IBM's Micro Channel Architecture

Technical Cornucopia, March–April 1989.

by Lloyd Borrett
Technical Cornucopia, March–April 1989

With the introduction of its Personal System/2 (PS/2) range in April 1987, IBM also introduced a new system architecture. The PS/2 model 30 used the familiar AT style bus and the models 50, 60 and 80 used the new Micro Channel Architecture (MCA). Since then we have seen the model 30 286 using the AT style bus, and the model 50Z and 70 386 which use MCA.

The problem that has dogged IBM ever since is that but a few existing and intending PC users are able to understand the concepts of MCA, and that almost no hardware add‑ons and software applications have been created to exploit the performance benefits of MCA. So why did IBM do it?

Why a new architecture?

Contrary to industry speculation, IBM didn't deliberately change the bus structure to ensure a proprietary PC design. Work on the Micro Channel Architecture (MCA) began in 1983. IBM were having problems meeting the FCC requirements regarding radio and television interference. IBM also wanted to respond to customer requests  for a "switch less" setup of add‑in boards.

Also the user environment was changing. Early PC users worked in a singular environment, with one PC supporting a single user. Today's PC user often works in a multi‑environment with multi‑tasking and/or multi‑user requirements. Increases in data traffic to and from I/O devices in the multi‑environment require increases in data throughput. Also the importance of data integrity increases in proportion to the number of users and tasks. And the consequences of a system crash become much greater.

Significant changes in technology had occurred since the release of the 64 KBytes RAM, 160 kb diskette, 4.7 MHz 8088 based IBM PC back in August 1981. Today's PS/2 Model 80 can support 4 MByte of memory on the motherboard, two 1.44 MByte diskette drives and up to 628 MByte of hard disk storage (even more if you use non IBM options). Processors in use today need 16 and 32 bit data buses.

IBM's priorities

Once the decision to produce a new system architecture was made, the next stage was to compile a list of priorities that would determine what features actually got used:

1. Safety
2. Data Integrity
3. System Integrity
4. Reliability
5. Compatibility
6. Functionality
7. Performance
8. Cost

Notice how low a priority IBM assigned to functionality and performance when compared to data and system integrity. Why? Because for most people the value of the data stored in their computer is worth far more than the value of the computer itself.

An example of how this priority list effected the decisions made is the switch from 5.25‑inch format diskettes to 3.5‑inch diskettes. The 3.5‑inch media is far more reliable which means increased data integrity (item 2 on the list). Since data integrity rates above compatibility and cost, IBM made the decision to switch.

Note that compatibility rates above functionality and performance. Obviously the designers could have been more innovative and thus increased performance had they been able to ignore the software compatibility issue.

The solution

A system architecture designed to meet the need for improved reliability, data integrity, higher performance, and error recovery in a multi‑environment. The advances inherent in the Micro Channel Architecture include:

• Level sensitive interrupt handling
   
• Error logging and channel check reporting
   
• Micro Channel timing specifications
   
• Watchdog timer
   
• Programmable Option Select
   
• Design for electromagnetic compatibility
   
• Surface mount technology and large scale integration
   
• Bus masters

The results

One of the most obvious differences is the "Programmable Option Select" feature. This makes it possible for users to forget about having to set DIP switches when installing add‑in boards. Instead the MCA replaces these switches with a set of memory registers on each card that contains setup information. This saves installation time. Incorrectly set switches are the number one cause of costly and time wasting service calls.

The use of LSI technology has greatly reduced the number of chips required to make a system and the number of interconnections  between components. For example, to support the same number of DMA devices as the MCA, a PC would need 16 more signal lines. This would mean 31 contacts on every connector and adapter. The MCA requires only 6 signals (25 fewer contacts). The 32‑bit PS/2 connector is approximately the same length as the 16‑bit AT connector.

Surface Mount Technology (SMT) permits more efficient manufacturing techniques to be used. It also allows the placement of up to eight times as many modules per adapter. The combination of LSI and SMT means a much higher mean time between failures can be expected. The statistical measure of reliability is more than twice as high for the PS/2 than for the AT. This is already being reflected in reduced maintenance costs to users.

Putting it all together

The MCA has plenty of room for growth. We are only just beginning to see ideas surface that utilise MCA to its full extent. And IBM has built in a lot of reserve potential to cater for future requirements.

Some system characteristics can easily be measured, such as performance and compatibility. Other important functions, such as data integrity, reliability, and maintainability are not as easy to quantify.

Only a very small proportion of today's computer users have experience with the mini and mainframe multi‑environment factors that are incorporated in the PS/2 Micro Channel Architecture. Most have no idea of the impact that multi‑environment applications will cause, and how their existing hardware environments will need to change.

At first many so called experts passed off the PS/2 and MCA as an unnecessary change. Now that they have begun to see the problems that the coming multi‑environment software solutions cause they are changing their tune.

Who knows, some of them may make the EISA architecture work, or maybe they won't. There is a "standard" solution today. It's called Micro Channel Architecture. Can you remember what happened to those companies who tried to bypass the "PC" standard? Or are they now a long forgotten memory?