Matrox Genesis High-performance imaging with speeds of up to 100 billion operations per second (BOPs). Main Board Key features Key features long PCI board integrating acquisition, processing and display flexible acquisition, real-time processing and Grab Module acquires from virtually any video device: color/monochrome, high-resolution display all on one Main Board analog/digital, frame scan/line scan at up to 140 MHz sampling rate scalable processing - add more power by incorporates one processing node with: cascading Processor Boards as required for - Texas Instruments' TMS320C80 DSP up to 100 BOPs performance - Matrox Neighborhood Operations Accelerator (NOA) user configurable multi-processing - 64 MB SDRAM memory (parallel or pipeline or a combination) 1 - 64-bit bus @ 400 MB/s high-throughput I/O between acquisition, - Matrox's Video Interface ASIC (VIA) processing and display sections; between display resolutions up to 1600 x 1200 @ 85Hz multiple boards and to external resources; true-color image display with non-destructive pseudo-color overlay handled by custom video interface ASIC (VIA) VIA offloads data management tasks from the Processor Board Key features vision processors allowing them to be completely long PCI board containing one or two processing nodes dedicated to processing used with Matrox Genesis Main Board for increased processing performance available software includes Matrox Imaging up to six Processor Boards can be used in one system Library (MIL)/ActiveMIL, Matrox Genesis Native ™ interfaces directly to Main Board over the Grab Port Interface, VMChannel Library (GNL), Matrox Genesis Developer's Toolkit and the PCI bus and Matrox Inspector can also be used as a powerful stand-alone processor board ® ® support for Microsoft Windows NT 4.0, ® ® ® Windows 2000, Windows XP and QNX (GNL only) Software Key features the most comprehensive software available for a vision processor board extensive set of pre-built, high-level commands for image processing, pattern matching and blob analysis over 300 imaging operations are available, all fully optimized for accelerated processing by the C80/NOA combination Matrox Genesis Native Library Developer's Toolkit provides seamless integration of custom C80 code Matrox Imaging Library (MIL) has a consistent API across Matrox's line of imaging hardware The most comprehensive software available Matrox Genesis for a C80-based board State-of-the-art hardware components Matrox has extensive experience in providing software tools to meet Matrox Genesis vision processor is a unique design based on the the needs of developers in scientific and industrial imaging. We know PCI platform that uses Texas Instruments' TMS320C80 DSP. Our design what functions are required to solve machine vision, medical imaging exploits the full power of the C80 multi-processor DSP, the PCI bus, and image analysis applications. Matrox Imaging Library (MIL) is a SDRAM memory, the MGA 2064W graphics accelerator and WRAM memory. comprehensive image processing library that includes ActiveMIL,a Also onboard Matrox Genesis is a powerful ASIC dedicated to accelerating collection of ActiveX controls (OCXs). Access to the board is also neighborhood operations. The advanced Grab Module can handle provided through the board-specific Matrox Genesis Native Library virtually any monochrome or color video stream. A key component in (GNL). All library functions have been fully optimized to exploit Matrox Genesis' design that links everything together is the Video the full processing power and functionality of the hardware. Interface ASIC. VIA: at the heart of Matrox Genesis Custom designed by Matrox, the VIA (Video Interface ASIC) is an intelligent controller that performs optimized transfers between acquisition, display and processing sections in the Matrox Genesis system, as well as to and from external resources. It is also a powerful data formatting engine and is responsible for device synchronization. The Main Board and the Processor Board each have two VIAs The VIAs let Matrox Genesis sustain high throughput I/O and allow the C80/NOA to be completely dedicated to processing, since the VIA off-loads all data management tasks. Main Board 2 Processor Board Interfaces to virtually any camera or input device Acquisition Matrox Genesis provides exceptional flexibility for Key features interfacing to a wide variety of devices, including analog acquisition at up to 140 MHz high-resolution cameras (color or monochrome) digital acquisition up to 32-bit wide line scan cameras including TDI (color or monochrome) TTL at up to 30 MHz, RS-422 at up to 25 MHz and LVDS at up to 40 MHz medical scanners (ultrasound, CT, NM, MRI, X ray, etc.) frame scan/line scan video input up to 64K pixels per line and 64K lines slow-scan devices (e.g., SEM) grabs color or monochrome video frame reset cameras simultaneous capture of up to 4 video streams single or multi-tap cameras (time multiplexed or parallel streams) grabbed data written simultaneously to display and processing memory custom-designed input devices grabbed data broadcast to multiple processing nodes simultaneously (multi-board configuration) standard monochrome (RS-170/CCIR) programmable acquisition synchronization standard component RGB cameras external trigger input (TTL/RS-422/LVDS) future video devices such as faster or higher resolution cameras exposure (timer) outputs (TTL/RS-422/LVDS) 3 Designed specifically for industrial, medical Processing or scientific applications Key features Matrox Genesis was built to handle high-speed machine vision that processing is performed by TI's TMS320C80 (C80) and the requires interfacing to double-speed progressive-scan cameras with Matrox-designed Neighborhood Operations Accelerator ASIC (NOA) single or dual video outputs. It also supports asynchronous reset C80 is capable of accelerating the full range of operations used in capture with optional exposure control. For continuous web inspection, imaging applications, for example: point-to-point, neighborhood, such as in the printing, wood, textile and steel industries, Matrox statistical, geometric, blob analysis and pattern matching operations Genesis interfaces to both color and monochrome line scan cameras NOA further accelerates neighborhood operations with very long line lengths (e.g., grab four streams of 64K lines) and can handle variable line rates. For medical imaging, Matrox Genesis TI's C80 key features: acquires high rate analog or digital video at up to 140 MHz (e.g., MRI, one 32-bit RISC Master Processor (MP) with integral FPU CT, NM, etc.). For applications such as semiconductor inspection and biomedical analysis, the board supports interfacing to devices such four 32-bit integer Advanced DSPs (parallel processors or PPs) as scanning electron microscopes. 32 KBytes of internal RAM shared among processors (50 KBytes total) Camera interface support crossbar for optimal internal connectivity The Grab Module interfaces to a variety of standard/non-standard video cameras and devices. For a more extensive list of supported transfer controller for high performance external I/O cameras, please refer to the Matrox Imaging web site, 50 MHz system clock www.matrox.com/imaging. Matrox provides digitizer configuration internal FPU capable of 100 MFLOPS files (DCFs) and detailed application notes for many cameras. up to 2 billion RISC-like operations per second For interfacing to other cameras, developers can use Matrox Intellicam 2.4 GBytes sustainable on-chip data transfer rate camera configuration software to buildor modify DCFs. 400 MB/s off chip peak transfer rate Analog video monochrome Matrox's NOA key features: Matrox Genesis Grab Module has 4 A/D converters and can be configured Performs: to capture standard or non-standard monochrome video signals in - convolutions 3 software-selectable modes: - 4 channel/8-bit mode at up to 35 MHz - gray-scale morphology - 2 channel/8-bit mode at up to 70 MHz - binary morphology - 1 channel/8-bit mode at up to 140 MHz - normalized gray-scale correlations - lossless JPEG codec For all modes, each A/D converter is software programmable for coarse and fine adjustment of level and reference voltages as well as clamping. consists of a MAC (multiplier/accumulator) array capable of performing 32 simultaneous sums of products at 50 MHz color accelerates operations up to 20 times compared to C80 In the 4 channel/8-bit mode, Matrox Genesis Grab Module can capture standard component RGB or non-standard component RGB color video Processing nodes signals. In addition to an RGB camera, which requires 3 input channels, Processing nodes are the basic building blocks of Matrox Genesis' a separate analog stream for sync input could use the fourth channel processing power. The Main Board has one processing node while the or a single-readout monochrome camera may be connected to the fourth Processor Board has one or two processing nodes. A node contains input channel. Both the RGB and monochrome cameras, however, the following: must be externally synchronized. The video signals are simultaneously - C80 DSP digitized to 8-bits at a sampling rate of up to 35 MHz. - Neighborhood Operations Accelerator (NOA) ASIC Digital video data - 64 MB SDRAM The Grab Module can acquire 32-bit digital video data at a rate of up to - Video Interface ASIC (VIA) 40 MHz in one of several configurations: four 8-bit channels, two 16-bit 1 The processing node's local bus transfers data at up to 400 MB/s channels or one 32-bit channel. An optional companion board is required between the C80/NOA and the local SDRAM memory. The VIA manages for digital video data acquisition in RS-422/LVDS format. these transfers and transfers to/from acquisition, display and other processing nodes. 4 TMS320C80 PP0-3 : Parallel processors 0-3 (advanced DSP, 32-bit integer units) MP : Master processor (32-bit RISC processor with an IEEE-754 FPU) FPU : Floating-point unit TC : Transfer controller (transfers data between external and internal memory) RAM : On-chip memory Crossbar Network: High-speed bus switching network between processors and RAM Scalable processing: add multiple processing nodes Configurable multi-processing: parallel or pipeline by adding processor board(s) processing configure a system with a Main Board and a Processor Board for a total developer has complete control to optimally configure the of two to three nodes processing across multiple nodes up to a maximum of 6 Processor Boards can be added for more power multiple nodes can be used in parallel, either using the SIMD (single instruction, multiple data) model or the MIMD incoming images are broadcast to all processing nodes so that each (multiple instruction, multiple data) model can perform the required operation on a designated part or whole of the image there are several different ways to divide an application between nodes, for example: Transparent multi-processing: processing - let each node grab a different part of the same input frame automatically optimized within a node and work only on that each node has multiple processors - one C80 with 4 PPs, 1 MP - let each node grab and process a complete frame with each and one NOA successive frame going to a different node intelligent image processing functions automatically divide up - dedicate one node to grabbing and let it do the first part of the a task among the multiple processors to achieve maximum speed processing before passing the partial results on to the next node the user does not have to optimize the use of processors within in the pipeline a single node Matrox Genesis supports parallel or pipeline topologies or commands for control are provided to override task distribution any combination multi-processing designed to be completely configurable since the best approach depends on the individual application 5 Processing Nodes Video interface ASIC (VIA) Data Management key features To achieve real-time performance, an imaging subsystem must be integrates the following interfaces: able to handle the demanding interface requirements for simultaneous acquisition, processing and display of high-speed video. Matrox Genesis - 32-bit/50 MHz grab port was specifically designed to sustain high throughput I/O. Dedicated - 32-bit/33 MHz PCI bus Master/Slave interface high-speed buses transfer data between devices on the board, as well - 32-bit/33 MHz Controller/Master/Slave VESA Media Channel as between boards (i.e., Main Board and Processor Board) and to and - 64-bit/50 MHz glue-less SDRAM or WRAM memory port from external resources (e.g., host PC). The video interface ASIC (VIA) manages all data interface and performs efficient transfers. It also manages up to two concurrent data streams: performs real-time data formatting. - grabs to onboard memory with concurrent transfer between onboard ™ memory and VMChannel or PCI Dedicated high-speed data buses - PCI data transfers can be concurrent with host processor access 200 MB/s (32-bit) grab port interface for acquiring high-rate video to onboard memory and, in a multi-board configuration, simultaneously broadcasting implements memory bus arbitration the image to all processing nodes with the VIA's grab port, grabbed data is queued in the on-chip FIFO ™ 132 MB/s (32-bit) VMChannel for transferring images between and then burst to SDRAM so the processing is not slaved to the speed processing and display sections and, in a multi-board configuration, of the acquisition ™ between processing nodes (the VMChannel supports connection takes non-consecutive pixel data and reconstructs it in memory in of up to 13 devices, e.g., one Main Board and six Processor Boards) real time (e.g., from time multiplexed or multitap cameras, etc.) 400 MB/s (64-bit) local memory bus between processors (C80/NOA) writes to WRAM, performing sub-sampling, zooming and tagging and memory (SDRAM) enables efficient inter-processor communication: one C80 has access 400 MB/s (64-bit) local memory bus between display controllers to all of the host memory and the memory bank of all other C80s that (MGA/VIA) and memory (WRAM) are present in a multi-board system 132 MB/s (32-bit) PCI bus for transferring data offloads data management tasks from the C80 so the DSP is (e.g., synchronization information) between boards completely dedicated to processing PCI bus master interface with support for scatter-gather DMA 6 Protects your investment in software development Display Matrox developed MIL specifically for leading-edge developers facing Key features increasing competitive pressures to reduce costs and decrease time to supports high-resolution monochrome or color display market. With MIL, original software development takes less time and separate image buffer and overlay buffer for non-destructive overlay provides long-term benefits. Matrox pioneered the concept of portable up to 1600 x 1200 x 8-bit or 24-bit image display software tools that allow a smooth upgrade to new hardware platforms. up to 1600 x 1200 x 8-bit pseudo-color overlay display The investment you make in development today will pay off long into the future. display memory completely independent of processing memory supports live display of processed images For most MIL functions there is a corresponding Native Library function. However, some functions are only available through MIL (e.g., gauging, Software OCR, bar and matrix code reading, calibration, and video-in-a-window Matrox offers the most comprehensive image processing software available management). Also, Native Library offers some board-specific functions. for a vision processor board. When used with Matrox Genesis, MIL functions make calls to GNL, which executes operations on the processing node(s). The objective in Programming levels developing mainly with MIL functions is that most of the application By providing three different programming levels, Matrox meets will be portable. Moving the application later to a different platform the needs of a range of developers. will require changing only the board-dependent portion of the code. 1. MIL/ActiveMIL MIL - complete programming library for image capture, transfer, Common API across hardware processing, analysis and display Hardware-independent MIL is compatible with any VGA card and Matrox's ™ - fully exploits Intel MMX technology or Matrox vision processors line of PCI imaging hardware, from low-cost frame grabbers to Matrox - applications easily ported to new hardware platforms Genesis vision processor. It not only works with Matrox's current line of hardware, but with future imaging boards as well. The applications - processing performed to sub-pixel accuracy you write with MIL today can be reused on the PC or Matrox platforms - multi-processing and multi-threading support of tomorrow. ® ® ® - available as DLL for Microsoft Windows NT 4.0, Windows 2000 ® and Windows XP Optimum use of hardware resources - includes Matrox Intellicam camera configuration utility When used with Matrox Genesis, MIL commands make calls to GNL commands with little overhead. Performance will be virtually the same using equivalent MIL or GNL commands. 2. Matrox GNL - board-specific 'C' library Transparent hardware management ® - available as DLL for Microsoft® Windows® NT 4.0, Windows 2000, MIL automatically executes operations using whatever hardware ® ® Windows XP and LIB for QNX is available in the system, host CPU or onboard processors. - works with Matrox Genesis only - accelerates MIL functions Multi-processing and multi-threading MIL supports multi-process and multi-tasking programming - fully optimized for accelerated processing by the C80/NOA models. Multiple MIL applications not sharing MIL data or a single combination MIL application with multiple threads sharing MIL data can run under - automatically divides up tasks between parallel processors ® Windows NT 4.0/2000/XP. and the NOA (override control commands provided, if required) - suits the needs of developers who do not require a portable MIL provides synchronization mechanisms to access shared MIL data application and ensure that multiple threads using the same MIL resources do not ® - Native Library commands can also be used within a MIL application interfere with each other. These capabilities, coupled with Windows NT 4.0/ 2000/ XP, enable the creation of applications that distribute - does not require extensive knowledge of the C80 or workload across several CPUs in a multi-processor PC or multi-node DSP programming Matrox Genesis platform. 3. GNL Developer's Toolkit (DTK) - used in conjunction with Texas Instruments' C8x software development tools - programs the C80 directly (MP or PPs) - suits the needs of developers who must integrate native C80 code - provides a model for developing custom functions that behave as existing GNL functions 7 Comprehensive set of imaging functions Matrox GNL DTK MIL has a modular architecture, with modules for application and board The DTK provides the ability to integrate Native C80 code into control, as well as: an application. This is required when porting the control application - Image Processing from the host to the C80's MP (so that the host does not have to be involved in real-time control) or developing custom PP functions. - Blob Analysis The DTK is used in conjunction with Texas Instruments' C8x - Compression/decompression development tools. - Measurement - Geometric Model Finder Programming the MP in 'C' does not require detailed knowledge of the C80 and the user has access to all Native Library functions. For - Pattern Matching programming the PPs, some knowledge of the internal architecture - OCR of the C80 is required, including the workings of the Transfer Controller - Bar & Matrix Code Module and the Parallel Processors. It usually also requires programming in - Camera Auto-focus Tool assembly language to achieve maximum performance. - Calibration Module The GNL DTK is designed to remove much of the complexity of programming at the lowest level. It presents a model for writing new Matrox Genesis Native Library processing functions that behave exactly as existing Native Library Like MIL, Native Library has over 300 commands for image processing, functions. By following this model, developers can concentrate blob analysis, pattern matching, etc., along with additional commands their efforts on the heart of the function - the relatively small that take advantage of the board's specific architecture. If portability amount of PP assembly code that implements the algorithm is not a requirement, the entire application can be developed with (the so-called critical loop). Native Library. See Matrox Genesis Developers’ Toolkit brochure for more information. Matrox Genesis Native Library is a set of stub routines, one for each supported opcode. Native Library commands are initiated by the host and make remote procedure calls to the actual processing functions on the C80 (the Native Library "Shell" resides in the onboard SDRAM Software Architecture processing memory). Tasks are automatically divided among the multiple processors (the C80's, PPs and NOA). A set of commands is provided for controlling the use of the processors. Additional functionality (beyond MIL) includes: functions for hardware-specific features of the board (e.g., some hardware formatting features of the VIA) compound functions that perform several operations with only one call (e.g., imIntTriadic() is a function that takes 3 input arguments and processes them in a single pass) functions for finer control of parallelism (e.g., execute one operation on one of the C80 parallel processors (PP) and one operation can be split between the other PPs) the ability to develop an application under an environment not supported by MIL 8 Data types Data interface Supported data types include binary (1-bit per pixel), 1-band and custom Matrox Video Interface ASIC (VIA) performs optimized 3-band (color) integer including 8-, 16- or 32-bit (signed or unsigned), transfers between acquisition, display and processing; between floating point and RGB packed; and commands for converting boards, and to external resources (i.e., host) between data types. VIA offloads C80 from all data management 1 up to 400 MB/s processing memory bandwidth (64-bit/50 MHz) Hardware Specifications 1 Acquisition up to 400 MB/s display memory bandwidth (64-bit/50 MHz) monochrome or color, standard or non-standard, analog or digital 1 up to 132 MB/s VMChannel™ bandwidth (32-bit/33 MHz) 1 frame scan or line scan up to 132 MB/s PCI bus transfer rates (32-bit/33 MHz) 1 up to 64K pixels per line and 64K lines up to 200 MB/s grab port interface (32-bit/50 MHz) analog Display section 3 software-selectable modes Matrox MGA 2064W graphics engine - 4 channel/8-bit mode at up to 35 MHz 6 MB WRAM image frame buffer - 2 channel/8-bit mode at up to 70 MHz 2 MB WRAM overlay frame buffer - 1 channel/8-bit mode at up to 140 MHz resolutions up to 1600 x 1200 @ 85 Hz refresh rate PLL range: 5-35 MHz supports 1200 x 1600 portrait display mode programmable clock phase: 0°, 90°, 180° and 270° non-destructive overlay of text and graphics programmable gain, offset and DC restoration Connector pinouts To see connector pinout diagrams for the Matrox Genesis family of digital boards, please visit the product section of the Matrox Imaging web site: onboard up to 32-bit wide TTL input at up to 30 MHz www.matrox.com/imaging/products/genesis/ optional digital input companion board provides up to 32-bit wide RS-422 at up to 25 MHz and LVDS at up to 40 MHz Dimensions and environmental information Main board preprocessing 2 31.4 L x 10.7 H x 1.99 W cm (12.4" x 4.2" x 0.78") 10 MHz low pass filter (can be bypassed) H: bottom edge of goldfinger to top edge of board 2 34.1 L x 10.7 H x 1.99 W cm (13.4" x 4.2" x 0.78") synchronization and control includes retainer hsync, vsync, composite sync, pixel clock input Power consumption: 4.7A @ 5V or 23.5W; 230mA @ 12V or 2.76W; or output (TTL/RS-422/LVDS) 145.4mA @ -12V or 1.75W (GEN/F/64/8/STD) external trigger input (TTL/RS-422/LVDS) two exposure (timer) outputs (TTL/RS-422/LVDS) Processor board auxiliary I/Os (TTL/RS-422/LVDS): two in/two out 31.4 L x 10.7 H x 1.73 W cm (12.4" x 4.2" x 0.68") H: bottom edge of goldfinger to top edge of board configurable look-up tables 34.1 L x 10.7 H x 1.73 W cm (13.4" x 4.2" x 0.68") four 256 x 8-bit or two 8k x 16-bit includes retainer Power consumption 4.13A @ 5V or 20.65W (GPRO/F/64/F/64) Processing node 2.29A @ 5V or 11.45W (GPRO/F/64) TI's TMS320C80 DSP @ 50 MHz Matrox's Neighborhood Operations Accelerator (NOA) @ 50 MHz Main board/processor board 64 MB SDRAM operating temperature: 0º C to 55º C (32º F to 131º F) Matrox Video Interface ASIC (VIA) relative humidity: up to 95% (non-condensing) up to 13 nodes in a system 9 Ordered separately: Software Environment ® ® ® Software host driver for Microsoft Windows NT 4.0, Windows 2000, ® ® GENESIS/SW/CD Matrox Genesis Native Library and DTK Windows XP and QNX (requires access code included in ® Windows development done using DLL interface (MIL and GNL) with Genesis/DTK),MIL-Lite/ActiveMIL-Lite, ® ® Microsoft Visual C++ or ActiveX interface (ActiveMIL) with Matrox Intellicam, MGA drivers, ® ® ® Microsoft Visual Basic or C++ and on-line/printed manuals. ® ® QNX (GNL only) development done using LIB interface with QNX IDE custom code development for the C80 done using GNL DTK with Texas GEN-SW/QNX/CD Matrox Genesis Native Library for QNX Instruments C80 software development tools and GENESIS/SW/CD. GENESIS/DTK Matrox Genesis Developer's kit Ordering Information (includes access code) Part number Description Hardware: Main Board MIL 7 DEV P or U Matrox Imaging Library (MIL) GEN/F/64/8/STD or LVDS PCI vision processor with C80/NOA/64MB, (see MIL brochure for more details). integrated 8 MB display and standard/ non-standard analog color/monochrome Maintenance program line/frame scan acquisition. Each purchase of GENESIS/SW/CD includes one year of the Genesis Native Library Maintenance Program. The program entitles registered users to GEN-DIG-BRD/X/Y Digital input companion board. X is R for one year of technical support and free updates, and may be renewed at RS-422 and L for LVDS (requires/LVDS the end of the first year. option). Y is P for PCI and I for ISA. GEN/MAINTENANCE One year program extension. GEN-CBLADT/PCI Digital cable adapter board for the PCI form factor. Analog input cables IMG-7W2-TO-5BNC 2.4 m (8') input cable, one 7W2 Hardware: Processor Board to five BNCs. GPRO/F/64/F/64 PCI processor board with two C80/NOA/64MB nodes. Digital input cables DBHD100-TO-OPEN 3 m (10') input cable, high density GPRO/F/64 PCI processor board with single DB-100 to an open end C80/NOA/64MB node. (requires customization). PCB back planes DBHD-68-TO-OPEN 3.6 m (12') DBHD-68 to an open end GEN-BUS/2 VMC and GP2 back planes for two boards. (for synch and control only). GEN-BUS/3 VMC and GP2 back planes for three boards. Custom camera cables are available. Contact Matrox Imaging Sales for more information. GEN-BUS/4 VMC and GP2 back planes for four boards. Notes: 1. Bus speeds given are maximum theoretical transfer rates. 2. Includes required space for screws and nylon washers which are non-conductive. For more information, please call: 1-800-804-6243 (toll free in North America) or (514) 822-6020 or e-mail: imaging.info@matrox.com or http://www.matrox.com/imaging Corporate headquarters: Offices: Canada and U.S.A. Europe, Middle East & Africa France Germany All trademarks by their respective owners are hereby acknowledged. Matrox Electronic Systems, Ltd. reserves Matrox Electronic Systems Ltd. Matrox VITE Limited Matrox France SARL Matrox Electronic Systems GmbH the right to make changes in specifications at any time 1055 St. Regis Blvd. Sefton Park 2, rue de la Couture Inselkammerstr. 8 and without notice. The information furnished by Matrox Electronic Systems, Ltd. is believed to be accurate and Dorval, Quebec H9P 2T4 Stoke Poges Silic 225 D-82008 Unterhaching reliable. However, no responsibility license is granted Canada Buckinghamshire 94528 Rungis Cedex Germany under any patents or patent rights of Matrox Electronic Tel: (514) 685-2630 SL2 4JS Tel: (0) 1 45-60-62-00 Tel: 089/62170-0 Systems, Ltd. Windows and Microsoft are trademarks of Microsoft Corporation. MMX and the MMX logo are Fax: (514) 822-6273 U.K. Fax: (0) 1 45-60-62-05 Fax: 089/614 9743 registered trademarks of Intel Corporation. Printed in Tel: 01753 665511 Canada, 05-12-2003. $IE-4915-B Fax: 01753 665599