aperez@cvbnet.UUCP (Arturo Perez x6739) (05/08/90)
From article <4698@uceng.UC.EDU>, by dmocsny@minerva.che.uc.edu (Daniel Mocsny): > In article <1990May3.153742.9750@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes: >>The invisible hand of Adam Smith right now is full of Intel x86 machines, >>and I *refuse* to believe that those are the world's best computer >>architecture! > > What architecture is better *right now* for all the Intel x86 > customers? I don't mean what architecture would be better if > only it ran the applications, or if only a mass market existed > to drive the prices down, or if only Joe Smallpotatoes could > get it serviced at the local compustore, or if only > programmers with Henry Spencer's talents worked for nothing, etc. > > Pervasive technologies are almost never cutting-edge, nor > particularly exciting. Mostly, they are standardized. A Now, I'm only familiar with VAXen, 680x0 and a little Sparc so please understand my bias :-) What I never understood about the widespread use of the Intel processors (x86) is that it is so widespread due to the "popularity" of 1 (that's right, count'em ONE) computer: IBM's PC-DOS machine. VAXen are proprietary and yet there ARE quite a few of them out in the market. MC680x0 processors are in Suns, Apollos, Macintoshes, Masscomps (maybe Concurrent now), BBN's original Butterfly, and I'm sure many more. So, I don't think arguments about the popularity of the Intel processor carry much weight (although I would be a fool to say that the popularity of DOS machines doesn't hold much weight). How many people buy cars based on the exact engine model under the hood? Arturo Perez ComputerVision, a division of Prime aperez@cvbnet.prime.com Too much information, like a bullet through my brain -- The Police
msp33327@uxa.cso.uiuc.edu (Michael S. Pereckas) (05/10/90)
aperez@cvbnet.UUCP (Arturo Perez x6739) writes: [quote deleted] >Now, I'm only familiar with VAXen, 680x0 and a little Sparc so please >understand my bias :-) >What I never understood about the widespread use of the Intel processors >(x86) is that it is so widespread due to the "popularity" of 1 (that's right, >count'em ONE) computer: IBM's PC-DOS machine. Well... the fact that there are a thousand clone-makers helped. The IBM name helped a lot to get things started. IBM has to a significant extent become irrelevant now. >VAXen are proprietary and yet there ARE quite a few of them out in the market. >MC680x0 processors are in Suns, Apollos, Macintoshes, Masscomps (maybe >Concurrent now), BBN's original Butterfly, and I'm sure many more. >So, I don't think arguments about the popularity of the Intel processor carry >much weight (although I would be a fool to say that the popularity of DOS >machines doesn't hold much weight). So does it cary weight or not? DOS and the programs that run under it are married to the Intel 86s. Ports to other architectures are a lot harder under DOS than under UNIX. Even if DOS ran on other chips, the programs are not easily portable. The conoptions needed to run with segments and with an "OS" that is essentially useful only for disk I/O (for all other I/O you are on your own, if you want tolerable performance) make portable code difficult. Sure there are programs that are available on other platforms, but those are only the biggest and most popular ones. >How many people buy cars based on the exact engine model under the hood? Most of use don't, but no matter what type of engine a car has (within reason) it can use the same roads. The cpu and surrounding machine determine to a large extent what software you can use in the world of PCs. While some major packages are available on multiple platforms, but a huge number of excellent programs are available only for IBM PC clones. The PC compatables are popular with actual users because of the large software base and the fact that machines are available at excellent prices, thanks to the clones. Programmers hate the 80x86, but they program for it because of the size of the market. Architecture types (like me) hate it for many of the same reasons, but I'm using one now, because I can afford it and it does what I need. The invisable hand is working, where it can. You don't see a lot of UNIX workstations using the Intel 80x86s. The RISC chips have tremendous advantages, but that is only because (most) UNIX workstations are not expected to run any of the thousands of IBM PC binaries. For machines that are, the invisable hand has to fight the invisable boat-anchor of binary compatability with an architecture rooted in the 8080. -- Michael Pereckas * InterNet: m-pereckas@uiuc.edu * NovaNet: m pereckas / imsa89 / cerl + My opinions are mine (no one else wants them, that's for sure). + + I will accept no criticism of my spelling of any words not in the OED2 +
hascall@cs.iastate.edu (John Hascall) (05/10/90)
In article <293> aperez@cvbnet.UUCP (Arturo Perez x6739) writes: }From article <4698>, by dmocsny@minerva.che.uc.edu (Daniel Mocsny): }> In article <1990May3.153742.9750> henry@utzoo.uucp (Henry Spencer) writes: }>>The invisible hand of Adam Smith right now is full of Intel x86 machines, }How many people buy cars based on the exact engine model under the hood? Wasn't GM sued for putting Chevy engines in Oldsmobiles? John Hascall / hascall@atanasoff.cs.iastate.edu ISU Comp Ctr
gerry@zds-ux.UUCP (Gerry Gleason) (05/30/90)
In article <LXL34DG@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: >> Are people going to let this go uncontested? I've *never* seen a system >> that couldn't be improved. IBM may dictate the interface - but do they >> dictate the implementation? >Yes. That is far from clear. At present, the AT bus standard has been all but abandoned by IBM, but the PC compatible computer and board companies are still building a lot of hardware based on this bus, and are designing new lines of systems and boards for an extension to the original bus. IBM chose to design a completely new and incompatible bus, MCA, and the clone makers, etc. got together with EISA which is every bit as good on the high end and somewhat compatible (obviously a 32 bit board won't work with an old bus, but you get to keep all your I/O boards). The market hasn't chosen a clear winner, but I'd put my money with EISA in the long run. >Ask the people with Victor 9000s, or the early TI machines, what happens if >you stray too far from IBM's design. Since I worked for Victor back in the beginning, I'll set the record straight on this. Victor didn't stray from IBM's design, they designed a better machine at the same time that IBM was designing their first PC. The first Victor 9000's shipped in late spring/early summer of 1981, a bit behind IBM, but the machine had a lot more features for less money. What Victor did not do is begin right away to design the next machine, and the next one, etc. It would have been pretty easy to redesign the 9000 with a dual format disk controller (IBM standard tiny and a 1.2M format of their own), and with a compatible bus. Next would have been an AT class machine, and so on. But then, hindsight is 20/20. Gerry Gleason