Path: utzoo!mnetor!uunet!husc6!mit-eddie!killer!chasm From: chasm@killer.UUCP (Charles Marslett) Newsgroups: comp.sys.ibm.pc Subject: Re: Extended Memory (and 80286) Message-ID: <3890@killer.UUCP> Date: 20 Apr 88 13:05:54 GMT References: <4710@bcsaic.UUCP> <6853@j.cc.purdue.edu> <1119@maccs.UUCP> <4704@teddy.UUCP> Distribution: na Organization: The Unix(R) Connection, Dallas, Texas Lines: 62 Keywords: ps2 model 30 Summary: P9 and other intel flops (not megaflops) In article <4704@teddy.UUCP>, jpn@teddy.UUCP (John P. Nelson) writes: > On a different subject, is anyone else hoping that OS/2 flops as strongly > as I am? I don't have any particular prejudice against Microsoft: what > I mean, is that I would hope that ANY operating system based on the > 80286 processor would fail, since the architecture is exceedingly > brain-damaged (I thought the 8088 was bad until I looked at the 80286!) > And I thought the '286 was bad til I tried to do some work with the '386. Seriously, the '386 is a great chip -- its a somewhat faster '286 with a neat feature to speed up MSDOS execution under a real OS. Mostly, though, it is not much faster than an equivalent clock speed '286 in real applications. My cursory examination of real software shows up branches every 15-30 bytes so the 32-bit wide fetch helps less than you'd expect (especially when you factor in the typical size of tight loops -- they are small so they are fast so the '386 buys you the least performance increase). 10 MHz '286 --> 16 MHz '386 resulted in a 60% increase in speed for a compute bound screen graphics generation program and a RAM based video BIOS doing text mode screens. That's not as good as the ratio of the clock speeds (the '286 is no waits, the '386 is Compaq's -- 1.2 waits?). And data accesses are faster only for alligned 32-bit operations -- once in a blue moon from the code I've seen. On top of that, the '386 looks like it has more bugs than features, it seems -- see Microsoft's MASM 5.0 manual. At least the newer '286s really do what they are documented to do. > I still don't see why IBM/Microsoft couldn't have based OS/2 on the > 80386 instead. Intel announced their P9 (80386 in a 80286 compatible > package) pretty early in the game: IBM could have supported their > AT base by insisting that customers upgrade their processor before > upgrading their operating system. The P9 is still not available to the unwashed masses, and all the rumors I have heard indicate it is not a drop-in replacement. The P9 is just an 8088-style '386. That is, it lets you build a 16-bit '386 computer (a cheap 386 box?) but not upgrade a current '286 box. It is not necessarily functional with an AT bus without custom glue either (comments from the guys at Faraday or C&T?). I'd like to hear comments on both of these issues from Intel (officially if possible, rumor if not!) -- this is as bad as predicting IBM announcements now. > > Certainly, I believe that bringing out the PS/2 models 50 and 60 with > 80286 processors was a serious mistake, wasn't it? Couldn't they have > put slower memory and slow 80386 processors into them, and still been > able to differentiate the high-end model 80? And make them that much more expensive than they are now. . . don't you like IBM at all? More seriously, again, the 386 (except for the P9 chip) is a 32-bit chip and that alone makes the 386 boxes more expensive, the 386 is sole sourced (and Intel can't or won't make enough of them) so it would limit sales if the box really did sell. There are lots of reasons for not using the 386 -- and lots of reasons I think the 386/286 ratio may remain smaller than most people think. 68000s are better than 386s for programming, in performance, and it uses more memory (all IBM pluses), but most code is for the 8088 today. Sorry. > -- > john nelson Charles Marslett chasm@killer.UUCP