All computers have a central processing unit (CPU), also known as a processor, or, when built onto a silicon chip, as a microprocessor. This device, usually with other ‘chips’ and hardware, performs the ‘core’ tasks of a computer.
The processor’s arithmetic logic unit (ALU) is used to perform simple logical tasks, assisted by several registers that temporarily store numbers during processing. The abilities of a processor are often set by the maximum number of binary digits or bits that these elements can handle.
The internal parts of a processor are linked by internal bus circuits. The number of wires used for each address bus or data bus are known as lines or bits and can limit the performance of the device. Connections to other hardware, such as a computer’s memory or hard disk drive, are made via external bus circuits. In this case, the number of bits can determine how many external devices can be used and the speed of data transfer. A lesser number of bits are often used for external buses, either to simplify the connection technology or because lower speeds are required for these circuits.
A processor’s clock rate sets the pace of numerical processing and has a significant influence on performance. However, there are numerous other considerations. For example, older processors are invariably based on complex instruction set computing (CISC), where a wide range of operations are used within the ALU. More recent devices are centred around reduced instruction set computing (RISC), involving repeated use of simpler instructions. A RISC device therefore requires a less complex design, making the processor run cooler and allowing it to work faster.
The subjective speed of a computer is also influenced by other factors. For example, most processors have an area of fast memory known as a cache, which stores recently-used data, allowing it to be used again quickly, but without the slow access of a hard disk drive. In some computers the cache isn’t inside the processor itself but is in a separate memory device.
Processors often contain a floating point unit (FPU), a special form of co-processor that’s designed for rapid calculations of real numbers. This can considerably speed up a machine that’s used for complex work, especially when processing graphical material. In older computers an FPU is sometimes fitted as a separate device, external to the processor.
Mac OS computers contain processors jointly developed by the AIM alliance, consisting of Apple, IBM and Motorola. ‘Classic’ Macs use the first generation of processors, retrospectively known as the G1 family. The first device in this range is the Motorola 68000, followed by the 68020, 68030 and 68040 processors, all known collectively as the 680x0 series.
The second generation, known as the G2 family, consists of early PowerPC processors, such as the 603e or 604e. However, more recent computers incorporate later PowerPC products, such as the G3, G4 and G5 series of processors.
The following table shows the number of bits used for the ALU and registers in some types of Macintosh processor, as well as for the external data and address buses:-
* Hardware can impose 4 MB limit on older machines
† Integer register: also has 64-bit register for floating-point numbers
‡ Effective addressing is 64-bit, virtual addressing gives 80-bit range
• Not pin-compatible with single-processor versions
The G2, G3 and G4 RISC devices are made by both Motorola and IBM. In addition, several of the processors shown above are also used in other computers. For example, 68030 and 68040 devices appear in Amiga machines and in NeXT computers, whilst the PowerPC series of processors are incorporated in various non-PC machines manufactured by IBM.
As mentioned above, a cache acts as a store for recently-processed data, allowing it to be reused by the processor. It can also as a high-speed buffer, accommodating rapid data flows to and from the disk drive, but without overloading the main RAM. Transferring data in and out of the processor can also be hampered by the computer’s bus speed, which is limited to the 33, 66, 133 or 200 MHz rate of the Peripheral Component Interconnect (PCI) bus in older PowerPC designs.
Some machines use a backside cache, often running at half the speed of the processor. An older computer with a 400 MHz processor may use a cache running at 200 MHz and a system bus operating at 100 MHz. The backside cache ratio indicates the difference between processor and cache speeds: for example, our 400 MHz processor with a 200 MHz cache has a ratio of
2:1. In other machines an inline cache is used, running at the same speed as the processor itself.
Modern PowerPC-based machines can contain three different caches:-
A very fast buffer, usually 32 or 64 KB in size, built into all modern processors.
This cache is built into G4 processors, although older devices require a separate SIMM or DIMM, typically with a capacity of 32 KB, 128 KB, 256 KB, 512 KB, 1 MB or more. In some models the module is soldered in place, preventing any upgrades.
A slower cache only found in G4 machines, often running at one third or one quarter of the processor’s speed. In a typical computer this cache has a capacity of 1 or 2 MB, sometimes in the form of Double Data Rate RAM (DDR RAM). Some recent Apple PowerBook models don’t have this cache at all, instead increasing the Level 2 cache from 256 KB to 512 KB.
Most Macs only have one processor. But, despite being out of fashion for a time, models with multiple processors are beginning to reappear. Early attempts at multi-processing in the Classic Mac OS used the Multi-Processing (MP) protocol, where one processor handled system tasks whilst less important work was dealt with by a second or third device.
The PowerPC 603 and G3 processors use the Modified, Exclusive and Invalid (MEI) system for a common cache, with processors and memory kept in step, while the PowerPC 604 uses the Modified, Exclusive, Shared and Invalid (MESI) protocol for multiple caches. The G4 uses MEI and MESI, as well as MERSI for improved processor communications. Both Mac OS 9 and Mac OS X accommodate symmetric multi-processing (SMP), which treats all processors equally.
The processors in older PowerPC-based machine can usually be upgraded, although it’s often more economic to buy a new computer. Some upgrades are also incompatible with later versions of the Mac OS, whilst the limited bus speed of older machines (usually 33 MHz, as compared with 100 MHz or higher in newer machines) can minimise the advantages.
Modern computers often have a zero insertion force (ZIF) socket that can accept a newer processor. Or you can put an upgrade card into a PCI slot of a recent machine or into the Processor Direct Slot (PDS) of a ‘classic’ model. A carrier card, which is fitted with the same ZIF socket as a modern machine, allows you to swap processors with other computers.
A processor can be made to run faster by the dubious practice of over-clocking. This technique, optional with some upgrades, shifts the operating speed closer to the absolute maximum that the processor can handle. Unfortunately, this makes the device run at a higher temperature, possibly shortening its life and reducing its general reliability. Over-clocking requires special software, as provided with the upgrade, and is often in the form a control panel for the Classic Mac OS.
Once you’ve over-clocked your machine you should check to see if it’s still reliable.
Although older Macs don’t have the processing power or memory capacity for complex audio-visual tasks, they’re perfectly useable for word-processing and other non-demanding work. In fact, they can be invaluable in charitable organisations or ‘third world’ countries where nothing else is available. As far as the hardware is concerned, Apple computers are highly reliable. And if something does go wrong you can usually obtain another machine and ‘cannibalise’ the parts.
The first Macintosh appeared in 1984. It’s a ‘compact’ machine, containing a small screen that provides a monochrome picture. This range includes the following models:-
where Mem is the minimum memory in KB, Clk is the clock rate in MHz and Disk is the diskette size in KB.
The Mac II series of computers, first introduced in 1987, employ a separate keyboard and monitor. These machines also support NuBus expansion cards. The following variations were produced:-
where Mem is the minimum memory and HD indicates the size of hard disk drive, both given in MB.
The Mac LC, a low-cost colour machine, first appeared in 1990. It’s fitted with a 68020 processor, lacking a co-processor. The full range of LC models are as follows:-
The Quadra and Centris ranges first appeared in 1991 and 1993 respectively, including:-
In 1992, a range of Performa machines were introduced, although these were mainly variations on existing models. The earliest types of Power Macintosh appeared in 1994, fitted with the 601 variety of PowerPC (PPC) processor, also known as a Reduced Instruction Set Chip (RISC) device. By 1995 machines were available with 603 and 604 chips, the latter being far more powerful than the older versions. Later Performas, such as the 5200 and 6200, also contain 603 processors.
The Mac Portable arrived in 1989, followed by the PowerBook (PB) in 1991. Early portables include:-
The processors in the IBM PC and its successors are entirely incompatible with those found in Mac OS machines. Unfortunately, it’s also impossible to make any sort of meaningful comparison between the two groups. The following table provides information on some of the devices used in PC-based machines:-
* Double-speed internal clock means that a 486dx2/50 runs at 50 MHz, although the machine itself operates at 25 MHz. This processor is used in DOS-compatible cards fitted in some older Macs.
• 386 and early 486 code can seem slow on a Pentium processor whilst 16-bit code is slow on a P6. 32-bit code runs at the expected speed. The FPU in Pentium processors prior to December 1994 have a bug: this can be fixed by using software to switch off the FPU or by replacing the processor.
The motherboard in a PC has one of several different processor sockets. A 423-pin socket is the minimum requirement for a Pentium 4 processor, whilst a 603-pin socket appears on cards designed for future compatibility (with the ability to have a large level 3 cache on or off the processor ‘die’). A 478-pin socket with Pin Grid Array (PGA) is also commonly used.
Mac-IBM-Compare, Bruce Grubb
Macintosh Hardware History, Rod Ackland, University of Glasgow Computing Service
©Ray White 2004.