The Intel Processor Family

First Generation 

• 8088: 8 bit processor, and 8 bit data bus for cheapness 
• 8086: identical architecture to 8088 but with a 16 bit data bus; 4.77 MHz and 8 MHz clock speed 
• 80186: half way between 8086 and 80286; not used in PCs, used (today) in embedded microcontrollers 


Second Generation 

• 80286: 16 bit processor; 16 bit data bus; wider address bus to accommodate 16mb of memory; 6 - 20 MHZ clock speed; the first AT (Advanced Technology) chip running in protected mode (a high performance mode) 


Third Generation 

• 803860X: 32 bit processor; 32 bit data bus; 32 bit address bus allowing (in theory) 4 GB of RAM; introduced pipelining (start execution of a second instruction before the first has been compleled) to allow increased clock speeds; up to 33 MHz clock speed (Cyrix and AMO equivalent chips had a 40MHz version); 
• 80386SX: 32 bit processor; 16 bit data bus; 16 - 33 MHz clock speed; up to 16MB RAM only (same as 80286); cheaper than OX, but slower 

Fourth Generation 

• 804860X: 32 bit processor but 32 bit data bus; 32 bit address bus; maths co- processor; Levell cache memory; deeper pipelining; burst mode memory transfer (automatically fetch the next memory contents before they are requested) to reduce the time processors spend waiting for memory access (the Wait state); 25 or 33 MHZ clock speed; 100-150% performance improvement over 803860X. 
• 80486SX: as for OX, but the maths co-processor is disabled, resulting in reduced performance, but a cheaper chip. 
• 804860X-2: the clock speed doubled to twice the rate of the internal bus system (the motherboard speed); needs a heat sink; 50 MHz or 66 MHz clock speed (AMO and Cyrix had 80 MHz version running with a 40 MHz motherboard); performance comparable with early Pentium chips; still a viable processor for routine office processing. 


Fifth Generation 

Pentium I 

A 32-bit microprocessor introduced by Intel in 1993. It contains 3.3 million transistors, nearly triple the number contained in its predecessor, the 80486 chip. A Pentium chip is double the performance of a 486 chip of the same clock speed. The Pentium I clock speed ranges from 60 - 233 (MMX) MHz. 


Superscalar Architecture 

The Pentium introduced Superscalar Architecture: two parallel execution units a pair of 80486 processors - on the one chip. This means that it can execute, in parallel, two instructions per cycle, for which it needs two 32 bit pipelines (data highways), one for each processor. For maximum efficiency, pipelines must be kept busy, so instructions are pre fetched. The next instruction in the sequence is fetched from memory ready for when it is needed.

Superscalar architecture requires instructions to be retrieved and delegated intelligently, otherwise the pipeline stalls and execution units are left idle. Problems occur when, for instance, an executing instruction is dependent on data generated by another instruction which has not yet completed. Thus, instructions must be found which can be executed simultaneously. 

Handling branches (when a jump occurs to an instruction which is not the next instructi0l1 in the sequence) is a problem for Superscalar architecture. Thus branch prediction is used - a 'guess' is made as to whether a branch will be taken or not -'the 'guess' determines which instructions will next be pre fetched. Branch prediction is correct over 90% of the time. Despite these strategies, however, it has been shown that superscalar architecture rarely operates to full efficiency, resulting in idle hardware, and thus lower overall execution speeds.

Cache Memory 

The Pentium has 16 KB of level 1 cache (split into Bus System twO 8 KB parts, one part for data anq one for instructions). See notes on cache memory. 


Bus System

The Pentium has a wider data bus of 64 bits to double the bandwidth for memory transfers. Additionally the system clock speed is 60 or 66 MHz - double that of the 80486 motherboard. The Bus system's speed is also increased: a Pentium I runs at 60 or 66 MHZ. 

Floating Point Unit 

The Pentium's floating point unit is also considerably faster than that of the 80486. Floating Point arithmetic is a method of carrying out calculations on real (fractional numbers), where the position of the radix (decimal) point in both numbers is equalized before the calculation, then moved to its proper place in the calculation result. 

MMX 

Intel introduced the Pentium with MMX Technology. MMX is designed to enhance the processing of multimedia and communications data. Additionally, the new Pentium chip doubled the level 1 cache to two 16 KB parts. These and other architecture improvements gave the MMX Pentium a significant advantage for performance improvement over the regular Pentium chip.

你可能感兴趣的:(System,performance,branch,parallel,generation,Numbers)