s, the Dimond Ring Translator was used in the #5Crossbar Switch, andTXE telephone exchanges. Dimond Ring was the basis for most later forms of transformer-coupled or "core rope" memory.
  • Transformer Read Only Storage (TROS) on the360/20,360/40 and peripheral control units), is a transformer matrix ROM technology operating in the same way as the Dimond Ring Translator. It is faster and more compact than IBM's CCROS used in theIBM System/360 Model 30, but slower than IBM's BCROS used in theIBM System/360 Model 50 andModel 65.
  • Core rope memory, also known as wire braid memory,^[25] which couples drive lines to sense lines through ferrite cores, used where size, weight, and/or cost were critical. Core rope stores multiple bits of ROM per core (unlike normal read/write core memory), and was programmed by weaving "word line wires" inside or outside offerrite transformer cores. Two different kinds of core rope memory, distinguished by whether the magnetization of the cores is flipped during operation, are known as the pulse-transformer technique and the switching-core technique^[26]
    • In the pulse-transformer technique, the drive lines are coupled to the sense lines through ferrite cores, but the core magnetization is not flipped, nor does this methoddepend on the magnetization hysteresis loop, using them only as transformers. This operates in the same way as the Dimond Ring Translator, and was used inDEC'sPDP-9 andPDP-16 computers, theHewlett-Packard 9100A and 9100B calculators,Wang calculators, and many other machines.
    • The switching-core technique does flip the magnetization of the ferrite cores. This is significantly different than the operation of a Dimond Ring Translator. This was used inNASA/MIT'sApollo Spacecraft Computers,^[27]
  • Inductively coupled printed circuit board memory, which uses inductive coupling but no ferrite cores, instead coupling between drive lines and sense lines on separate planes of a printed circuit board. This operates on the same principle as the Dimond Ring Translator, and was used in theHewlett-Packard 9100A and 9100B calculators for the main control store (in addition to a pulse-transformer core rope memory used for the microinstruction decoder).^[28]

Although the relative speed of RAM vs. ROM has varied over time, as of 2007^[update] large RAM chips can be read faster than most ROMs. For this reason (and to allow uniform access), ROM content is sometimes copied to RAM orshadowed before its first use, and subsequently read from RAM.

For those types of ROM that can be electrically modified, writing speed has traditionally been much slower than reading speed, and it may need unusually high voltage, the movement of jumper plugs to apply write-enable signals, and special lock/unlock command codes. Modern NAND Flash can be used to achieve the highest write speeds of any rewritable ROM technology, with speeds as high as 10GB/s in an SSD. This has been enabled by the increased investment in both consumer and enterprise solid-state drives and flash memory products for higher end mobile devices. On a technical level the gains have been achieved by increasing parallelism both in controller design and of storage, the use of large DRAM read/write caches and the implementation of memory cells which can store more than one bit (DLC, TLC and MLC). The latter approach is more failure prone but this has been largely mitigated by overprovisioning (the inclusion of spare capacity in a product which is visible only to the drive controller) and by increasingly sophisticated read/write algorithms in drive firmware.

Because they are written by forcing electrons through a layer ofelectrical insulation onto afloating transistor gate, rewriteable ROMs can withstand only a limited number of write and erase cycles before the insulation is permanently damaged. In the earliest EPROMs, this might occur after as few as 1,000 write cycles, while in modern Flash EEPROM the endurance may exceed 1,000,000. The limited endurance, as well as the higher cost per bit, means that Flash-based storage is unlikely to completely supplant magneticdisk drives in the near future.^{[citation needed]}

The timespan over which a ROM remains accurately readable is not limited by write cycling. The data retention of EPROM, EAROM, EEPROM, and Flash may be time-limited by charge leaking from thefloating gates of the memory cell transistors. Early generation EEPROM's, in the mid-1980s generally cited 5 or 6 year data retention. A review of EEPROM's offered in the year 2020 shows manufacturers citing 100 year data retention. Adverse environments will reduce the retention time (leakage is accelerated by high temperatures orradiation). Masked ROM and fuse/antifuse PROM do not suffer from this effect, as their data retention depends on physical rather than electrical permanence of the integrated circuit, although fuse re-growth was once a problem in some systems.^[29]

The contents of ROM chips can be extracted with special hardware devices and relevant controlling software. This practice is common for, as a main example, reading the contents of older videogame consolecartridges. Another example is making backups of firmware/OS ROMs from older computers or other devices - for archival purposes, as in many cases, the original chips are PROMs and thus at risk of exceeding their usable data lifetime.

The resultant memory dump files are known asROM images or abbreviatedROMs, and can be used to produce duplicate ROMs - for example to produce new cartridges or as digital files for playing inconsole emulators. The termROM image originated when most console games were distributed on cartridges containing ROM chips, but achieved such widespread usage that it is still applied to images of newer games distributed onCD-ROMs or other optical media.

ROM images of commercial games, firmware, etc. usually contain copyrighted software. The unauthorized copying and distribution of copyrighted software is a violation ofcopyright laws in many jurisdictions, although duplication forbackup purposes may be consideredfair use depending on location. In any case, there is a thriving community engaged in the distribution and trading of such software for preservation/sharing purposes.

• Flash memory
• Random-access memory
• Read-mostly memory (RMM)
• Write-only memory