RapidCAD-1 & RapidCAD-2 , 一套两个(CPU+FPU) =========================================== intel RapidCAD是英特尔有史以来第一款为旧款个人电脑所提供的升级套件（也就是OverDrive的始祖）。原386的使用者不需要更换主机板，只要把RapidCAD买回来将主机板上旧有的中央处理器晶片（CPU）替换掉，就可以享受接近486的运算能力。RapidCAD其实就是把486 DX晶片去掉内部快取记忆体然后装入386的封装里面，RapidCAD也不支援486增加的新指令。不过由于386封装的频宽限制，RapidCAD对整体的效能提升比不上直接升级到486 DX。相同时脉下，486 DX可以有比386/387快上两倍的速度，而RapidCAD在整数运算方面最多只能提升35%，在浮点运算方面，则可以提升将近70%。
Intel RapidCAD特殊的地方在于，它是由两颗晶片组成，缺一不可。这归咎于486 DX内建浮点运算器（FPU），而386则是将浮点运算器分开（就是387）。由于RapidCAD-1本身就含有浮点运算器（因为它就是486 DX阉割版），根本不需要387，所以RapidCAD-2就是用来替代原来主机板上的387晶片。RapidCAD-1负责所有的运算，而RapidCAD-2则是负责假装浮点运算器，以防止旧有主机板以为没有安装浮点运算功能（尤其在执行286/287的程式时）。市面上有时候把RapidCAD-1与RapidCAD-2分开卖，这是就是不了解RapidCAD运作方式的结果。
RapidCADis a specially packagedIntel 486DXand a dummyfloating point unit(FPU) designed as pin-compatible replacements for anIntel 80386processor and80387FPU. Since the DX variant has a working on-chip FPU, a dummy FPU package is supplied to go in theIntel 387FPU socket to provide the FERR signal. Despite being able to execute instructions in the same number of cycles as an i486DX, integer performance on RapidCAD suffers due to the absence of level 1 cache and the bottleneck of the 386 bus. RapidCAD offers minimal improvement in integer performance over a 386DX (typically 10%, at most 35%), but provides substantial improvement in floating point operations (up to 80% faster) for which it was marketed. Floating point performance is mostly improved by moving the FPU onto the CPU core. | 
发表于 2010-3-7 16:57
| 
