Web ProForums
CompactPCI
PDF of this tutorial5. Frequently Asked Questions about CompactPCI
- What is the CompactPCI bus?
CompactPCI is a modern, very–high performance industrial bus based on the standard PCI electrical specification on a rugged 3U or 6U Eurocard packaging. Unlike its desktop cousin, the CompactPCI board uses a high-quality 2 mm metric pin and socket connector and can be front-loaded into a rack mount system. - From where does the CompactPCI specification come?
Ziatech Corporation initiated the CompactPCI effort in late 1994 under the auspices of the PICMG. The CompactPCI specification is the result of a concerted effort of the CompactPCI subcommittee composed of the following companies: Digital Equipment, GESPAC, I-Bus, Pro-Log, Teknor, VMIC, and Ziatech. - What is the PICMG?
The PICMG is a consortium of industrial computer product vendors who develop specifications for PCI–based systems and boards for use in industrial computing applications. The member companies have combined sales of more than $500 million of industrial computers and include industry leaders such as Texas Microsystems, Industrial Computer Source, DEC, GESPAC, Pro-Log, Teknor, and Ziatech. Membership in PICMG is open to any organization or individual with a legitimate interest in helping to extend the PCI standard in the industrial marketplace. The PICMG can be contacted at 781-246-9318. - What other specifications has the PICMG issued?
The first effort of the PICMG has been to issue a specification for passive backplane computers using both PC–style (card edge connector) ISA and PCI bus. The PICMG has also issued a specification for a PCI–to–PCI bridge allowing passive backplane computers to extend the number of PCI slots. The CompactPCI project is a more recent effort driven by a subcommittee of companies within the PICMG. - Which applications does CompactPCI target?
CompactPCI is intended as an industrial bus for applications in real-time machine control, industrial automation, real-time data acquisition, instrumentation, military systems, or any other application requiring high-speed computing, modular and robust packaging design, and long-term manufacturer’s support. Because of its extremely high bandwidth, the CompactPCI bus is particularly well suited for many high-speed data-communication applications, such as routers, converters, and switches. A hot-swap feature has been included in the CompactPCI specification and will be particularly well suited for the telecommunications industry. - What are the unique features and benefits of CompactPCI?
Compared to the standard desktop PCI, CompactPCI supports twice as many PCI slots (eight versus four) and offers a packaging scheme that is much better suited for use in industrial applications. For example, CompactPCI cards are designed for front-loading and removal from a card cage. The cards are held firmly in position by their connector, card guides on both sides, and a faceplate that solidly screws into the card cage. Cards are mounted vertically, allowing for natural forced air convection cooling. Finally, the pin-and-socket connector of the CompactPCI card is significantly more reliable than the card edge connector of the standard PCI cards. The power and signal pins on the CompactPCI connector are staged to allow the specification to support hot swapping, a very important feature for fault-tolerant systems that is not possible on standard PCI. - Which processors can be implemented on CompactPCI?
Although PCI has gained most of its recognition as a local bus for x86-based PCs, PCI is at the core of all modern microprocessor designs. PowerPC and DEC’s Alpha, for example, are supported with PCI–compliant chip sets and can benefit from CompactPCI. In fact, CompactPCI is the industrial bus that does the most justice to these very–high performance new chips, giving them a system bus with all the bandwidth that these chips are capable of. - Are CompactPCI products less prone to early obsolescence?
Yes. Unlike the desktop PC market, which is driven by volume and fast-changing consumer demand, CompactPCI is driven by professional customers who value product stability and long-term availability. Nearly all major CompactPCI manufacturers have at least 10 years of experience serving the OEM marketplace and have made a reputation of protecting their customers from the dangers of early obsolescence. This is achieved by a careful selection of components and their suppliers and even, in many cases, by stocking several years’ worth of key components. - What are the software implications of PCI and CompactPCI?
The PCI architecture has been carefully planned to simplify the software integration of a peripheral device. For example, all PCI or CompactPCI devices have a set of 256 registers that contain information on the device identity, as well as a great deal of software programmable parameters, such as address maps or interrupt types and levels. As a result, the system CPU can automatically detect and identify a device on the bus and configure it without jumpers on the peripheral. PCI is a key element of the plug-and-play concept. - What are the system implications of CompactPCI?
Every modern computer architecture has an internal PCI bus, whether or not it supports PCI add-on slots (which it usually does). This is the case for nearly all Pentium® PCs, Alpha workstations, and PowerPC systems based on the PowerPC reference platform (PREP) or common hardware reference platform (CHRP) standard. CompactPCI makes it possible to build any computer compliant with these hardware system designs. As a result, CompactPCI systems can be built using standard components and can run practically any operating systems and hundreds of thousands of application software. - Which operating systems can run on CompactPCI computers?
Pentium®-based CompactPCI computers can run all operating systems that have ever been ported to the PC, including MS–DOS*, Windows3.11*, Windows95*, WindowsNT*, OS/2*, Berkeley Software Distribution (BSD) UNIX*, Linux*, OS–9000*, and QNX*. PowerPC–based CompactPCI computers will be able to run AIX*, SOLARIS*, WindowsNT*, Mac OS*, and OS–9*. - Will CompactPCI system run real-time operating systems well?
The performance of CompactPCI is particularly well tuned to real-time applications, from machine control to machine vision, fast data acquisition, and data acquisition. Operating systems such as OS–9, PSOS, and VxWorks will run very well on CompactPCI hardware and will be instrumental in the expected success of CompactPCI in these markets. - What products are, or will soon be, available on CompactPCI?
Pentium® CompactPCI CPU boards are available in 3U format and 6U format from three CompactPCI vendors, with more on the way. System-level add-on functions, such as 100 Mbps Ethernet, fast small computer system interface (SCSI), and accelerated video graphics array (VGA), are also currently available. Other companies have announced plans for PowerPC CPU boards, fast fiber-optics networks, image acquisition and processing, and motion controllers. - What other functions could benefit from CompactPCI?
CompactPCI can benefit all applications requiring very-high data-transfer rates. Data communication interfaces, such as asynchronous transfer mode (ATM) and broadband integrated services digital network (ISDN), are examples. In the field of high-energy physics research, very fast, multichannel, data-acquisition cards could also benefit from CompactPCI. Many of the most exciting applications are probably yet to be invented, but if history is any indication, the sophistication of systems will rise to use all available computing bandwidth that CompactPCI computers have to offer. - Are CompactPCI products expensive?
No. Even though they offer unprecedented performance, they are based on broadly available silicon implementation produced in very high volume. Additionally, because the PCI bus is not terminated, no external bus drivers are required to interface a PCI peripheral. An Ethernet controller, for example, connects directly from the leads of the controller chip to the PCI connector. CompactPCI products are typically priced below equivalent VME products but above desktop PCI or PCI products. - How can I get a copy of the CompactPCI specification?
You can get a copy by calling the PICMG at 617-224-1100. PICMG members receive free copies of the specifications and updates. You can get a membership application by calling the same number. - Can CompactPCI cards be plugged on a desktop PC?
No. Even though both use the same pins with the same electrical signals and are software equivalent, they are mechanically different and use a different style of connector. The desktop PCI card uses a card edge connector at the bottom of the board, and the I/O connectors are on the side. The CompactPCI card uses a 3U (100 mm × 160 mm) or 6U (233 mm × 160 mm) board format with a pin-in socket connector at the bottom and the faceplate and I/O connectors on the top. CompactPCI cards can be front-loaded or removed from a card cage. - What are the key specifications of CompactPCI?
3U or 6U Eurocard board format, 2 mm hard metric pin-in socket connector, 32- or 64-bit synchronous data bus, 32-bit address bus (four gigabytes total addressing), 133 or 266 Mbytes/s data transfer rate, multiprocessing, up to eight slot backplanes, bridging from PCI to PCI or from PCI to other buses, plug-and-play ready. - What makes the PCI and CompactPCI bus so fast?
The PCI and CompactPCI get most of their speed from three factors. First, the data bus is 32 or 64 bits wide, which means that it can move four or eight bytes within a single cycle. Second, the bus is clocked at 33 MHz, which means that a basic cycle is 33 nanoseconds long. Finally, data transfer is based on synchronous burst transfer in which the address is given once at the beginning of a read or write cycle, followed by a string of continuous data. - What is the PCI and CompactPCI top speed (really)?
The theoretical top speed of PCI and CompactPCI is 133 Mbytes/s using 32-bit-wide data or 266 Mbytes/s using 64-bit-wide data. Note that the most current revision of the PCI specification allows a bus clock speed of 66 MHz, which results in 266 and 533 Mbytes/s, but this faster version is not part of the CompactPCI specification. The 133 Mbytes/s (in 32-bit data mode) top speed could be achieved in the case of very-long data-block transfer with a processor and memory sufficiently fast to handle such a continuous transfer. Practically, the memory or CPU is not always ready to send or receive data when the bus is, and the overhead needed to initiate a transfer cycle (address setup) is not always negligible. As a result, a real-speed figure for PCI or CompactPCI is in the range of 80 to 100 Mbytes/s (with 32-bit data), which is still several times faster than the “real” throughput of most other high-performance buses. - Why is there a limit in the number of slots in PCI?
To achieve this very high speed using low power, the PCI bus is not terminated, which means that not only are reflections not eliminated as in most other fast buses, but reflections are also expected and used by the bus drivers. In addition, all bus timings are kept within a very tight tolerance of two nanoseconds. The combination of tight clock timing and reflection characteristics imposes narrow margins on trace length and load impedance. A PCI bus driver is designed to drive only 10 loads. A connector pin accounts for one load and so does the pin of a PCI peripheral chip. As a result, there can be only four PCI slots in a typical desktop PC system. CompactPCI allows backplanes with up to eight slots. - How many slots can a CompactPCI system have?
The CompactPCI cards use a very–high performance DIN–style pin-and-socket connector with five rows of 47 pins plus a metallic shroud that acts as a shield. This design gives the connector exceptional impedance characteristics and a capacitance of 2 pf versus 10 pf for a typical desktop PCI connector. The CompactPCI bus driver, card, connector assembly, and backplane assembly have been fully characterized using a detailed simulation and have been proven to support up to eight slots (instead of four for other PCI implementations). - How can a CompactPCI system be expanded to more slots?
CompactPCI systems (and PCI bus systems in general) can be expanded past their eight-slot limitations (past four-slot limitation for standard PCI) using PCI–to–PCI bridges. These bridges are devices that connect to the PCI bus on one end and can drive another set of eight (or four) PCI slots at the other. They act as a repeater, “repeating” what they see on the main PCI bus to the PCI bus on the other side. PCI bridges could conceptually be added indefinitely to build very large systems. - Is there a penalty for expanding PCI with bridges?
Practically none. During the initialization phase, the PCI–to–PCI bridges must be configured so that all devices on the buses can be mapped. Once this is done, the CPU does not need to know about the bridges and can access a device three bridges downstream in exactly the same fashion as a device on the main PCI bus. Each bridge adds a clock cycle (33 ns) to a data block transfer. If the amount of data that is moved is low (one or two words) this additional cycle may be significant overall. If a block of 100 words is moved in a transfer, this added cycle would typically degrade the performance by 1 percent or less. - How well does CompactPCI or PCI handle simple I/O transfer?
CompactPCI is technically able to access simple and slow I/O devices like a parallel port or an A/D converter, for example. The transfer would not be particularly efficient because such devices transfer 8 or 16 bits at the time and would not benefit from the burst capabilities of the PCI bus, but a complete write cycle could be completed in around 100 ns, which is still faster than any I/O bus in the market today. Practically, however, CompactPCI and PCI introduce a level of complexity (configuration registers, first in first out [FIFO], identification register, etc.) that may add unjustified complexity and cost if the I/O peripherals are very simple. This is why, in a typical system, a PCI–to–I/O bus is used. In the desktop PC, the I/O bus is the ISA bus. In CompactPCI systems, the I/O bus may be PC/104, G-64, STD, or VME. - Does CompactPCI support multiprocessing?
Yes. It is possible to have boards with processors on a CompactPCI (or PCI) backplane. These boards can take control of the PCI bus and send data directly onto another PCI peripheral on the bus. The PCI bus on the host processor (system controller) arbitrates the access to the PCI bus if several intelligent peripherals (slaves) attempt to take control of the PCI bus at the same time. This type of multiprocessing is typically designed to let intelligent peripherals perform dedicated tasks to off-load the host CPU. - Can there be more than one master in a CompactPCI system?
Yes. PCI supports arbitration for peripheral masters to acquire the bus and perform transfers directly. The system master in CompactPCI is responsible for providing arbitration services for peripheral masters. Peripheral masters are boards, such as SCSI interfaces or Ethernet interfaces, which need high-bandwidth access to memory. Peripheral masters may also include intelligent boards with microcontroller or microprocessor capability. - How are interrupts handled in PCI and CompactPCI?
PCI supports shared interrupts where multiple devices can generate an interrupt using the same line and are then differentiated using an interrupt acknowledge cycle. This resolves an important limitation of the traditional PC architecture where one interrupt line had to be used for every interrupting peripheral, thus limiting the number of such devices and allowing dangerous conflicts. The shared interrupt support makes PCI and CompactPCI much better suited for I/O intensive real-time applications than was the ISA bus. - What are the differences between CompactPCI and VME?
From an appearance point of view, there exist very few differences. Both use the same style of rugged Eurocard packaging. Both are available in 3U and 6U versions. CompactPCI uses a single 220-pin 2 mm connector for the complete 64-bit specification. VME requires two connectors and a 6U card for a 32-bit implementation. CompactPCI is rated at 266 Mbytes/s compared to 40 for the original VME specification. A recent extension to the VME specifications to 64-bit (VME–64) has narrowed the gap, but CompactPCI remains the speed champion when it comes to raw data movement. More fundamentally, however, VME is a bus that is very closely tied to the 68000 microprocessor family. As a result, it is not well suited as a system bus for Pentium®-based processors. For example, it is impossible to have system-level PC functions (VGA, networking, disk controller) on the VME bus. Similarly, PowerPC processors now expect their peripheral functions to be on a PCI local bus to seamlessly integrate with the BIOS and operating system. As the 68000 family continues to retire, PCI and CompactPCI will be increasingly better suited for system-level and high-speed extensions, whereas VME will remain an excellent I/O bus for many years. - Is CompactPCI destined to replace VME?
Some applications may continue to use the VME architecture for some time because of the specialized interface requirements. - How does CompactPCI differ from the public mobile carrier (PMC)?
PMC is a mezzanine bus concept based on the PCI bus specification. PMC cards are designed to be stacked on a VME CPU board, and typically, no more than two PMCs may be mounted on a 6U CPU. By contrast, up to eight CompactPCI cards can be mounted on the same backplane. PMCs can only be installed or removed by first extracting the carrier board from the backplane, which makes it less modular and more difficult to handle. More fundamentally, the PMC concept admits that the local piggyback bus may be more powerful than the main VME backplane bus. CompactPCI is founded on the assumption that a fast backplane bus is required to take best advantage of the power of modern microprocessors.
