The AMD K-Series chips started it all for AMD. Previously, their 286, 386, and 486 chips were almost exact copies of the Intel designs. With the K5, they came up with their first independent design. Of course, being new to it, they had some difficulties and the K5 was late coming into market. It was limited to cheaper systems. The K6, on the other hand, was a very successful chip.

The K5

Although beyond extinction, the K5 still has the power to run your basic business programs. The K5, like the Cyrix 6×86, uses PR rating system in order to help convince customers to purchase the K5 over more expensive Intel chips.

The K5 started out as the SSA/5 and came in 75, 90, and 100 MHz versions. These chips, unfortunately, suffered from a very poor FPU and a weak core design. The result was a disappointing chip. AMD, though, saw that there was a market demand for a faster chip, and the K5 release had already been delayed. After the initial release, AMD went back to the drawing board and tweaked up the original K5. The result? The K5 most of us think of today.

The K5 was a nice Pentium alternative and is socket compatible with the Pentium. The chip comes in 100, 120, 133, and 166 MHz versions. The newest PR166 K5 runs faster than a Pentium 166, and in some cases, roughly the same speed as an average Pentium 200. It, as is the case with the other K5\’s, is named with the PR system which compares it with the equivalent Pentium performance. The K5 PR166 actually works at 116.7 MHz with a 66 MHz bus. This leaves it with a 1.75 multiplier which kicks in whenever an external multiplier of 2.5x is used, which breaks the normal multiplier of 1.5 for all the other K5\’s. It lacks MMX, but can perform 2 instructions per clock cycle, like most other chips. It has 16K instruction and 8K data L1 cache and contains 4.3 million transistors.

For business users, the K5 was a nice performer at the time. The FPU, even in the improved version, does not compare with that of the Pentium. Having a high-end graphics card will help, but you will still not get the performance of the Pentium on such applications.

For more information…

• AMD-K5 Processor

The K6

The K6 is AMD\’s sixth generation MMX processor. It started out as NexGen\’s 686, but when AMD got off to it\’s rocky start with the K5, it decided to buy NexGen for the 686 design.

Its fully compatible with 16-bit and 32-bit applications and is Socket 7 compatible. It has more onboard cache: 32K for both instruction and data, for a total or 64K. It has RISC86 superscalar architecture. This is the key to it\’s MMX capability. It implements the x86 instruction set by internally decoding x86 instructions into RISC86 operations that support the x86 instruction set. Rather than directly executing complex x86 instructions, it performs the simpler, fixed-length RISC86 instructions while satisfying the requirements for x86 programs.

With the K6, AMD abandoned the PR-Rating system and went back to the standard MHz rating. No longer was there a standard Intel Pentium to compare it too with the release of the standard Pentium, Pentium Pro, and the Pentium II. Besides, the PR rating system was simply confusing potential buyers of the K6.

Its technical features are below, provided by AMD.

• Sixth generation performance
  
• Advanced RISC86 superscalar microarchitecture
  
  
  
  • Seven parallel execution units
    
  • Multiple sophisticated x86 to RISC86 decoders
    
  • Two level branch prediction
    
  • Speculative execution
    
  • Out-of-Order Execution
  
  
• 64K on-chip level one cache
  
  
  
  • 32K instruction cache
    
  • 32K writeback data cache
  
  
• High performance floating point unit
  
• MMX capability
  
• SMM mode
  
• Socket 7 compatible
  
• .35 micron architecture

Based on benchmarks, the AMD outperforms the Intel Pentium MMX equivalent with the 166 and 200 MHz versions. The 233 MHz K6 falls slightly short of the performance of the Pentium II/233. PC Magazine claimed that it was faster than the Pentium MMX/233, but not the Pentium II. The K6\’s larger L1 cache helps to boost the chip\’s performance, but it\’s MMX capabilities and FPU fall short. Although it\’s MMX capability are good, it can only process one MMX instruction at a time, while Intel\’s can do two. The FPU, or floating point unit, falls short of Intel\’s. It\’s throughput is not as high and therefore can only start a new operation every two clock cycles. Therefore, on applications such as image-editing or AutoCAD, the K6\’s performance will be slower. The FPU of the K6 is good and very well designed, but it is not quite up there with that of Intel. Although a good graphics card will help close in the gap, it cannot catch up to Intel on graphical software.

For more info, check out AMD\’s page on the K6…AMD-K6 MMX Processor