Mac-IBM-Compare Version 2.1.5

Changes: Mac and OS information updated; Pricing section eliminated in favor of links.

Note: for proper reading off line this document should be in a monospaced font such as Monaco 9pt or Courier 10pt.

The reason for this general data sheet is to consolidate and condense the information out there, so that people in both camps can be clear and accurate about what they are saying about their machines.
Since computer technology is always changing there are always going to be points in which the sheet will be lacking or incorrect on information. So, please just don't say the sheet is incomplete or incorrect but also give me clear and concise information to make the needed corrections. To keep this data sheet accurate please provide article citations,if possible, for the information provided or corrected and keep opinions to a minimum. As this is a general data sheet, keep the info provided short and simple.
{For example, if you want more details on PC hardware the comp.sys.ibm.pc.hardware FAQ is a much better source than this sheet is.}

Please note that all magazine dates are in mm/dd/yy formats.
Finally, keep the information relevant to the section corrected.
Thank you.

Home pages:
mac-ibm-compare.html
mac-ibm-OS-compare.html

Contents

The CPUs

Note: I am only showing Motorola & Intel CPUs used in Macs and most IBM/PC clone machines. Years indicate use of CPU chip in new machines.
Cache is "where data can be stored to avoid having to read the data from a slower device such as a disk" (Dictionary of Computer Terms:61-DTC). Both IBM and Mac use caches external to CPUs which increase the speed of the CPU but are not a part of it. Since there are many different external caches, each having a different effect on CPU performance, with some built-in {present Macs}, other optional but installed {IBM}, and are machine, seller or expansion dependent, I have decided to leave them out of the list.

SPECMarks: OSG SPECMarks measure the IDEAL performance of processors NOT real world performance. In some cases hardware configuration can effect SPECMarks themselves. For example, the presence or absence of an optional L2 cache card effected SPEC92marks for BOTH the Pentium and MPC601 by 10-15%. OS-software design and interaction resulted in certain crossplatform programs on MPC601 Macs outperforming their Pentiums counterparts by almost 50% even though the SPEC92ints were within 10% of each other.

Note:  ALU is industry's de-facto standard for CPU bit classification.

IBM     ALU  Registers    External     CPU     Features/
CPU                     data  address cache    Notes
8088(6) 16      16     8 (16)   20    none     {1981-9} {197?-89}
80186   16      16       16     20    none     {198?-9?} 8088(6) segmenting
80286   16      16       16     24    none     segmenting + Protected Mode*
386sx   32      32       16     24    none     80386*
80386   32      32       32     32    none     MMU & 32-bit Protected Mode 
486sx   32      32       32     32   one 8K    80486 w/o FPU
80486   32      32       32     32   one 8K    new CPU core {~386 + FPU}
486dx2  32      32       32     32   one 8K    doubled internal clock rate**
486dx3  32      32       32     32   one 16K   80486 w/o FPU; IBM chip
 [Blue Lightning]                              between 486dx2 and Pentium 5 
                                               lines in speed (BYTE 04/94:22)
DX4     32      32       32     32   one 16K   80486; Intel's version of the
                                               486dx3.
Pentium 32      32       64     32   8K code,  2 instructions/cycle max
 [P5]                                8K data,  2-issue superscalar, 386 
[P54c]                               Branch    Write-Back, 64-bit FPU,
                                     target    pipelining***
P6      32      32       64     32   8K code   CRISC
                                     8K data  
                                    256K SRAM 
P7                                             PA-RISC or VLIW with a
                                               hardware x86 code translator,
                                               1997-1998****

386sl: low power (3.3V) 386sx with built-in power management. Laptop use.
386slc: IBM 5V 386sx with a 16k on-chip cache added (John H. Kim).  As far as
 John H. Kim knows it is only used on IBM models.
486slc: Neither of two chips that have this name have a FPU.  Cyrix: basically
 486sx in 386sx socket with 1k cache and improved integer math speed.  IBM: 
 equivalent to 486sx with a 16k on-chip cache.
486slcs: IBM chip equivalent to 486dx2 - FPU with a 16-bit external data path
 and 16k on-chip cache.
486dlc2: IBM chip equivalent to 486dx2 - FPU and with 16k on-chip cache.
P24T{Pentium}:  64 bit internally, 32 bit for system I/O.


* 16 MB maximum RAM
** ex. for 486dx2/50, chip runs 50 MHz, rest of machine runs at 25 MHz.
 The chips are used in Apple's DOS compatible card for some Macs.
*** CPU Info & System Performance Summary  
 All Pentiums made before Dec 94 have a bug in the FPU; there was an offer
 for replacement chips and a software solution that turns off the FPU. 
 386 and early 486 code can run slower on a Pentium then expected; 16-bit code
 runs slow on P6s, 32-bit code runs at expected speeds.
**** Computergram International, 06/02/94; BYTE 4/95:54.  VLIW (Very Long
 Instruction Word) is supposed to be faster and cheaper than RISC but it has
 a major problem in that VLIW binary code is _so_ CPU specific as to be
 TOTALLY incompatible with _any_ future VLIW CPUs {it would be as if 386 code
 would not run on 486 and 486 code would not run on a Pentium}
 (InfoWorld, 01/17/94 v16 n3 p29(2); Microprocessor Report, 02/14/94 v8 n2
 p18(4); BYTE 4/95:54)

Mac     ALU  Registers  External bus   CPU     Features/
CPU                     data  address cache    Notes
68000   16      32       16     24     none    {1984-93} 16 MB limit*
68020   32      32       32     32   256 code  {1987-92} parallel processing
68030   32      32       32     32   two 256   {1988-94} 68020 + MMU, 16K
                                               burst mode.
68EC040 32      32       32     32    two 4K   68040 w/o FPU and MMU {~68020}
68LC040 32      32       32     32    two 4K   68040 w/o FPU {~68030}
68040V  32      32       32     32    two 4K   68040 w/o FPU, PowerBooks-1994
68040   32      32       32     32    two 4K   MMU, FPU, pipelining, doubled 
                                               internal clock rate**

Note: The Amiga and the now defunct NeXT used the 68030 and 68040.
* Earlier 68000 Mac designs created a 4 MB limit.
** (Electronic Buyer's News Aug 20, 1993; pointed out by Bradley Lamont and
   Motorola 68040 data book in 92) Apple marketed 68040 Macs as 'clock
   doubled' though speed is NOT in the 486dx2's class {see CPUs Comparison
   List}.  The 68060 never really took off as a Mac CPU or as third party
   upgrade cards.

The following PowerPCs are to be in both IBM and Mac machines.  They are Motorola/IBM RISC CPU chips.  More details on the PPC chip and it uses can be found at the PowerPC FAQ.

PowerPC ALU  Registers  External bus   CPU     Features/
CPU                     data  address cache    Notes
MPC601  32    int 32     64     32     32K     3 instructions/cycle max,
[.6]          fp  64                 combined  80 MHz - 11 W
[.5]                                   I/D     100 MHz [.5] uses 5 W.
MCP601+                                        100 MHz - 2 W*
MPC603e 32    int 32     32/    32    two 16k  100 MHz - 3.5 W 
              fp  64     64            I/D     ~300 MHz - {limit}*
MPC604e 32    int 32     64     32    two 16k  4 instructions/cycle max
              fp  64                   I/D     ~300 MHz - {limit}*
MPC620  64      64      64/     40    two 32k  133 MHz - 30 W {simulated}*  
                        128    64 e           
                               80 v
G3                                             200-400 MHz, 97-2000*
G4                                             500-700 MHz, 98-2001*
Project K2                                     undefined, 2001*
                                                  
MPC630 - in development*

e - effective  v - virtural

* (Motorola PPC; CPU Info & System Performance Summary); Power PC future)
Note that the SPEC web site gives a more detailed view of SPECMarks than the above sites do. Code using POWER commands can run ~50% slower than expected on the 603/604 chips.

CPUs Comparison List

As a general rule of thumb Motorola chips are faster than Intel chips at the same frequency {030/25 ~= 386/33; 040/33 ~= 486/50}, but Intel has chips at higher frequencies than Motorola, so this evens out. The Macintosh Bible 6th edition and IBM System User, 1/92 v13 n1 p43(1) support the comparisons made between Intel and Motorola chips below and statements made here.

<=80186 ~ 68000

16-bit vs 16/24/32-bit chip. The 4 MB limit on the 68000 Macs brings the chip in them down to the 80186 and lower chips, otherwise the 68000 would compare to the 80286.

286 ~ 68020

hardware segmenting vs. 68020's 32-bit ALU and having no usable built-in MMU unlike their successors [80386, 68030]. The use of the hardware segmenting and the 16-bit nature put the 286 between the 60000 and 68030 in features and the LC's 16-bit data path strenghthens the 286 ~ 68020 comparison.

386 ~ 68030

32-bit chips with MMUs, and protected memory. At present application protected memory is limited to A/UX 3.0. System 7.x uses this feature to protect a RAM disk created by the Memory control panel which is supported only on Powerbooks, Quadras and Power Macintoshes. 68030 Macs with 16-bit hardware paths are comparative to 386sxs.

486sx ~ 68LC040

same as 486 and 68040 without the FPU; used as a low cost solution for people who do not need the FPU. Only in comparison with Windows programs does the 68LC040 approch 486SLC2 - chip cache speeds.

486 ~ 68040

32-bit microprocessors with built-in FPU, MMU, 8K internal cache (which is implemented as two 4K caches in the 68040 and one in the 486). The 486dx2 is in Macs with the DOS card installed. Please note that MHz for MHz a 486dx2 outperforms a 68040. For example a 68040/40 is ~95% the speed of a 486dx2/66 (Ingram 93 report)

Pentium ~ None

The 68060 wasn't a standard Mac CPU and PowerPC chips running native code are ~1.50 times as fast as a Pentium of the same MHz (Ingram 94 and 95 reports)

PowerPC = PowerPC

This CPU line is planned to run programs from OS/2 and MacOS through the CHRP platform. The 615 is planned to have a x86 section.

Hardware

PowerPC machines

Apple and clone Macs web pages

The MacOS Buyers Guide features a guide to all current MacOS systems available in the U.S. and their prices.

APS Technologies - The M·Power systems are customizable. Go on to http://www.apstech.com/aps-byob.html for more details.

Daystar - Daystar Digital systems: Multi processor Mac clones.

Marathon Computer offers make-your-own systems

Motorola - Motorola StarMax

Power Computing - systems are customizable. See http://store.powercc.com/ for more details.

Monitor/Display support

Both Mac and IBM have graphic accelerators. In Macs VRAM can act like a graphic accelerator but only boosts things by <=10%. The range and scope of these is accelerators beyond the purpose of this web page.

Mac

VRAM: Video RAM.

Standard for present non-PowerBook Mac's handling of built- in video. 640x480 x 24-bit is effectively replacing and 8-bit and 16-bit as the base standard though old modes still exist.

IBM

Because of the wide range of OSes available for IBM, the use of drivers bypassing BIOS, video hardware inconsitancies {see Super VGA below} and nonstanderzation of clone BIOS have left resolution of video display hardware/OS/program interaction up to the OS and video hardware in question (Faisal Nameer Jawdat). In addition, IBM and clone makers never bothered to provide a standard hardware mechanism for software to determine what display mode is actually present (Matt Healy) nor a standardized screen-drawing toolbox {like Mac's QD}. As a result detecting some modes and/or use them consistantly is a challange, especially with some third party cards. Things were so dependent on the interaction of the program, OS, print driver and monitor card that editing 32-bit pictures regardless of color mode, program, and monitor type/card combination as one can do with the MacOS was near impossible (Fortune 10/04/93:112). DOS had the biggest problem, Windows is better, and NT-OS/2 are the best but there is still some degree of inconsitacy from program to program.

MDA: Monocrome Display Adapter

Original character-mapped video mode, no graphics, 80x25 text.

CGA: Color Graphics Array

320x200x2 or 640x200 b/w with text-mode support of 80x24 color or 40x24 color, 16 color palette, bad for the eyes.

EGA: Enhanced Graphics Array

Resolutions are 640x350x4, 640x200x4 and all CGA modes {from 64 color palette}. Additional text-modes of 80x43 color and 40x43 color. Some versions could run at 256 colors, bearable on the eyes.

VGA: Video Graphics Array*

Resolutions are 320x200x8, 640x480x4 and all EGA modes. Additional text modes of 80x50 color, 40x50 color, and emulation of all MDA text modes. Can be programmed for many non-standard resolutions. All modes have 256 colors, from a 18-bit {IBM} to 24-bit {IBM/Mac} color palette. 25.175 MHz Pixel Clock (Mel Martinez). Monitors use analog input, incompatible with TTL signals from EGA/CGA etc.

MCGA: Multi-Color Graphics Array*

Subset of VGA that provides all the features of MDA & CGA, but lacks some EGA and VGA modes, like VGA 640x480x4 (DCT). Common on the initial PS/1 implementation from IBM and some PS/2 Models.

SVGA: Super VGA {Quality of 99% of color Mac video monitors}*

This is not a standard in the way the others were, but instead was a 'catch all' category for a group of video cards. In an effort to clean up the chaos resulting from each manufacturer using their own implementation scheme VESA was established and is used in the newer units, but things are still a mess with some people still debating as to what is SVGA and what is not. Video is either 512K [~1990], 1MB [1992], or 2MB [today], resolution of 800x600 and 1024x768 at 256 and 32,768 colors are common with most 24b at 640x480. Speedwise, too much variation and change.

Other non-SVGA standards:

8514/a

IBM's own standard, graphics accelerator with graphics functions like linedraw, polygon fill, etc. in hardware. IBM version interlaced.

TMS34010/34020

high end graphics co-processors, usually >$1000, some do 24-bit, speeds up vector-oriented graphics like CAD.

XGA: eXtended Graphics Array {May be used in IBM PowerPC}

newer and faster than 8514/a, originally only available for MCA bus-based PS/2s. Emulates VGA, EGA, and CGA (DCT). Max resolution: 1024x768x8, also some 16 bpp modes.

XGA-2

Accelerates graphics functions up to 20 times faster than standard VGA in Windows and OS/2, including line draws, bit and pixel-block transfers, area fills, masking and X/Y addressing. Has an intelligent way to detect and co-exist with other XGA-2 cards, so multiple desktops like on the Mac may not be far away. Since this is an architecture, its resolution and color depth isn't fixed {IBM implements only 16-bit [65,536] color, while other companies can have 24-bit color through IBM technical licenses}. Refresh rates up to 75 Hz, ensures flicker free, rock solid images to reduce visual discomfort, and is VGA compatible. Up to 1280x1024 on OS/2.

*some monitor types usable by Mac. See Mac section above for specific details.

Expansion

{Speeds are baced on: throughput = (bus_clock_speed x byte_width) / transaction_overhead (Mel Martinez)}

Both Mac & IBM (* = Part of CHRP/PPCP)

IDE: Integrated Device Electronics

Asynchronous {~5 MB/s max} and synchronous {8.3 MB/s max} transfer. Currently the most common standard which makes its drives ~10% cheaper than SCSI. Outperforms SCSI in single-tasking OSes while underperforms SCSI in multitasking enviroments {it is not clear if this also applies to co-operative multitasking}. (BYTE 08/94:116) Limited to two drives per controler and 528 MB.

Enhanced IDE*

Under under ideal conditions this sustains about 5 MB/s. Supports larger hard drive sizes and non-hard drive devices. Apple uses EIDE internal drives that support logical block addressing in certain Mac series or models.

USD (universal serial bus)*

Daisy-chainable hot-pluggable 12 Mbps 127 device maximum port. Targeted at keyboards, mouses, modems, ISDN connections, joysticks, and floppy-disk drives. macuser

SCSI*

First external device expansion interface common to both Mac and IBM. Allows the use of any device: hard drive, printer, scanner, Nubus 87 card expansion {Mac Plus only}, some monitors, and CD-ROM. SCSI stared out at 5 Mhz with fast SCSI rasing that to 10 Mhz. SCSI is limited to a total of 8 devices and an ideal total cable length of 6 m which cheap cables, terminators, connectors, and device impedance can shorten substantially. (BYTE 8/94:112) Conversely the delopment of 'repeaters' is expected to increase the cable length.
Main problems: Since SCSI is supposed to be terminated ONLY at the begining and end of the SCSI chain, internal terminated external devices are the biggest bane of SCSI. This along with other problems are planned to be fixed with SCSI-3.

SCSI-1

8-bit asynchronous {~1.5 MB/s max} and synchronous {5 MB/s max} transfers. 8-bit SCSI-2 is often mistaken for SCSI-1 {see SCSI-2 for details}.

SCSI-2 (Includes Ultra-SCSI)

Fully SCSI-1 compliant though timing margins where "shaved" so 8-bit SCSI-2 is a little faster than SCSI-1. 16 and 32-bit SCSI-2 require different ports, electronics, and SCSI software drivers from SCSI-1. Ports are 68-pin {16-bit} and two 68-pin/one 104-pin {32-bit}. Single cable 16-bit SCSI-2 is accually 16-bit SCSI-3. Transfer speeds are 3.0 MB/s max asynchronous and 4-6 MB/s with 10 MB/s burst synchronous {8-bit}, 8-12 MB/s with 20 MB/s burst {16-bit}, and 15-20 MB/s with 40 MB/s burst {32-bit} (BYTE 08/94:114). Ultra-SCSI is a method for doubling the synchronous 8-bit and 16-bit transfer speeds (Ultra-SCSI white pages)

SCSI-3

Is really this is a family of standards which includes no internal terminated devices allowed, access to more than 8 devices, standarized a single cable for "Wide SCSI-2" tranfers, and extended SCSI commands to non-SCSI devices (nickname Serial SCSI). Speedwise not much difference from SCSI-2 on Parallel SCSI side but a wide range on the Serial SCSI side (http://www.zdnet.com/macuser/mu_0496/myslewski.html):

FireWire [IEEE P1394; Serial Bus] - SCSI-3 Serial Bus Protocol (SBP)

Was intended to replace ADB, RS-232, RS-422, parallel and SCSI (BYTE 07/93:90). Has six shielded wires with 100 Mb/s {12.5 MB/s}, 200 Mb/s {25 MB/s}, 400 Mb/s {50 MB/s]. Did not require terminators and ID numbers and had limits of 63 devices per port and up to 1022 buses being bridged together. (Microprocessor Report, Mar 07/94 v8 n3 p18(4); Byte 07/94:37; Byte 08/94:120)

SSA (Serial Storage Architecture) - SCSI-3 Serial Storage Protocol (SSP)

Uses only four signal wires, requires no address switches, external drivers, receivers, or discrete terminators and allows up to 127 devices on the chain with 'hot pluggability' as well as multiple simultaneous reads and writes. Maximum speed: 80 MB/s. (SSA overview)

FC-AL (Fibre Channel - Arbitrated Loop) - SCSI-3 Fibre Channel Protocol (FCP)

Supports 30 meters between devices with coaxial cable 1, 126 devices, 'hot pluggability', and Maximum throughput of 200 MB/s. (Fibre Channel Whitepapers)

Mac SCSI

The Mac’s 25-pin SCSI port was designed from earlier SCSI-1 specs and is not true SCSI-1. Older systems had an SCSI Manager that limits some 8-bit SCSI-2 Macs to synchronous SCSI-1 speeds. The rise of EIDE devices has effectivley eliminated the cost advantages of the old 25-pin SCSI-1 port.

IBM SCSI

SCSI-1 is now pretty common but it is generally not bundled with systems, except as add-on with EISA and VESA Local Bus adapters avalable {See IDE information}. Like the Mac, 8-bit SCSI-2 is used as a very fast SCSI-1 by many controllers out there. Unlike the Mac, IBM had no exact SCSI controller specifications {until CorelSCSI} which resulted in added SCSI incompatibilities (Byte 10/92:254). This combined with the Mac’s non-standard 25-pin SCSI port sometimes makes PC SCSI-1 incompatible with Mac SCSI-1.

PCI: Peripheral Component Interconnect

32-bit {64-bit expandable}, up to 33 MHz bus clock, combining EISA and VLB advantages; supports up to 10 slots {5 cards} (Jay C. Beavers; MacWeek 5/31/93). Intel's version of Local Bus is designed with a PCI->ISA/EISA/MCA bridge in mind (PCI spec (rev 2.0)).
PCI is slightly different on the Mac and PC sides. On the Mac side PCI is always 33 MHz, on the PC PCI MHz is whatever multiple of the system bus speed that is 33MHz or less; ie 75 MHz CPU -> 25 MHz, 100 MHz -> 33 MHz, and 120 MHz -> 30 MHz. Replaced Nubus.

Mac

Memory expansion

The Mac does a complete memory check at startup by writing/reading every memory location; if something is seriously wrong with a SIMM/DIMM the Mac will not boot and give a sound chord indicating what the problem is.
Memory data width must match the memory controller's width. Non-PowerBook Macs have three basic memory expansion set-ups:
Non-parity 30-pin 8-bit SIMMs {Before Feb 15, 1993} - While 30-pin 9-bit parity SIMMs could be used in these Macs, only special IIcis could make use of the parity feature. Later 30-pin 16-bit SIMMs replaced these.
64-pin SIMMs with a parity option - IIfx only.
72-pin 32-bit SIMMs - Non-PowerBook and non-PCI Macs made after Feb 15, 1993. 168-pin 64-bit DIMMs - PCI Macs.

CPU expansion

Handled through a daughter card, PDS or NuBus. Unlike PDS, Nubus CPU cards could allow the use of multiple processors at the same time {Like MCA; example-RocketShare} via parallel processing. Each NuBus card needed its own memory but most NuBus cards of this type came with 8 MB RAM of SIMMs on the card standard. The 630 68x00->PPC daughter cards allowed use of the old 68x00 CPU reducing the need for the software emulator.