0-9, A - B

0-9

• 16-, 24- and 32-bit colour (pixel format)
  Monitors use 3 different coloured pixels (red, green and blue or RGB) to produce an image and graphics adapters do the same, using 3 or 4 colour channels.
  16-bit colour rendering normally uses one of two modes – (1) 5-bits for the red and blue channels and 6-bits for the green channel or (2) 5-bits for the red, green and blue channel with 1-bit for the alpha channel. Since 5-bits gives 32 levels of colour, all 3 channels together offers up to 32,768 colours with 256 alpha levels or up to 65,536 colours with no alpha. Other modes are possible (depending on the video adapter) such as just using 4-bits for all 4 channels.
  24-bit and 32-bit colour rendering uses 8-bits for the RGB channels (and therefore offers no more colours than the others do) but you have 8-bits for the alpha channel to play with in 32-bit mode. Many cards offer other pixel formats but most games stick with the standard 16- or 32-bit rendering modes.
  So why the need for more bits per channel? Contrary to popular belief, it is not about how many colours the bits give you (although this does have some importance). The reason is about colour gradients: imagine a scene that is nothing but blue. The top of the scene is very pale and the hue deepens as you go further down the screen. The number of bits per channel restricts the number of steps that can be used in this colour gradient. 5-bit channels only give you 32 steps, making the gradient very banded. 8-bit channels are better with 256 steps but once you start going higher than 10-bits, the gradient steps become much finer.
  
• 64-, 128- and 256-bit graphics adapters
  The processing chips on graphics adapters are sometimes classified by the amount of pixel data processed per clock cycle, so a 128-bit graphics adapter processes 128-bits in one clock cycle. Most graphics chips have multiple pipelines, using parallel processing to get through so much data.
  A few companies also classify the graphics adapters by other means: memory interface, vertex shaders, etc but in general, it can be assumed that as the "bits increases", so does the potential of the chip (i.e. a 128-bit chip is "better" than a 64-bit one).
  
• 8-, 16- and 24-bit sound card
  The number of bits that a sound card is used to be categorised by often varies from source to source. It can mean the number of bits used in the digital-to-analogue conversion of any FM or wavetable sound; it can also refer to the internal "width" of processing within the audio adapter. Either way, more bits equals better quality sound.
  
• 2D card
  Somewhat rare these days, but a 2D card is a video adapter that offers no hardware acceleration for 3D graphics rendering. An example of such a device would be the ye-olde 16 colour graphics adapters from the 1980s!
  
• 3D card (3D accelerator)
  See Acceleration. A video card that offers hardware acceleration for the rendering of 3D graphics (typically polygon-based); virtually all modern video adapters combine 2D and 3D acceleration on the same chip.
  
• 3D glasses
  Some graphics adapters can simulate a "true" 3D image by a special rendering technique and the use of "3D glasses". Previous attempts of using red and green filters over each eye have been superseded by ones that have shutters, which alternately rapidly open and close.
  
• 3DNow!
  AMD's (Advanced Micro Devices) trademark name for the set of additional registers and extensions to the x86 instruction set in their CPUs. 3DNow! is an example of SIMD (single instruction, multiple data). Applications that are floating-point intensive, such as 3D games, can make use of the extra instructions to gain an increase in performance.
  
• 3D Sound
  Unlike a 3D game, which is really a 2D frame that is drawn in such a way to give an impression of 3-dimensions, true 3D sound has genuine depth and width to it. Using additional channels (and obviously speakers too), the listener is surrounded by sound or music, seemingly coming from all directions.
  However, just like 3D graphics, 3D sound can also be recreated by just two speakers and some very clever mathematics! These methods are clearly not as good as the real thing but it can successfully produce basic effects such as stereo panning.
  
  

  A

• A3D
  This is an audio API initially created by Aureal, but now owned by Creative Technologies. It was a popular rival to DirectSound and in the space of two revisions boasted an impressive array of 3D sound features. The development of the API appears to have halted, so it is likely to become less popular with game developers in the future.
  
• AC-3
  Audio Coding 3. This another name for Dolby Digital, a technique used to code audio into a digital format but requiring smaller amounts of data to do so. AC-3 is a popular format for digital media as it offers 6 channels of audio - 5 surround lines and 1 low frequency line.
  
• Acceleration
  If a piece of hardware can perform specific functions that would otherwise be done by the CPU, it is said to accelerate that function. Typically, such a function would be done faster by the graphics adapter than the CPU, as the former is custom-designed to do that job. An example of this would be the transformation of vertices where the graphics adapter performs the calculations and not the CPU.
  
• ACPI
  Advanced Configuration and Power Interface. It is an industry-developed standard that specifies how the operating system manages the power and operations of the various devices in the PC. An official site for the technology exists: http://www.acpi.info
  
• Accumulation buffer (A-buffer)
  This is a type of memory buffer that stores multiple frames together without them being visible. It works by having the graphics chip rendering a frame to the back buffer and then passing the pixel values onto the accumulation buffer; for each frame that is to be included in the buffer, the pixel values from the successive back buffers are "accumulated" into this special buffer. The contents can then be sampled or adjusted in some way to produce certain effects, such as temporal anti-aliasing, motion blur, and depth of field.
  
• ADC
  Analogue-to-digital converter. A device that reads a varying, analogue voltage and converts it into a sequence of digital values. Such devices are found on sound cards (to record audio) and some graphics cards (to record video signals). ADCs are sometimes classified by the number of bits used to sample the analogue signal - more bits produce a more accurate representation of the signal.
  
• AGP
  Accelerated Graphics Port. AGP is a PC bus architecture developed to supplement PCI. It is a 32-bit bus, just like PCI, but runs quite differently. Data can be transferred through the port at different rates, depending on what version of AGP is being used and what motherboard settings have been applied.
  • AGP 1.0 has two different operating modes: 1x and 2x. In the first mode, it operates like a fast PCI bus, transferring data at a rate of 66MHz, while in 2x mode the data transfer rate is 133MHz. It does this by timing when to move data about using the periods when the clock signal is changing, either when the signal is rising or falling. Since it does this twice in one normal clock "tick", 2x mode times the transfers at twice the rate too.
    
  • AGP 2.0 also has 1x and 2x transfer modes, but in addition, it allows 4x mode producing a transfer rate of 266MHz. This faster transfer rate requires a more complex signalling system than in 2x mode - this time, two sets of signals are used together and the points where they crossover each other time the transfers.
    
  • AGP 3.0 is the latest version of the specification. It is similar to AGP 2.0 but it also offers an 8x mode, giving a transfer rate of 533MHz, although this requires additional hardware support from the motherboard and graphics card.
  These faster transfer speeds lets the system move more data around per clock cycle (i.e. more bandwidth) but doubling the AGP clock speeds is not the same as doubling the speed of a processor. The faster modes also require more accurate voltage tolerances, especially with 4x transfers.
  As well as having faster transfer rates than the PCI bus, there are some additional tricks that the AGP bus can play - sideband addressing, fast writes, special arbitration methods, etc - to help boost the operating performance. An official site for the technology exists: http://www.agpforum.org
  
• AGP Aperture
  See AGP Memory. The AGP Aperture is actually the correct name for the AGP Memory, as laid down in the specifications but most people use the term to refer to a setting within the motherboard BIOS. This setting tells the PC the maximum amount of the system's memory, including virtual memory, which can be allocated to be used as AGP memory. Most motherboards default to 64MB, but many allow you to change it from 4MB to 256MB. The minimum size has to be 12MB (see AGP Memory for the reason why) but once above this, the actual value does not have a great impact on the overall rendering performance if your graphics adapter has 32MB or more of local memory. Contrary to popular belief, there is no formula to work out what setting you should use but many people will tell you that a good starting place is to keep it at half the value of your system memory (e.g. 256MB of system RAM, use an AGP Aperture Size of 128MB).
  
• AGP Fast Writes
  A feature of the AGP 2.0 specification. When not using Fast Writes, the CPU sends all the video data to the system RAM and then notifies the Northbridge chip and graphics adapter that it is available for use. The graphics adapter then instructs the Northbridge chip that it is ready for this data and copies across the relevant material into the local memory. If Fast Write transactions are enabled, the CPU simply writes data directly into the local memory without caching it in the system RAM.
  The advantage behind this is that if the bandwidth of the system RAM is particularly weak, then potential additional speed can be gained by not having to constantly write-read data to and from the system RAM. The disadvantage is that very tight signalling tolerances are required by all the components for fast write transactions to run successfully. At the moment, only a handful of graphics chips support the use of AGP Fast Writes and many manufacturers choose to disable the function in the video BIOS to prevent any potential problems occurring.
  
• AGP Memory (AGP Aperture)
  A feature of all AGP specifications. An AGP graphics adapter can use part of the system memory to store data and/or directly render from. Once it has been set by the operating system, the amount of memory allocated to the use of AGP memory is then dynamic and changes as according to use by the graphics adapter and application using it. The maximum size of memory that can be "sidelined" is determined by the AGP Aperture value in the motherboard BIOS.
  So why not just use the normal system memory? Pages that are allocated to AGP memory cannot be paged to the hard drive (swapped into the virtual memory/swap file) and the addresses for each page are stored in a table that can be accessed very easily by the Northbridge chip. All of this helps to increase the performance of read/writes of data stored in the AGP memory.
  The graphics adapter can access the contents of the AGP memory by several methods:
  • DMA (Direct Memory Access): the graphics adapter is not permitted to render directly from the AGP memory but it can take data from it as and when it needs to. This is currently the standard mode used as virtually all graphics cards have sufficient local memory to store the data it needs.
  • DIME (Direct Memory Execute): the graphics adapter must always render directly from the AGP memory in this mode. Considering that using the local memory is far faster than the AGP memory for rendering, this is only used with onboard graphics adapters that are forced (by design) to use system memory instead of having their own local memory.
  • DIMEL (Direct Memory Execute and Local): the graphics adapter can be made to render from either the local or AGP memory. The idea behind this mode is that data that is not going to be needed very often can be left in the AGP memory, whereas frequently accessed data is kept in the local memory. In reality, it is limited in use as graphics cards today can read/write data across the local memory bus so much faster than across the AGP bus that the AGP memory becomes a hindrance rather than a performance aid.
  For a particular mode to be used, both the graphics adapter and system (operating system, Northbridge, etc) must support it.
  
• AGP Read Synchronisation
  This setting forces the Northbridge chip to flush all data caches to ensure that all read/writes across the AGP bus are synchronous with the system memory transfers. Such a setting may not be desirable for benchmarking or 3D games as it means that no data is ever queued for the graphics adapter - only one frame is prepared and then processed at a time, potentially wasting many CPU cycles.
  
• AGP Sideband Addressing (SBA)
  A feature of all AGP specifications. In normal usage, all instructions and data between the graphics adapter and system are transferred via the primary AGP bus. However, the bus cannot be used to send and receive at the same time - if the graphics adapter needs more data, it must wait until it has finished receiving the last lot before requesting some more.
  The AGP Sideband bus is an additional bus that runs parallel to the main AGP one. Its primary function is to allow the graphics adapter to request more instructions, address locations, etc while still collecting data down the main line. The speed of the sideband bus is the same as the AGP bus (1x, 2x, etc), although it is much "narrower" being only 8-bits wide. Although it takes at least two cycles of the SBA clock for an instruction to be transferred (due to the narrow width of the bus), the graphics adapter never has to request ownership of it before making any such requests - it has full use of the bus at all times.
  Even though the sideband bus offers a potential performance increase, it is not a particularly large one by modern day standards. Several manufacturers choose not to enable the use of SBA - offsetting the small loss in speed, with a gain in overall stability.
  
• AGP Pro
  This is a version of the standard AGP slot that has additional voltage pins, in order to offer extra power for those graphics cards that need it. The standard AGP slot currently permits a maximum power consumption of around 45~50W via the usual pins - the longer slot and extra pins of the AGP Pro slot allows up to 110W of power! The only graphics cards that need that amount of operating current are the rendering beasts that live in the workstations of the 3D professional.
  
• AI
  Short for Artificial Intelligence. When used in conjunction with 3D games, it usually refers to the way that characters react to the environment, player, etc. It can also refer to the game environment itself, such the movements and reactions of a group of NPCs. Good AI is computationally expensive, often beyond the means of a typical home computer, so short-cuts are often employed such as having a limited range of reactions. When done well, it can make a big difference to a game.
  
• Algorithm
  This is a mathematical procedure that is used to take data, work through a sequence of steps, and give a required answer. An example of one can be found with texture filtering; the graphics chip takes various samples from a texture map and then uses an algorithm to blend the values of these samples together, producing a final, single number.
  
• Aliasing
  See Anti-aliasing. A term mostly used to describe the jagged look to the edges of objects in a 3D scene, but it also applies to textures and simple lines. The jagged look is caused by a combination of low screen resolution and the sudden change in pixel colour at the edge of an object. Texture aliasing is caused by the textures themselves not being sampled correctly (either the pattern is not suitable for the conditions or not enough samples are being taken). There are various methods used to counter all of these issues; collectively they are called anti-aliasing techniques.
  
• Alpha (value, component, channel)
  The "fourth" colour of a pixel. The alpha value describes how transparent the pixel is - many textures often have a 1-bit alpha channel, as this is all that is needed for a "see-through" hole. Materials such as glass or water require a higher level of accuracy (i.e. more bits) to blend the colours of the pixels behind the object correctly.
  
• Alpha Blending
  See Alpha. This is the process where the pixels of a transparent object are blended with those behind it, in such a way that the final scene allows the viewer to look through the object (e.g. a glass window). Alpha blending is typically done last in the rendering pipeline, as it requires all the objects in the scene to be drawn with the further object done first.
  
• Alpha Testing
  This is a process that can be used in the rendering pipeline to decide whether a pixel is declared visible, and therefore written to the frame buffer (or any other target surface).
  
• Alpha textures
  These are texture maps that contain transparent or semi-transparent pixels; therefore, the surface format must contain an alpha channel. A popular use for them in games is for mapping leaves or blades of grass; unless the texture is of a high resolution though, the visual effect decreases rapidly in quality as the camera approaches the map.
  
• ALU
  Arithmetic Logic Unit. Part of a microprocessor that performs mathematical operations and instructions; in a CPU they work on integer values and operations, whereas the FPU works on floating point values and calculations. For a graphics chip, it is a little different as the ALUs perform all the operations, with units specific to vertex and pixel processing.
  
• Ambient Lighting
  See Shading. An object that is being viewed under ambient lighting has the same light intensity all over it - a process called flat shading calculates this per vertex and then applies the same value over each pixel in the object.
  
• Anisotropic Texture Filtering
  See Texture Filtering. This method of texture filtering is better than linear filtering and can be used with or without mipmaps, as it takes into account of the angle at which the texture map is being viewed. The anisotropy of a texture map is a measure of how "angled" it appears to the viewer - higher values of anisotropy require either more texture samples to be taken, a different sampling pattern to be used or possibly both.
  
• Anti-aliasing
  See Aliasing. Anti-aliasing is a single or multiple method of removing the noticeable effects of aliasing. These days it often just is called FSAA (Full Scene Anti-Aliasing) and various manufacturers have their own way of performing anti-aliasing. Texture aliasing is reduced by the process of texture filtering but for edges and lines, the two most popular methods are:
  • Supersampling: there are various implementations of this method. One is where the graphics adapter renders the entire scene at an increased resolution into a non-visible buffer in the local memory. The adapter then samples this frame into the back buffer, and once complete it is flipped to the front buffer. Another method is to render several copies of the same frame over several passes, adjusting the geometry of each successive copy, and then sampling each stored frame into the back buffer. Both methods require lots of fill rate and memory bandwidth to be effective.
  • Multisampling: the DirectX version of this anti-aliasing method is similar to the OpenGL multi-sampling anti-aliasing, but the term is used in various ways by different companies. Although it is described as being an example of FSAA, multisampling only affects the edges of the polygons within a frame; no texture detail is affected (unlike supersampling). This is done by testing the frame before it is textured; each pixel is checked and those at a visible polygon edge are then sampled into a new buffer. Once complete, the frame is updated within the pass by blended all of the samples from the various buffers into the back buffer. Multisampling requires lots of memory bandwidth to be effective but since pixel colours and operations are only performed once, this method is not as dependent on fill rate.
  There are several other methods too, which each one having its good and bad points. However, for the time being, any form of anti-aliasing has some impact on the rendering performance of the graphics adapter.
  
• API
  Application Programming Interface. An API is the initial interface between an application (e.g. 3DMark) and the hardware. It basically "talks" to the device drivers, telling them what the program wants it to do. Over the years, there have been countless APIs for all sorts of software and hardware; but for the graphics in PC games today, they are made using only one of two: DirectX and OpenGL.
  
• Architecture
  The in-depth design of a chip or system.
  
• Artifact
  This is a term used to describe an image on the screen that should not otherwise be there - it could be a single green patch or a series of white dots, or just misaligned textures. Such faults are often caused by overclocking or software bugs.
  
• Assembler
  An assembler is a low-level program that converts instructions from another program into the fundamental patterns of bits, called machine code, to be processed by the appropriate hardware.
  
• ATA
  AT Attachment. This is a interface design that was primarily used to connect hard drives, but it is now employed with many other storage devices such as CD-ROMs. Although the original design has been superseded by the likes of ATAPI and UDMA, they are all still based off the first ATA format.
  
• ATAPI
  AT Attachment Packet Interface. The original ATA interface was only used with hard drives to begin with but as its popularity rapidly increased, it was expanded to include storage devices such as CD-ROMs or tape backups. To do this, ATAPI incorporated a new set of instructions/commands were designed that allowed the interface to correctly control these additional devices.
  
• Audio
  Another term for sound! Games would be very dull without it - graphics often grab the attention first and therefore all the glory, but well crafted audio can make a big difference.
  
• Audio Adapter
  This is the device (or devices) that are responsible for producing audio. A modern audio adapter will sport many functions; offering recording abilities, a large range of 3D sound features and support for a wide range of audio formats.
  
• Audio Card
  Another term for sound card.
  
  

  B

• B-spline (Basis spline) Curves
  Shapes that use vertices (control points) to define the curvature of specific parts of the objects. They are a generalised form of Bézier curves but permit a much greater variety of shapes and control over them. The complicated mathematics involved often rules them out for current 3D games.
  
• Back Buffer
  This is a section of the local memory that a graphics chip will usually render a frame into. When the chip has finished and completed the frame, the contents of the back buffer are swapped to the front buffer so that it can be displayed.
  
• Back Face Culling
  See Culling. The back face of a primitive always faces away from the viewer (unless it is being reflected in a mirror!), so it is not going to be visible. The graphics adapter is just going to be wasting performance by rendering things that are non-visible, so this method of culling removes the back faces from the rendering process before they are rasterised and textured.
  
• Banding
  This is a term to describe the noticeable "steps" seen in area where one colour blends into another one. For example, a bright blue sky might appear to have distinct levels as the dark blue fades in a pale blue. This effect is typically caused by the use of low colour depths (i.e. not enough bits for each colour) or poor blending, but it can be reduced by dithering or using more bits for the colours.
  
• Bandwidth
  A measure of how much data per second can flow between two devices. The theoretical maximum of any bus can be calculated by multiplying the width of the bus (in bits) by the rate at which data transfers can take place.
  
• Bézier Curves
  Shapes that use vertices (control points) to define the curvature of the object. Although they are not really used for 3D games (due to the more complex mathematics involved), you can see their use in the drawing functions of most word processors.
  
• Bézier Patches
  See Bézier Curves. An expansion on the Bézier curves, whereby the shape is defined by a greater number of functions and control points. They can be applied to a simple polygon mesh to increase the number of vertices, and increase the detail at the same time - "patches" are an easier way of increasing the polygon count without having to send lots of data to the graphics adapter.
  
• BGA
  Ball Grid Array. The name for the way in which a chip is attached to a socket or circuit board. The physical connections are little balls of solder - the small size reduces electrical resistance and capacitance, helping to retain signalling integrity.
  
• Bilinear Filtering
  See Texture Filtering. This method of texture filtering is probably the most popular as the blending algorithm is quick and easy. Bilinear filtering only takes samples from one texture map and so does not use mipmaps.
  
• Billboard
  The name given to a quadrilateral polygon that is most commonly used (when textured) for effects such as lens halos and coronas, flames, smoke, clouds, etc. It gets the name because the same face of the polygon is always displayed towards you, even if you move around - just like a person holding a billboard for your attention.
  
• Bit
  The smallest size of piece of information can be, in terms of computing, is equal to 1 bit or 1b for short. Since computers are binary, bits "hold" two states of information so the size scales geometrically - i.e. 2, 4, 8, 16, 32, 64, and so on
  
• Bit rate
  This means "the number of bits per unit of time" - for example, bits per second. A larger value means more data being transferred in one unit of time.
  
• Bitmap
  A 2D image used, for example, as a texture map.
  
• Blending
  The process of combining multiple values for something (e.g. pixel colour) in such a way that a single value is produced. An example of this is the multiplication of a lightmap and a base texture map - a pixel value is taken from each map and blended together.
  
• Blinn Shading
  See Shading. This is a method of vertex lighting similar to Phong shading, which calculates the diffuse and specular colour value of each pixel on a surface.
  
• Boolean
  This term is normally used to refer to the system of logic/algebraic processes developed by George Boole, during the 19th century. The most well-known examples of Boolean are the AND, OR and NOT operators; computers use logic gates within their processors to carry out the Boolean instructions.
  
• Botmatch
  A type of game where you play against computer-controlled, rather than real-life, opponents. The difference between a bot and a normal computer "enemy" is that a bot is designed to take the place, and therefore mimic, of a human player.
  
• Bottleneck
  A term commonly used to describe something that "holds back" the performance of a PC. For example, in a system that has the very latest graphics adapter but only a 500MHz CPU, the processor would be the bottleneck.
  
• Branching
  Virtually all programs undergo branching as they run. This is when they alter what instructions they issue as according to the results from previous instructions. Branching causes problems for modern pipelined processors; a pipeline can be full of instructions for one type branch but then receive a totally different branch on the next cycle, forcing the processor to clear its pipeline and restart again. Branch prediction can help to reduce this problem.
  
• Branch Prediction
  See Branching. Modern CPUs perform branch prediction to increase the performance of pipelining. The processor runs logical processes to predict how a program will branch and issue instructions, helping to reduce the number of times the pipeline has to be cleared and refilled. Branch prediction is only useful, of course, when the prediction turns out to be correct but at least a failed prediction causes no performance loss compared to the non-usage of predictive branching.
  
• Bucket Rendering
  Another name for tile-based rendering. This is a rendering process whereby a frame is processed in cubic or rectangular sections called "buckets" or "tiles". The term appears more often in professional systems such as Pixar's RenderMan.
  
• Buffer
  This is a piece of memory that is used to temporarily store data. In the world of graphics adapters, there seems to be a huge list of them! Such examples are frame buffers, vertex buffers, texture buffers, depth buffers, stencil buffers, index buffers, execute buffers and so on. They are all just "memory" but their uses are as varied as their names.
  
• Bump Mapping
  This is a method where the flat image of a texture map is altered to give the impression of being non-smooth but without changing the physical shape of the surface. Bump mapping is becoming more popular in 3D games because it can be used to add small scale detail (such as wrinkles, creases, scuff marks, etc) without the need to add extra polygons.
  It does this by using another texture map (sometimes called a bump map) to change the brightness of the pixel in the main texture on the surface. There are various methods of bump mapping - 3 such popular ones are:
  • Emboss: the bump map is called a height map - it is a simple greyscale version of the texture to be bumped, whereby the texel values are used to alter the colour of the original pixel. The height map is rendered twice, with the second one moved slightly. The two maps are then combined together, in such a way that the resulting new bump map is the difference between the two. This is then multiplied with the required base texture.
  • DOT Product 3: the bump map is called a normal map - this texture contains values that alter the surface normals of the original pixels. A surface normal is like an arrow that points in the direction that the surface is facing in; as it angles away from a light source, the colour of the pixel gets darker (and vice versa). This method generally gives better results than embossing but at a greater cost to performance.
    Direct3D offers a "set" dot 3 bump mapping routine but a more general method exists too; normal mapping basically uses the same process but it is done via a programmable pixel shader.
  • Environmental: this method uses a mixture of techniques, embossing, and environment mapping. A height map is created (as described above) and then blended with the original base texture; this is then followed by blending this result with an environment map (which can be anything from a simple lightmap to a complex cubic map) to give the final bumped surface.
  It should be noted that bump mapping never alters the position or amount of vertices in the surface. Edge-on views and shadows can spoil the illusion but more complex bump mapping procedures can help to reduce this problem.
  
• Burst
  A term used to describe a method of data transfers, whereby only one address/command is issued, but several cycles of data follow. Burst transfers are more efficient because less clock ticks are wasted getting control of the bus.
  
• Bus
  A collection of signalling paths between devices, used to transfer instructions, addresses and data.
  
• Byte
  8 bits is equal to 1 byte or 1B for short. Note the difference between the symbols for bits and bytes - b and B.