Is Pentium The Only Way? 
 
     by Mike Hagerty, President 
     Monterey Bay Area Users Group 
 
     A number of people have asked me if it is true that the 386 is dead. 
     They have been told that they will not be able to function without 
     a 486 or even a Pentium (i.e., 586) in their computer.  This is 
     malarky. A little history may put this into perspective. When Morse 
     and Ravenal first designed the 8086 for Intel, one of the major 
     design constraints was that it reportedly had to be more or less 
     "compatible" with the 8080, the 8-bit chip that was the brains 
     inside a CP/M system.  This limitation was levied upon the designers 
     because Intel wanted an easy port of software onto the new chip and 
     to make it easy to use other chips which already worked with the 
     8080. 

     The result was a cobbled up architecture which would only support 1M 
     of memory, unless one resorted to strange mapping schemes. Instead 
     of being a nice flat address space (easy for programmers to use), 
     it was divided into segments.  The maximum size of any segment was 
     64K, exactly the address space of the 8080. Programs could span 
     multiple segments, if care was taken, but not data structures. 

     Motorola took the opposite approach, essentially burning the bridge 
     from their earlier processors and started from scratch with the 
     68000 (the processor in the original Macintoshes).  Unfortunately 
     for a lot of programmers who would have had an easier time, 
     Motorola was incapable of shipping the 68000 until after Intel had 
     captured the IBM PC design and was thus relegated to perpetual 
     second place. 

     Intel developed the 80286 to provide a number of enhancements 
     requested by developers.  Although the 80286 was technically an 
     improvement over the 8086, it did not provide the flat address space 
     which developers really wanted. This would have required discarding 
     the segmented address space.  It would have burnt the bridge from the 
     8086 and potentially opened the market to the 68000.  After all, if 
     the new chip is incompatible with the old, it is no more difficult 
     to go to the competitor's chip.  The 80286,  was the brains inside of 
     the AT.  Because of its strong ties to the 8086 and additional 
     design limitations, the 286 was described as being "brain-dead". The 
     time for bridge-burning had come, but Intel was playing for time.  

     The 386 was for Intel what the 68000 was for Motorola, a break with 
     the past.  The 386 is capable of addressing a large amount of memory 
     providing programmers with a reasonable architecture. To preserve 
     their investment, Intel made certain that programs written for the 
     8086 could run on the 386. However, programs written to take 
     advantage of the architecture of the 386 could never run on earlier 
     processors.  

     This is why WinNT and OS/2 will never run on your old AT. The 486 
     merely added Intel's floating point co-processor and a little cache 
     to the 386 resulting in a faster chip.  Unless you are doing a lot 
     of graphics or math, the floating point co-processor is useless.  If
     you do need one, a 387 can be added to any 386 system for less than 
     $100.  Thus, the 486 provides essentially nothing but a speed-up.
     The real reason why Intel is pushing the 486 is that they have very 
     little competition for 486 market.  

     There are many firms making 386s faster and all cheaper than Intel's.  
     The margin in the 386 market has shrunk and Intel would like us to 
     buy their more costly and higher-margin 486s instead.   The Pentium
     (586) is not presently in volume shipment and no one else has made 
     any. You can rest assured that they will be very pricey indeed. 

     So, to answer the question posed to me, "No, you do not necessarily 
     need a 486.  A 386 will probably work just fine."  The 486 is faster, 
     but it is not appreciably different.  The 386 architecture will be 
     around for a long time (in computing terms, that is).