spp@zabriskie.berkeley.edu (Steve Pope) (04/24/91)
Perhaps the obvious is worth stating in this discussion -- not only is the 68000 a lower-cost part then the 68020, but it has a 16-bit bus vs. the 68020's 32 bit bus, and a somewhat simpler set of control signals. Therefore, for a low end system, the parts count and parts cost will be considerably lower for a 68000 system. Therefore, which part to use primarily depends on whether you're aiming for a 16-bit or 32-bit system. As I recall the 68010 is a 68000 with some hack that some VM system designers demanded. steve
byron@cc.gatech.edu (Byron A Jeff) (04/24/91)
In article <12964@pasteur.Berkeley.EDU> spp@zabriskie.berkeley.edu.UUCP (Steve Pope) writes: > >Perhaps the obvious is worth stating in this discussion -- >not only is the 68000 a lower-cost part then the 68020, >but it has a 16-bit bus vs. the 68020's 32 bit bus, and >a somewhat simpler set of control signals. Therefore, for a >low end system, the parts count and parts cost will be >considerably lower for a 68000 system. I'd have to disagree with on on this point. I've built systems with each of the 68008, 68000,68010,68020 and '020 is much easier to hook up because of the dynamic bus sizing. The parts count is actually less than a 68000 or '010 based system because everything can be done on a single 8 bit bus (Including EPROM, RAM, and I/O). Also I was able to pick up 2 20 Mhz '020s for $25 apiece (used). Considering I get a more than %25 increase in speed using only half the bus width (due to the effects of double bus speed, 3 instead of 4 cycles, and the instruction cache) and reclaim more board space, I'd use a '020 any time. In my opinion a '020 system can be costly if you choose to exploit all of the features available. However in minimal and small configured systems a '020 can beat out 68000 anytime. My '020 system consists of the following: 20 Mhz '020 1 32K battery backed static ram (70ns running at no wait states) PC parallel port interface for program loading PC keyboard interface LCD driver interface 1 Mbyte of dynamic ram (70ns running at no wait states, configured as as 1 32 bit x 256K bank) All this fits on a 6 x 8 inch perfboard with enough space to fit another Mbyte of DRAM on board. To exploit the 32 bit data bus all I need is a single 16L8 pal to decode the SIZE and A1 and A0 address lines. Also note that since the '020 does not force the use of the entire data bus that all the program can fit in a single ram without having to be split. This simplifies the programming of the board. > >Therefore, which part to use primarily depends on whether >you're aiming for a 16-bit or 32-bit system. > The '020 (and '030) can give the best of both (and also 8 bit) because of dynamic bus sizing. >steve BAJ --- Another random extraction from the mental bit stream of... Byron A. Jeff - PhD student operating in parallel! Georgia Tech, Atlanta GA 30332 Internet: byron@cc.gatech.edu
spp@zabriskie.berkeley.edu (Steve Pope) (04/25/91)
In article <27178@hydra.gatech.EDU> byron@cc.gatech.edu (Byron A Jeff) writes: #>In article <12964@pasteur.Berkeley.EDU> spp@zabriskie.berkeley.edu.UUCP (Steve Pope) writes: #>> #>>Perhaps the obvious is worth stating in this discussion -- #>>not only is the 68000 a lower-cost part then the 68020, #>>but it has a 16-bit bus vs. the 68020's 32 bit bus, and #>>a somewhat simpler set of control signals. Therefore, for a #>>low end system, the parts count and parts cost will be #>>considerably lower for a 68000 system. #> #>I'd have to disagree with on on this point. I've built systems with each #>of the 68008, 68000,68010,68020 and '020 is much easier to hook up because #>of the dynamic bus sizing. The parts count is actually less than a 68000 or #>'010 based system because everything can be done on a single 8 bit bus #>(Including EPROM, RAM, and I/O). #> I'm surprised to hear that the dynamic bus sizing applies to instruction fetches. Motorola's documentation, so far as I can tell, specifically talks about dynamic sizing for operand transfers only (for the 68020). If the dynamic sizing also applies to fetches then I agree with the above response. steve
byron@cc.gatech.edu (Byron A Jeff) (04/25/91)
In article <13017@pasteur.Berkeley.EDU> spp@zabriskie.berkeley.edu.UUCP (Steve Pope) writes: >In article <27178@hydra.gatech.EDU> byron@cc.gatech.edu (Byron A Jeff) writes: >#>In article <12964@pasteur.Berkeley.EDU> spp@zabriskie.berkeley.edu.UUCP (Steve Pope) writes: >#>> >#> [ Steve writes of a 68000 being easier to interface than a 68020. >#> I responded that '020 are easier to work with because of dynamic >#> bus sizing. ] > >I'm surprised to hear that the dynamic bus sizing applies to >instruction fetches. Motorola's documentation, so far as I >can tell, specifically talks about dynamic sizing for operand >transfers only (for the 68020). If the dynamic sizing >also applies to fetches then I agree with the above response. > >steve Any device can specify its bus width (8,16, or 32 bits) for each and any fetch, instruction fetches included. Ask your local Motorola representative for the Application Note "A Minimum Configuration 68020 System". It contains a 68020 system that only has 8 bit ports for everything (EPROM, RAM, and IO). BAJ --- Another random extraction from the mental bit stream of... Byron A. Jeff - PhD student operating in parallel! Georgia Tech, Atlanta GA 30332 Internet: byron@cc.gatech.edu