cheong@uicsrd.CSRD.UIUC.EDU (11/22/85)
/* Written 11:49 am Nov 20, 1985 by brad@looking.UUCP in uicsrd:net.arch */ /* ---------- "Re: What if IBM Had chosen the 6800" ---------- */ In article <428@ecn-pc.UUCP> wdm@ecn-pc.UUCP (Tex) writes: >In article <456@looking.UUCP> brad@looking.UUCP (Brad Templeton) writes: >>Ok, just what would have happened under these circumstances? I won't >>say that this is gospel truth, but there is some evidence for it: >> >>1) The 68000 was only 16 bits at the time, no 68008 was to be had for >> several years. This would have resulted in either special bus >> multiplexing hardware (slow) or a 16 bit bus. This all adds up to >> *cost*. The PC then would cost what the IBM-AT costs now. The >> higher cost equipment means fewer people buy the machine, and very >> few non-business customers buy it. How many hobbyists have ATs? >> Result, little hacking in the mass market. > > Get serious!! Does the macintosh cost as much as the AT? The Amiga? > The ST? No of course not. Face it, the bus interface hardware accounts > for a tiny fraction of the overall cost. Given the cost/performance > ratio, adding a sixteen bit bus would make a lot of sense. > Please don't tell me to "get serious." I am quite serious. IBM is not Atari. IBM doesn't feel its place is to make the cheapest micro. Their goal is to make the best supported one - that's their reputation. The IBM-AT has a 16 bit processor and support chips, 16 bit bus with 8 slots, 1.2 meg floppy, nice keyboard and a disk controller card. The same computer with a 68000 in it would cost just as much, or more than the 286 version. How can you seriously suggest they would charge less? >> >>2) CP/M Software (8080) is given no place to migrate. CP/M programs and >> 6502 programs all have a high degree of processor loyalty that C programs >> for 16 bit CPU's don't. You *can't* port a cp/m program to a 68000 >> without a total rewrite. (This may be a good thing!) What this means >> is that CP/M doesn't die, and maintains strength the same way the Apple >> ][ and Commodore Architectures hang on. The result: CP/M and the 6502 >> are the only serious contenders against IBM. > > Is a sizable percentage of ms-dos software old cp/m software? It > would surprise me if it were. Or maybe I should say it would sur- > prise me if it weren't rewritten in a major way, seeing as how the > operating systems are not at all alike. I would have rather they > were rewritten for the 68000 environment. No, lots of new software (mostly in HLLs) has been written for ms-dos. The point I made is that all the wide-selling CP/M software was available quickly for the 8086, providing a quick migration. The software was not rewritten, it was ported, which is to say they made adjustments instead of totally reworking the project. >> >> [This is the most serious consequence. In order to advance the industry >> to a new generation of architectures, you must *kill* the previous >> generation. > > I guess that explains why there are no ibm 370- type systems around > anymore. What? You say there are? Well I guess that means we haven't > advanced since the mid-60's. There is absolutely no need to "kill" > off the previous generation of architectures. In fact, this comment > contradicts your point 2. You miss the point. The 360 has been killed, in that there are almost none of them out there. You can still run 360 and 370 programs on the more recent machines, but the software headache, "I have to get this to run on the 360 if I want it to sell" is gone. By "kill" I mean eliminate the old processors, not the ablility to run their software. > >> >>3) 68000 programs are a lot larger than 8086 programs. A lot of programs >> that might have shown up don't fit. On the plus side, this means a >> bit of a push for larger memory, but only to achieve the same results. > > If you mean there are alot of large 68k application programs out there, > then I agree. If you mean that a program for a given task is larger on > the 68k than on the 808/8/6 then I disagree totally. Look at any > of a number of benchmarks that have been written. See other notes. Please show your benchmarks. I have yet to see a 68000 Unix box that runs reasonably with much less than 1 meg of ram. I used to own a Z-8000 Unix machine (ONYX) that was a very reasonable single user system with 256K and a 3 user system with 512K. Coherent ran at PDP-11 levels on an IBM-PC with an 8088! Xenix on the 286 performs at a similar level to 68000 systems costing more than twice as much, and can do it with less memory. In fact, since we know IBM isn't a "price cutter", isn't it amazing that they make by far the cheapest machine for Unix? (forgetting the clones, for the time being) > >> >>4) Unix on micros is delivered a real blow. Chances are the IBM 68000 >> has no memory managment. It's expensive and slows things down. >> This means no Unix on this one. Sure there will be Unix for more >> expensive 68000 boxes with no MMUs, but they will always be there. >> Other multi-tasking systems that need an MMU like QNX are also hurt. > > The IBM PC has no memory management. What is your point here? It makes Incorrect > no sense. You are incorrect in saying that because this imaginary > machine has no mmu, there would be no unix (as you point out > in the next sentence) since there are versions of unix right now > that do not need memory management. Anyway, who said our ibm 6800 Show me these versions of Unix... If they exist, don't they have to make very harsh restrictions on user code, ie. making it position independent? > has to run unix? Why not os-9, or any or a bunch of operating systems? I was making a point. Perhaps OS-9 would do the job. I was just saying, "If you value unix, here's something to consider." >> >>If your goal is to make most people use a "nice" architecture (where "nice" >>is subjective but usually means "easy to get programs running under") then >>you must do three things: >> >> 1) Have a nice architecture! > > Ok, I suggest the 68000 here. > >> 2) Get people to stop using the old (not-nice) architectures > > Not all that many people were using the 808X before IBM anyway. 8080 (Z-80) and 6502 were almost the whole industry prior to IBM. Today there are more 6502 computers sold than any other, inluding IBM. > >> 3) Get people to use the nice architecture > > Anything that IBM sells is going to be used, if they support it like > they did the PC. >> >>#1 is engineering. #2 and #3 are marketing. To reach your goal, they >>CAN'T be ignored. You just can't wish them away. > > Who is wishing them away? My scenario is what if IBM chose the 68000 > instead of the 8088 for their PC. Am I to believe that they wouldn't have > marketed it as ardently as they did the 8088-based PC? I'm not saying it would have been bad to see IBM choose the 68000. In fact, they *did* market a 68000 machine at the same time as the PC. They mostly aimed it as a "lab" machine. That's one reason it didn't sell well. The others are (surprise, surprise) that it cost like an AT and couldn't run any of the old software! It couldn't compete with the price of the 8088 machine. It wasn't pure goodness for IBM to choose the 8086. But it would not have been all sweetness and light if they had chosen the 68000, and that's the notion I am trying to correct. -- Brad Templeton, Looking Glass Software Ltd. - Waterloo, Ontario 519/884-7473 /* End of text from uicsrd:net.arch */