gnu@sun.uucp (John Gilmore) (11/02/84)
> > Falcone is setting up something of a "straw man" here. A conscientious > > 68020 design MUST use a data cache, ... > > I find it absurd to state that a XXXXX design MUST do YYYYYY. > Let's face it, you have no idea what the goals of anyone elses design > are. Are we talking cost? Multiple processors? Single tasking? They're both right. Maximum performance in a high speed 68020 system will only come with a cache. Mimimum cost will come without one. There is some area of overlap, where an intelligent design can run relatively fast and have a relatively low cost, and could even beat a badly-designed expensive cache system. It took the market a few years to realize how to build fast 68000 systems (look at old Wicats with 13 wait states!); the same will be true of the 68020 though hopefully not to the same degree. For example, the Apple Mac is a low-performance minimum-cost design. When they come out with a 68020 machine it could continue to alternate memory cycles with the video (cheap, slow) or could have a separate video buffer (medium to fast depending on other things). Given that they have only one row of RAM chips, only at the highest performance levels would they need a cache.