Am386

An AMD 80386DX-40 in a 132-pin PQFP, soldered onboard
Produced	1991
Marketed by	AMD
Designed by	AMD
Common manufacturer(s)	AMD
Max. CPU clock rate	20 MHz to 40 MHz
FSB speeds	20 Mhz to 40 Mhz
Min. feature size	1.5 µm to 0.8 µm
Instruction set	x86 (IA-32)
Microarchitecture	80386
Product code	23936
Cores	1
L1 cache	Motherboard dependent
L2 cache	none
Application	Desktop, Embedded (DE-Models)
Predecessor	Am286
Successor	Am486
Package(s)	DX variant: 132-pin PGA 132-pin PQFP SX variant: 88-pin PGA 100-pin PQFP

The Am386 CPU was a 100%-compatible clone of the Intel 80386 design released by AMD in 1991. It sold millions of units, positioning AMD as a legitimate competitor to Intel, rather than being merely a second source for x86 CPUs (then termed 8086-family).^[1]

History and design [link]

While the AM386 CPU was essentially ready to be released prior to 1991, Intel kept it tied up in court.^[2] AMD had previously been a second-source manufacturer of Intel's Intel 8086, Intel 80186 and Intel 80286 designs, and AMD's interpretation of the contract, made up in 1982, was that it covered all derivatives of them. Intel, however, claimed that the contract only covered the 80286 and prior processors and forbade AMD the right to manufacture 80386 CPUs in 1987. After a few years in the courtrooms, AMD finally won the case and the right to sell their Am386 in 1992. This also paved the way for competition in the 80386-compatible 32-bit CPU market and so lowered the cost of owning a PC.^[1]

While Intel's 386 design peaked at 33 MHz, AMD released a 40 MHz version of both its 386DX and 386SX, extending the lifespan of the architecture. The AMD 386DX-40 was popular with small manufacturers of PC clones and with budget-minded computer enthusiasts because it offered near-80486 performance at a much lower price than a real 486.^[3]

The 386DX-40 could match or even slightly outperform a 486SX-25 in popular benchmarks and many real-world applications, while costing less. Integer performance at 40 MHz thus approached that of low-end 486 CPUs, but rarely exceeded it. This is because the 486 needed fewer clock cycles per instruction, thanks to its tighter pipelining (more overlapping of internal processing) in combination with a crucial on-chip CPU cache. However, because the Am386DX-40 had the same 32-bit width on its data bus as an 80486, but operating at the same 40 MHz rate as the processor, (rather than the 25 to 33 MHz buses of the 486 DX-2s) it had comparatively good memory and I/O performance.^[4]

• 

  An Am386DX-25

• 

  The Am386DE-33 is an embedded version of the Am386DX-33

• 

  A PGA Am386DX-40

• 

  A PQFP Am386DX-40 on a 132-pin PGA adapter

Am386DX data [link]

• 32-bit data bus, can select between either a 32-bit bus or a 16-bit bus by use of the BS16 input
• 32-bit physical address space, 4 Gbyte physical memory address space
• fetches code in four-byte units

AM386 SX [link]

In 1991 AMD also introduced advanced versions of the 386SX processor -again not as a second source production of the Intel chip, but as a reverse engineered pin compatible version. In fact, it was AMD's first entry in the x86 market other than as a second source for Intel.^[5] AMD 386SX processors were available at faster clock speeds at the time they were introduced and still cheaper than the Intel 386SX. Produced in 0.8 µm technology and using a static core, their clock speed could be dropped down to 0 MHz, consuming just some mWatts. Power consumption was up to 35% lower than with Intel's design and even lower than the 386SL's, making the AMD 386SX the ideal chip for both desktop and mobile computers. The SXL versions featured advanced power management functions and used even less power.^[5]

• 

  An Am386SX-25

• 

  An Am386SX-33

• 

  An Am386SX-40

Am386SX data [link]

• 16-bit data bus, no bus sizing option
• 24-bit physical address space, 16 Mbyte physical memory address space
• prefetch unit reads two bytes as one unit (like the 80286).

80387 coprocessor [link]

Floating point performance of the Am386 could be boosted with the addition of a 80387DX or 80387SX coprocessor, although performance would still not approach that of the on-chip FPU of the 486DX. This made the Am386DX a suboptimal choice for scientific applications and CAD using floating point intensive calculations. However, both were niche markets in the early 1990s and the chip sold well, first as a mid-range contender, and then as a budget chip. Although motherboards using the older 386 CPUs often had limited memory expansion possibilities and therefore struggled under Windows 95's memory requirements, boards using the Am386 were sold well into the mid-1990s; at the end as budget motherboards for those who were only interested in running MS-DOS or Windows 3.1x applications. The Am386 and its low-power successors were also popular choices for embedded systems, for a much longer period than their life span as PC processors.

• 

  An IIT 387SX-25 Coprocessor

• 

  A Cyrix FasMath 387DX-33 Coprocessor

• 

  An ULSI 387SX-40 Coprocessor

References [link]

External links [link]

Wikimedia Commons has media related to: AMD Am386

• AMD.com: Am386 Family 32-bit Processors
• AMD Am386®SX/SXL/SXLV Datasheet
• cpu-collection.de: Pictures
• AMD: 30 Years of Pursuing the Leader. Part 2
• Shvets, Gennadiy (5 November 2011). "AMD 80386 microprocessor". CPU-World. https://www.cpu-world.com/CPUs/80386/MANUF-AMD.html. Retrieved 29 November 2011.