Matrox Imaging - News & Events - Feature Stories Archives

PCI Bus Imaging Boards are here

Long-anticipated, and now arriving for scientific/industrial, videographic and graphic arts and multimedia imaging

by Laura Robinson, Matrox Imaging

Original article featured in Advanced Imaging Magazine, June 1994.

For some time, imaging OEMS, integrators and end users alike have been hearing that PCI bus image processing boards for varied markets are on the way. Now they're here--and could have more impact on image processing performance in more ways than might even have been anticipated. By offering cards for the PCI bus, board suppliers enable OEMs and end users to take full advantage of today's 486, Pentium and RISC-based desktops, which have ample processing power for many imaging applications.,/div>

The PCI bus standard was specifically designed to keep pace with the new CPUs, and to match the demands of throughput-intensive applications such as image processing by moving major amounts of information at breakneck speeds.

Last month, Matrox released the first PCI frame grabber to the imaging market. The "Matrox Comet" is a high performance color and monochrome grabber with SVGA display and live video-in-a-window designed for industrial/scientific image analysis applications. Matrox Marvel II also hit the market last month, as the first multimedia card on the PCI bus; Matrox introduced MGA Ultima PCI, a graphics accelerator card for the PCI bus, last November. From scientific/industrial imaging to videographics to multimedia, PCI power is at hand. These introductions are part of an industry-wide trend towards the PCI bus standard in high performance computing.

Access to advanced PC power

PCI boasts a 32-bit data path and a 33 MHz clock speed, which yields a maximum theoretical data transfer rate of 132 Mbytes per second. Using the speed of the PCI bus, a PCI frame grabber can provide rapid image transfer from the on-board grabber to the host computer, where images can be processed by today's increasingly powerful CPUs. PCI technology also facilitates the display of true color, high resolution images and graphics, and full-motion video-all of which, of course, are now much in demand.

An imaging system can be configured with a PCI frame grabber and powerful imaging software (for example, Matrox Comet supports ITOOLS) running on a high-end PC, for a cost-effective solution in image analysis or machine vision. The PCI bus has paved the way for inexpensive, yet potent, host-based imaging systems.

Power for multiple platforms

The PCI (which stands for Peripheral Component Interconnect) bus standard was developed by a consortium of vendors in the computer industry, led by Intel. It stems from the need to remove the bottleneck in moving data from the CPU to peripheral device-which tend to be far from "peripheral" in the imaging marketplace. As processors have become more powerful, I/0 buses have been left behind.

In a 486 CPU, data moves around at 33 MHz, but slows right down once it enters the I/O bus. For example, the AT bus, developed in 1984, and the EISA bus, which came out in 1989, both have speeds of 8 MHz. (Although the EISA does offer DMA burst mode which can achieve a throughput of up to 33 MB/s.) MCA, which was first introduced in 1987 has a speed of 10 MHz. The PCI bus removes the bottleneck between processor and peripheral and delivers the performance required to work with advanced processors such as the 486 and beyond.

PCI was developed based on local bus architecture, which takes peripherals off the I/0 bus and connects them, together with the CPU and memory subsystem, to a wider, faster pathway for data transfer between the CPU and the peripherals. Significantly, the PCI bus is processor independent, it is one step removed from a system's processor bus, occupying an inter-mediate level between the processor bus and the standard expansion buses, with electronic bridges spanning the gap.

Since PCI is processor-independent, it can be easily hooked into a wide variety of CPUs, including Intel's Pentium, DEC's Alpha, MIPS R4400, Motorola, Apple and IBM's PowerPC. The PCI bus will enable different processor-based systems to use the same expansion cards. For example, when the second generation of Apple's PowerPC-based Macintosh arrives, it will utilize PCI instead of Apple's own NuBus. This means that the Mac world will be able to choose from standard off-the-shelf PC peripherals.

The PCI designers originally decided to isolate the PCI bus from the CPU's local bus in order to resolve some of the problems with a previous design for a local bus, the VESA Local (VL) Bus. For example, the PCI design can support more devices than a VL design because PCI devices do not electrically load down the CPU bus.

Another drawback of the VL Bus is that it was a "quick-to-market" solution that lacked tightly defined standards, so it has had more than its share of incompatibility problems. The PCI standard has robust specifications which are carefully administered by the PCI Special Interest Group. Moreover, when running a true multitasking operating system such as OS/2 or Windows NT, a PCI bus system can achieve higher performance than a VL bus system.

Plug-and-play auto-configuration is another benefit of the PCI bus standard. A PCI board has configuration specs set in memory on-board and provides installation information to the system during the boot-up phase. BIOS routines then automatically configure each PCI device. This means users can plug in add-on boards without having to deal with for example, DIP switches and jumpers to select I/O channels, etc.

Built-in future

PCI is definitely prepared to be the bus of the future. The PCI bus specification includes a well-defined migration path to a more powerful 64-bit bus architecture, which will double the maximum data rate to 264 MB/s. The specification also defines a 3.3-volt system as well a 5-volt system.

Because of PCI's many advantages, industry experts expect PCI to be the dominant bus standard in the industry for the long term. One telling sign of this is the industry-wide support that PCI has achieved. All major system vendors are supporting this standard, including Texas Instruments, Hewlett Packard, DEC, IBM, Compaq, Dell, Gateway 2000, AST, NEC, Apple, and NCR.

One of the reasons PCI is catching on is that it offers lower development costs for system developers, since they can reuse system designs across CPU platforms. Also, because the specifications include low-volt-age and the 64-bit interface, PCI facilitates development of multiple product lines.

PCI imaging

As a supplier of imaging hardware and software, Matrox's strategy in developing products for the PCI bus is to allow customers to make the most of today's PCs by coupling a PCI grabber with powerful software tools running on the host.

The Matrox Comet supports ITOOLS, comprehensive imaging application development software, containing a high level library with functions such as: point-to-point, statistical, neighborhood, convolution, morphology, blob analysis, and pattern matching and alignment.

It should be understood that for certain imaging applications, a high performance PC can handle the image processing requirements, making it unnecessary to purchase any additional accelerator hardware, which does, of course, save OEMs and end users money.

But despite that fact, or the appearance of "PCI bus PCs", some applications will always require intelligent boards. Matrox plans to provide a range of PCI solutions from frame grabbers to accelerator boards.

For more information on obtaining the PCI specification, call the PCI Special Interest Group at: (800) 433-5177.

For more information, contact our Media Relations Team.