From: Rui Pedro Mendes Salgueiro Newsgroups: comp.arch,alt.folklore.computers Subject: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: 8 Jul 1999 05:40:06 +0100 Organization: Universidade de Coimbra Lines: 76 Message-ID: <7m1a36$d3f$1@rena.mat.uc.pt> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> NNTP-Posting-Host: localhost X-Trace: rena.mat.uc.pt 931408810 13424 127.0.0.1 (8 Jul 1999 04:40:10 GMT) X-Complaints-To: usenet@rena.mat.uc.pt NNTP-Posting-Date: 8 Jul 1999 04:40:10 GMT User-Agent: tin/pre-1.4-981114 ("The Watchman") (UNIX) (FreeBSD/2.2.7-RELEASE (i386)) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!rccn.net!rena.mat.uc.pt!not-for-mail I crossposted to alt.folklore.computers, since we are now talking about computers 3 or 4 generations back. Stephen E. Halpin wrote: > On Sat, 03 Jul 1999 23:35:31 +0200, Bernd Paysan wrote: >>Bill Todd wrote: >>> Not having been very focused on x86s at that time, my memory is sketchy in >>> this area, but unless the 386 was significantly slower than existing 286s in >>> executing existing applications *and* faster 286s were still in the pipe >>> such that this disparity would continue for some time, the situations may >>> have similarities but aren't identical: >>IIRC, the 386 wasn't faster than the 286 for comparable codes (until >>Intel decided to put the 386SX as replacement for the 286, and ramp up >>the clock frequency of the 386), and for a long time, people still >>bought 286 machines. > I vaguely remember upgrading 12MHz 286s to 16MHz 386s with a reasonable > cache, and no one looked back. Yes, I think the 386 had a performance advantage since day one (IIRC, there was also a 20 MHz and later a 25 Mhz). I think that, for a while, 286s (heck, even 8086s *) were still being sold due to the simple fact that 386s were expensive. But the other chipmakers (AMD and others) which had second-source agreements with Intel were not very willing to drop the 286. They pushed it to 20 MHz (CMOS implementation, while the Intel 12 MHz was TTL ?), and so were kind of competitive with the 386. I remember some comparative review of such computers in one magazine (probably Byte). But I think everyone knew that the 286 was a dead-end and those 20 MHz 286s did not convince many people. Even if one was not using software which made use of the 386, one expected to eventually use it before the computer was retired. I don't think we have any 286 still in use (maybe one which is connected to the PBX ?) but there are 386s which were passed down the hierarchy when the original owners upgraded and are only now being retired. > Even if the ALU timing was the same, the > 386 could move more data and instructions to and from memory, > and with the cache the latency was reduced as well. BTW, about cache, since at the time most PC users were still blocked by the 640 KB limit, some manufacturers had the idea of making systems with 640 KB in SRAM. You don't need cache if all your memory is in the same technology as cache memory. * around 1988 or 1989 our departamental computer lab bought the first PC for use by students (there were other PCs in use in other parts of the department, including a number of ATs (IBMs with EGA ! A big thing at the time) for faculty and a number of PCs in a separate students lab, but we were then a strictly un*x lab, with a lot of terminals, many of which are still in use (the others mostly died)). I don't remember for sure if the 386 was already available, but 286s were. But due to the usual budget limitations, we had to settle for a 8088 (or was it a 8086 ? I am not sure) with a monochrome monitor (Hercules graphics ?). It was a Philips because Philips had had the very good idea of making a computer in which one could fit both 3.5" and 5.25" disquetes drives. Long after that computer was too slow for anything else, it was still being used for exchanging files from one format to the other and to and from the Unix systems (via ethernet at that time, but first it was via RS-232). The next PC we bought was a 486 in 1992 to run BSDI. It is still running. That one and later ones have not get really obsolete if you think of them as X-terminals with some local commands. -- http://www.mat.uc.pt/~rps/photos/ an ex-tifoso since 95/11/13 Mark Sandman - Morphine, RIP (July 3th, 1999, Italy) .pt is Portugal| `Whom the gods love die young'-Menander (342-292 BC) Europe | Villeneuve 50-82, Toivonen 56-86, Senna 60-94 ###### From: Terje Mathisen Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Thu, 08 Jul 1999 09:01:51 +0200 Organization: Hydro Lines: 40 Message-ID: <37844CDF.88A9BA4D@hda.hydro.com> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> NNTP-Posting-Host: 136.164.10.89 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.06 [en] (WinNT; I) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.belnet.be!news1.carrier1.net!newsfeed.wirehub.nl!148.122.208.67.MISMATCH!newsfeed1.online.no!newsfeed.online.no!hydro.com!not-for-mail Rui Pedro Mendes Salgueiro wrote: > BTW, about cache, since at the time most PC users were still blocked > by the 640 KB limit, some manufacturers had the idea of making systems > with 640 KB in SRAM. You don't need cache if all your memory is in > the same technology as cache memory. Cheetah started making memory boards for (probably) 286 machines, using faster ram than the motherboard version. To get maximum speed you'd disable all (or as much as possible) of the factory-installed ram so you could use the faster add-on version instead. > > * around 1988 or 1989 our departamental computer lab bought the > first PC for use by students (there were other PCs in use in > other parts of the department, including a number of ATs (IBMs > with EGA ! A big thing at the time) for faculty and a number of > PCs in a separate students lab, but we were then a strictly > un*x lab, with a lot of terminals, many of which are still in > use (the others mostly died)). > > I don't remember for sure if the 386 was already available, but Compaq delivered the first 386 machines around 1986, I remember getting one of the first available here in Norway, running at a blazing 16 MHz. By 88-89 the 286 was pretty much gone. The most important (???) benefit of 32-bit registers was that it allowed me to calculate more digits of pi, since my series evaluation of arctan required division using divisors greater than 64K. :-) Even later, using Watcom's Dos extended to allow a true 32-bit program, that same program turned out a nice million digits. Terje -- - Using self-discipline, see http://www.eiffel.com/discipline "almost all programming can be viewed as an exercise in caching" ###### From: Rui Pedro Mendes Salgueiro Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Followup-To: alt.folklore.computers Date: 8 Jul 1999 12:09:55 +0100 Organization: Universidade de Coimbra Lines: 78 Message-ID: <7m20u3$4kp$1@rena.mat.uc.pt> References: <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <37844CDF.88A9BA4D@hda.hydro.com> NNTP-Posting-Host: localhost X-Trace: rena.mat.uc.pt 931432199 4762 127.0.0.1 (8 Jul 1999 11:09:58 GMT) X-Complaints-To: usenet@rena.mat.uc.pt NNTP-Posting-Date: 8 Jul 1999 11:09:58 GMT User-Agent: tin/pre-1.4-981114 ("The Watchman") (UNIX) (FreeBSD/2.2.7-RELEASE (i386)) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!rccn.net!rena.mat.uc.pt!not-for-mail In alt.folklore.computers Terje Mathisen wrote: > Rui Pedro Mendes Salgueiro wrote: >> * around 1988 or 1989 our departamental computer lab bought the >> first PC for use by students (there were other PCs in use in According to our records, it was in 1989. Just now, I took a look at him (at its remains, more exactly. The power supply, the 20 MB hard disk and its controller, the drives, were all removed for use in other machines) and it is a 8088-2 ! The socket for the 8087 is (as usual) empty. >> other parts of the department, including a number of ATs (IBMs >> with EGA ! A big thing at the time) These were from 1986. >> I don't remember for sure if the 386 was already available, but > Compaq delivered the first 386 machines around 1986, I remember getting > one of the first available here in Norway, running at a blazing 16 MHz. Around November 1986, was when the Amstrad 1512 was arriving here (I remember a lot of people being quite worried about the computers they had in stock, having suddendly got very hard to sell, since the Amstrad was half their price). The Amstrads (the 1512, first and later the 1640) were the computers which made PCs common in Portugal. Although at the same time, the IBM ATs were already available (maybe even not uncommon), the Amstrads (and other clones) were more significant in installed base. > By 88-89 the 286 was pretty much gone. Hmm, I think it was the summer of 1990 when I borrowed (from a shop which belonged to some friends) a computer to do some schoolwork at home. It was a 286 (brand new), which I think was still the common PC at the time. Maybe the situation in richer countries was different. > The most important (???) benefit of 32-bit registers was that it allowed > me to calculate more digits of pi, since my series evaluation of arctan > required division using divisors greater than 64K. :-) > Even later, using Watcom's Dos extended to allow a true 32-bit program, > that same program turned out a nice million digits. In 1990, one of my teachers and me, we presented at some conference on optimization "An Incremental Chain Algorithm for the Maximal Flow Problem". One interesting bit was that our test programs (which were only used to solve random-generated problems, meaning, artificial problems with only a tenuous relation to anything real) were run in a Unix machine (Dec System 3100*), with 16 MB of RAM. And we really used lots of RAM, since the algorithm (and the computer, a R2000 processor) were fast enough that we needed large problems to have measurable times. In contrast a women who worked for a big chemical factory, presented her program to schedule the transport of chemical fertilizers (the factory was big enough to actually own railways waggons). She could only optimize for a few months at a time, since she was using DOS. She was planning on upgrading to Windows 3.0 (which IIRC, extended the limit to 16MB (286 limit)) to overcome that limitation. (For some reason, it seems that DOS extenders never were popular around here.) I found it quite curious that the "real world" was using inadequate technology, when there were several solutions available. * That Dec System 3100 is still working and it even does some useful stuff. And last year it was still our main mail server (as such I thought of mentioning it in that thread about the oldest machine used to connect to the Net). (This has become a little irrelevant to comp.arch so I have set followups to alt.folklore.computers.) -- http://www.mat.uc.pt/~rps/photos/ an ex-tifoso since 95/11/13 Mark Sandman - Morphine, RIP (July 3th, 1999, Italy) .pt is Portugal| `Whom the gods love die young'-Menander (342-292 BC) Europe | Villeneuve 50-82, Toivonen 56-86, Senna 60-94 ###### Path: chonsp.franklin.ch!usenet From: Neil Franklin Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: 08 Jul 1999 20:50:05 +0200 Organization: My own Private Self Lines: 117 Sender: neil@chonsp.franklin.ch Message-ID: <6ulncr2cyq.fsf@chonsp.franklin.ch> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> X-Newsreader: Gnus v5.3/Emacs 19.34 Rui Pedro Mendes Salgueiro writes: > > >>IIRC, the 386 wasn't faster than the 286 for comparable codes For 8088 code (as used on 99.x% of DOS programs they were roughly equals at equal clock. So the 12->16MHz made about 33%. The real strength of the 386 was its 32 bit bus and 32 bit at a time instructions, which were promptly ignored for years. > >>(until > >>Intel decided to put the 386SX as replacement for the 286, and ramp up > >>the clock frequency of the 386), 386SX was about 5-10% below 386DX for DOS programs, but about 20% cheaper. > >>and for a long time, people still > >>bought 286 machines. They were a _lot_ cheaper. In 1988 when I got my first PC a 286-12 was about 30% more than 8088-10. OTOH 386-16 was near double the 286-12. I payes abotu half the difference 286->386 for an T800-20 Transputer card for in the PC and got 486-25 speed out of that. > > I vaguely remember upgrading 12MHz 286s to 16MHz 386s with a reasonable > > cache, and no one looked back. Cached 386-16 were a lot faster than 286-12 because they did not suffer them pesky wait states which killed about 1/5 of the 286-12s speed. A cached 286-12 would have been interesting. A 286-12 with RAMs that allowed an 286-8 to run full was as fast as an 286-19 with the better RAMs that required for full, but cheaper due to the slower RAMs. OTOH many 386-16 were also cache-less, the same many 386-20 and some 386-25. Only with 386-33 did caches become standard. Note that all 386 chips have no internal caching whatsoever, cache was entirely an motherboard feature and fairly costly. > Yes, I think the 386 had a performance advantage since day one > (IIRC, there was also a 20 MHz and later a 25 Mhz). I think that, It started with 386-16 (later renamed to 386DX16). I still have such an chip marked 368-16 in operation (as backup server). > But the other chipmakers (AMD and others) which had second-source > agreements with Intel were not very willing to drop the 286. They > pushed it to 20 MHz (CMOS implementation, while the Intel 12 MHz > was TTL ?), and so were kind of competitive with the 386. I remember Intel chips were NMOS since the 8080. Before (4004, 8008) they were PMOS, they never made TTL processors. They switched to CMOS with the 486 IIRC. > But I think everyone knew that the 286 was a dead-end and those 20 MHz > 286s did not convince many people. Even if one was not using software > which made use of the 386, one expected to eventually use it before Hmmm. I still knew times when 286-20s outsold all 386 flavours together. And that was in the days when I worked at an PC dealer :-). > (maybe one which is connected to the PBX ?) but there are 386s which > were passed down the hierarchy when the original owners upgraded > and are only now being retired. Mine is still in operation, sort of. It hangs randomly after 10min-2hours after boot, after crashing the AHA-1542B SCSI its boot disk is on, due to an broken DMA chip (Chips&Technology 82C206, anyone got one laying around?) > > Even if the ALU timing was the same, the > > 386 could move more data and instructions to and from memory, > > and with the cache the latency was reduced as well. But the cache was optional on all earlier models. My 386-16 has cache, but it was in its days (1989) designed to be used in an server (Novell Netware 2.x thingy). > with 640 KB in SRAM. You don't need cache if all your memory is in > the same technology as cache memory. Outch, that must have been expensive. Mine has a scant 32k of cache. > * around 1988 or 1989 our departamental computer lab bought the > first PC for use by students (there were other PCs in use in > other parts of the department, including a number of ATs (IBMs > with EGA ! A big thing at the time) My 286-12 came with an Genoa 800x600 Super-EGA (16 colours). VGA only came into existance with the 1988 PS/2s and became widely cloned about 1990. > I don't remember for sure if the 386 was already available, but About 1987 IIRC. The chip came 1984/5 but it took 2/3 years (!) to get into PCs because everyone was waiting for IBM to lead. In the end Compaq made their AT/386 and so set the standard. -- Neil Franklin, Nerd, Geek, Unix Wizzard and Guru, Hacker, Mystic neil@franklin.ch.remove http://neil.franklin.ch/ Computer: toy that speeds work, so you have more time to play with it ###### From: tmg@BAIT.bc.sympatico.ca (Trent Gamblin) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> Reply-To: tmg@BAIT.bc.sympatico.ca Followup-To: alt.folklore.computers Message-ID: User-Agent: slrn/0.9.5.6 (UNIX) Lines: 7 Date: Fri, 09 Jul 1999 06:02:16 GMT NNTP-Posting-Host: 209.52.208.62 X-Complaints-To: news@bctel.net X-Trace: news.bctel.net 931500136 209.52.208.62 (Thu, 08 Jul 1999 23:02:16 PDT) NNTP-Posting-Date: Thu, 08 Jul 1999 23:02:16 PDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!isdnet!logbridge.uoregon.edu!newsfeed.direct.ca!newsfeed.bctel.net!news.bctel.net!tmg Neil Franklin wrote: > I payes abotu half the difference 286->386 for an T800-20 Transputer > card for in the PC and got 486-25 speed out of that. What the heck is a T800-20 Transputer card? Sounds like something you'd make up, when arguing about who's computer is better :) ###### From: dpeschel@u.washington.edu (Derek Peschel) Newsgroups: alt.folklore.computers Subject: Re: 286-386 transition Date: 9 Jul 1999 07:29:03 GMT Organization: University of Washington, Seattle Lines: 30 Message-ID: <7m48bv$ia6$1@nntp1.u.washington.edu> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2c NNTP-Posting-Host: saul1.u.washington.edu X-Trace: nntp1.u.washington.edu 931505343 18758 (None) 140.142.17.37 X-Complaints-To: help@cac.washington.edu NNTP-Posting-User: dpeschel X-Original-Newsgroups: comp.arch,alt.folklore.computers Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!logbridge.uoregon.edu!news.u.washington.edu!dpeschel In article , Trent Gamblin wrote: >Neil Franklin wrote: >> I payes abotu half the difference 286->386 for an T800-20 Transputer >> card for in the PC and got 486-25 speed out of that. > >What the heck is a T800-20 Transputer card? Sounds like something >you'd make up, when arguing about who's computer is better :) The Transputer is definitely not made up. In essence a Transputer is a chip containing a small and relatively simple CPU and some communications ports. The CPU is a bit like RISC except that it has variable instruction lengths; it also has a stripped-down priority system in hardware (and maybe a few other exotic features which I forget). The communications ports allow the chip to be connected to other Transputer chips via short wires. In reality a Transputer is a board containing one or more of those CPUs plus memory and possibly some I/O hardware (e.g., disk controllers, graphics). Transputer boards without the fancy I/O hardware rely on a host system such as an IBM PC clone. The Transputer architecture's strength is its parallel-processing ability. Inmos (the designer and manufacturer of the Transputer) also designed a programming language called occam to take advantage of the parallel processing. -- Derek ###### From: richard@cogsci.ed.ac.uk (Richard Tobin) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: 9 Jul 1999 11:30:00 GMT Organization: HCRC, University of Edinburgh Lines: 22 Message-ID: <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> NNTP-Posting-Host: doyle.cogsci.ed.ac.uk X-Trace: pc-news.cogsci.ed.ac.uk 931519800 75535 129.215.110.29 (9 Jul 1999 11:30:00 GMT) X-Complaints-To: usenet@cogsci.ed.ac.uk NNTP-Posting-Date: 9 Jul 1999 11:30:00 GMT X-Newsreader: trn 4.0-test72 (19 April 1999) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news1.sunrise.ch!news.imp.ch!netnews.globalip.ch!news-lond.gip.net!news.gsl.net!gip.net!nntp.news.xara.net!xara.net!server6.netnews.ja.net!newsfeed.ed.ac.uk!news.cogsci.ed.ac.uk!not-for-mail In article , Trent Gamblin wrote: >What the heck is a T800-20 Transputer card? Sounds like something >you'd make up, when arguing about who's computer is better :) The Transputer was a processor made by Inmos during the 80s (and possibly still made), intended for use in parallel. You could get cards with a bunch of transputers to plug in to PCs and other hosts. Like many "novel architectures" from that period - promising efficient parallelism through some paradigm promoted with ideological zeal (in this case CSP, Communicating Sequential Processes) - it failed to live up to its promise. I'm sure a web search would have turned up this information. -- Richard -- Spam filter: to mail me from a .com/.net site, put my surname in the headers. "The Internet is really just a series of bottlenecks joined by high speed networks." - Sam Wilson ###### From: timothy.mccaffrey@spam2filter.unisys.com.takethisoff (Tim McCaffrey) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: 8 Jul 1999 21:41:55 GMT Organization: A series networking Lines: 32 Message-ID: <7m35v3$dvm$1@mail.pl.unisys.com> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> NNTP-Posting-Host: mccafftm.tr.unisys.com X-Newsreader: WinVN 0.99.9 (Released Version) (x86 32bit) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!isdnet!newsfeed.tli.de!do.de.uu.net!uunet!ams.uu.net!ffx.uu.net!bbnews1!bbnews1.unisys.com!plnews.pl.unisys.com!not-for-mail In article <6ulncr2cyq.fsf@chonsp.franklin.ch>, neil@franklin.ch.remove says... > > >About 1987 IIRC. The chip came 1984/5 but it took 2/3 years (!) to >get into PCs because everyone was waiting for IBM to lead. In the end >Compaq made their AT/386 and so set the standard. > IIRC, Compaq 386 was introduced August 1986, at almost the same time as the official announcement of the 386. The IBM PS/2 was introduced several months later, and was unimpressive. The Compaq 386 used static column DRAM, and no cache. Compaq claimed that this gave them effectively a 0.8 wait state system. I think the system shipped with a 40 Meg MFM drive as standard. Again, IIRC, expansion memory was through a proprietary 32 bit card. Max was probably something like 8 megabytes, or maybe 16. Things like Fast Page Mode DRAM, VGA, cache (in PC processors), IDE and 3.5" drives were still in the future. I don't recall if even SoundBlasters were out yet, I think Adlib cards were available, although expensive ($100-$200). I think Windows 1.0 was introduced the same year. -- Tim McCaffrey NOT speaking for Unisys. ###### From: Terje Mathisen Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Fri, 09 Jul 1999 09:03:26 +0200 Organization: Hydro Lines: 31 Message-ID: <37859EBE.1188E2D6@hda.hydro.com> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m35v3$dvm$1@mail.pl.unisys.com> NNTP-Posting-Host: 136.164.10.89 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Mailer: Mozilla 4.06 [en] (WinNT; I) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!newsfeed.tli.de!news.algonet.se!algonet!newsfeed1.online.no!newsfeed.online.no!hydro.com!not-for-mail Tim McCaffrey wrote: > IIRC, Compaq 386 was introduced August 1986, at almost the same > time as the official announcement of the 386. The IBM PS/2 was > introduced several months later, and was unimpressive. Indeed. April 1987, together with the OS/2 announcement. Do you want to pay USD 3000,- for the priviledge of debugging one of the first betas of OS/2 V1.0? > The Compaq 386 used static column DRAM, and no cache. Compaq > claimed that this gave them effectively a 0.8 wait state system. > I think the system shipped with a 40 Meg MFM drive as standard. 40MB default, 60 MB as an option. A little while later they came out with the _huge_ 130 MB drive. At that point the Compaq w/130MB cost about the same as other vendors wanted to charge for just a 100+ MB hard drive. :-) We bought a bunch of those and used them for NetWare servers (V2.0a and later). Terje -- - Using self-discipline, see http://www.eiffel.com/discipline "almost all programming can be viewed as an exercise in caching" ###### From: Paul DeMone Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Fri, 09 Jul 1999 14:00:39 -0400 Organization: IGS - Information Gateway Services Lines: 34 Message-ID: <378638C7.4C24AB89@igs.net> References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> NNTP-Posting-Host: ttyc1a.ott.igs.net Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Trace: news.igs.net 931543174 3576 216.58.19.106 (9 Jul 1999 17:59:34 GMT) X-Complaints-To: abuse@igs.net NNTP-Posting-Date: 9 Jul 1999 17:59:34 GMT X-Mailer: Mozilla 4.61 [en] (Win98; I) X-Accept-Language: en Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.mathworks.com!news-out.cwix.com!newsfeed.cwix.com!206.191.82.231!rockie.attcanada.net!attcanada!nntp.igs.net!news.igs.net!not-for-mail Neil Franklin wrote: [a good description of cost & performance differences between early generation x86s snipped] > Intel chips were NMOS since the 8080. Before (4004, 8008) they were PMOS, > they never made TTL processors. They switched to CMOS with the 486 IIRC. The 386 was actually first fabricated in a 1.5 um P-well CMOS process. For some historical perspective and chuckles here are some process params (from ICCD '86) Ldrawn 1.5 um Leff 1.0 um Gate oxide thickness 250 A Metal 1 pitch 5 um Metal 2 pitch 6 um VTN 0.6 V VTP 0.9 V SRAM cell size 572 um2 (My god, talk about your bear skins and stone knives ;-) [snip] All opinions strictly my own. -- Paul W. DeMone The 801 experiment SPARCed an ARMs race of EPIC Kanata, Ontario proportions to put more PRECISION and POWER into demone@mosaid.com architectures with MIPSed results but ALPHA's well pdemone@igs.net that ends well. ###### Path: chonsp.franklin.ch!usenet From: Neil Franklin Newsgroups: alt.folklore.computers Subject: Re: 286-386 transition Date: 09 Jul 1999 23:24:57 +0200 Organization: My own Private Self Lines: 88 Sender: neil@chonsp.franklin.ch Message-ID: <6ubtdlsehi.fsf@chonsp.franklin.ch> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2c <7m48bv$ia6$1@nntp1.u.washington.edu> X-Newsreader: Gnus v5.3/Emacs 19.34 dpeschel@u.washington.edu (Derek Peschel) writes: > > >Neil Franklin wrote: > > >> I payes abotu half the difference 286->386 for an T800-20 Transputer > >> card for in the PC and got 486-25 speed out of that. > > > >What the heck is a T800-20 Transputer card? Sounds like something > >you'd make up, when arguing about who's computer is better :) The only "make up" involved consisted of an soldering iron and 800 pins to be done. Professional cards such as Inmos's original B004 cost about $8000, then c't brought an do it yourself project called Tek4/8 for $1500. Grab! > The Transputer is definitely not made up. Definitely not. The board is about 1.5m from where I am now. In my old hardware pile. > In essence a Transputer is a chip containing a small and relatively simple > CPU and some communications ports. The CPU is a bit like RISC except that > it has variable instruction lengths; Actually it is an 32bit stack based CPU (3 element hardware stack, plus 2 pointer registers) with single byte instructions. All instructions were 4 bit code and 4 bit operand. Larger operands and more instructions required prefix instructions (up to 7). And it had an 64bit FPU which ran asynchronous to the interger psrt (so it could calculate addresses while FP was working. It also had an 4k SRAM on chip (not an cache, as it was statically mapped, programs had to be explicitely compiled to use this SRAM). Oh, and no MMU. But the full DRAM multiplex and timing was on chip. > it also has a stripped-down priority > system in hardware (and maybe a few other exotic features which I forget). Round robin preemptive sheduled "low priority" threads and non deshedulable "high priority" threads. Any thread could wait on internal timers (2) or other threads (data exchange with them) or IO (data exchange with other transputers) or the external "interrupt" pin (called "event"). > The communications ports allow the chip to be connected to other Transputer > chips via short wires. 4 bidirectional 10MBit serial links. DMA driven. Also with "boot from link instead of external ROM" pin to make ROM-less systems. > In reality a Transputer is a board containing one or more of those CPUs plus > memory and possibly some I/O hardware (e.g., disk controllers, graphics). > Transputer boards without the fancy I/O hardware rely on a host system such > as an IBM PC clone. The B004 and the Tek4/8 were of the later type, with 1 Processor and 2MByte RAM and an C004 chip that provided the PC with on 10Mbit link. > Inmos (the designer and manufacturer of the Transputer) also designed a > programming language called occam to take advantage of the parallel > processing. And an extremely efficient compiler for it. My Mandelbrot program (the calculating part that run on the Transputer) compiled to 727 bytes: http://neil.franklin.ch/Projects/Mandel/occam/mandel.occ The PC part is: http://neil.franklin.ch/Projects/Mandel/qc/mandel.c Transputer control is: http://neil.franklin.ch/Projects/Mandel/qc/trproc.h This combo did Mandelbrot at 640x480 with 32 depth in 26 seconds. -- Neil Franklin, Nerd, Geek, Unix Wizzard and Guru, Hacker, Mystic neil@franklin.ch.remove http://neil.franklin.ch/ Computer: toy that speeds work, so you have more time to play with it ###### From: tmg@BAIT.bc.sympatico.ca (Trent Gamblin) Newsgroups: alt.folklore.computers Subject: Re: 286-386 transition References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2c <7m48bv$ia6$1@nntp1.u.washington.edu> Reply-To: tmg@BAIT.bc.sympatico.ca Message-ID: User-Agent: slrn/0.9.5.6 (UNIX) Lines: 14 Date: Fri, 09 Jul 1999 20:23:17 GMT NNTP-Posting-Host: 209.52.208.72 X-Complaints-To: news@bctel.net X-Trace: news.bctel.net 931551797 209.52.208.72 (Fri, 09 Jul 1999 13:23:17 PDT) NNTP-Posting-Date: Fri, 09 Jul 1999 13:23:17 PDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!feed2.news.luth.se!luth.se!sunqbc.risq.qc.ca!cyclone.bc.net!newsfeed.bctel.net!news.bctel.net!tmg Derek Peschel wrote: > In essence a Transputer is a chip containing a small and relatively simple > CPU and some communications ports. The CPU is a bit like RISC except that > it has variable instruction lengths; it also has a stripped-down priority > system in hardware (and maybe a few other exotic features which I forget). > The communications ports allow the chip to be connected to other Transputer > chips via short wires. So how exactly do they interface with a regular PC? Do they take over all processing or do they (somehow) split between themselves and the regular CPU? Do they require special software? Are they still being produced, and most importantly, where do I get one? ###### Newsgroups: comp.arch,alt.folklore.computers From: ben@eecg.toronto.edu (Benjamin Gamsa) Subject: Re: 286-386 transition X-Nntp-Posting-Host: isis.eecg.toronto.edu Message-ID: <1999Jul9.172258.4744@jarvis.cs.toronto.edu> Organization: University of Toronto References: <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> Date: 9 Jul 99 21:22:59 GMT Lines: 13 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news1.sunrise.ch!news.imp.ch!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.cwix.com!130.185.14.36!torn!utnut!utcsri!eecg.toronto.edu!ben In article , Mark Booth wrote: > >What would be really nice, would be a 'New Transputer'. How about 0.18um >die, 300-600MHz 64bit superscalar core, 4MB SRAM/cache, 4x1Gbps links, >single cycle floating point operations (or SIMD), built in support for >SDram or multi-channel Rambus memory interface (no support chips). The >SHARC gets pretty close. > Sounds like the Alpha 21364. ###### From: Mark Booth Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: Fri, 9 Jul 1999 21:46:32 +0100 Organization: OxIM Ltd. Message-ID: References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> Reply-To: Mark Booth NNTP-Posting-Host: oxim.demon.co.uk X-NNTP-Posting-Host: oxim.demon.co.uk:158.152.26.40 X-Trace: news.demon.co.uk 931553345 nnrp-09:237 NO-IDENT oxim.demon.co.uk:158.152.26.40 X-Complaints-To: abuse@demon.net MIME-Version: 1.0 X-Newsreader: Turnpike Integrated Version 4.02 M Lines: 106 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!isdnet!newsfeed.icl.net!dispose.news.demon.net!demon!news.demon.co.uk!demon!oxim.demon.co.uk!markbFrom2 In article <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk>, Richard Tobin writes >In article , >Trent Gamblin wrote: >>What the heck is a T800-20 Transputer card? Sounds like something >>you'd make up, when arguing about who's computer is better :) To be honest, there was a time when the worlds fastest computers were massively parallel systems based on hundreds/thousands of T800 Transputers. At the time they were outstanding. They were a fully 32bit processor which came with built in communications links[1], built in memory[2], fast memory interface[3], simple interfacing[4], scheduler and communications primitives built into the microcode, and a unique stack based huffman encoded RISC instruction set[5] which led to extremely compact code. [1] Which allowed them to communicate with 4 neighbours or peripherals at 10 or 20Mbps each. [2] 4KB SRAM. [3] For static and dynamic rams. [4] 5v power & single 5MHz clock. [5] The most common instructions were 8bit, less commonly used instructions were 2 or more bytes. >The Transputer was a processor made by Inmos during the 80s (and >possibly still made), intended for use in parallel. The classic Transputers were discontinued finally in December of last year. They can still be found, and indeed if you have a set top box for digital TV then you probably have a great-great-grandson of a Transputer in it. They were originally intended to be used in embedded systems or as the processors in application specific peripherals like graphics boards. But when the supercomputer guys found out about them they just started bolting hundreds together into huge systems. It took Inmos quite buy surprise. >You could get >cards with a bunch of transputers to plug in to PCs and other hosts. The best thing is, you could string as many as you wanted/needed/could afford together. I have worked with systems with over one hundred T8's, and at the time they blew away a whole network of DEC Stations and Suns when doing distributed ray tracing. In 1989, when my poor 386 was ages hours to render a single screen of a mandelbrot set, I could connect to the MEiKO computing surface and have it calculate it faster than my RS232 connection could transport the data back. >Like many "novel architectures" from that period - promising efficient >parallelism through some paradigm promoted with ideological zeal (in >this case CSP, Communicating Sequential Processes) - it failed to live >up to its promise. It may have failed to live up to some peoples hype, but it certainly performed. What was disappointing was that the software techniques and tools were not available to fully exploit the power. People had difficulty relating to the idea of running more than one program at once, let alone multiple threads within a single program or distributing those threads of execution over multiple processors. Now days 2 to 8 way SMP systems are common place, and 2 way systems are almost household. As an individual processor the Transputer family was better bang for the buck than anything else at the time, a T400 20 MIPS processor for $20! and a T800-30 could pump out 4.3 MFLOPS! As a processor for massively parallel systems it performed far better on embarrassingly parallel problems (ray tracing, mandlebrot etc.) than many real world problems, but then so do all massively parallel machines. And as an embedded processor it was great too. You could boot it down one of its communications links, debug it that way, it had an exceptionally low latency context switch and interrupt latency, and all you needed for a usable processor was 5v power and 5MHz clock. In the end it is a shame that Inmos screwed up with the T9000, and ST screwed up rest. What would be really nice, would be a 'New Transputer'. How about 0.18um die, 300-600MHz 64bit superscalar core, 4MB SRAM/cache, 4x1Gbps links, single cycle floating point operations (or SIMD), built in support for SDram or multi-channel Rambus memory interface (no support chips). The SHARC gets pretty close. >I'm sure a web search would have turned up this information. Too true, but any excuse......... Take care, Mark......... -- Mark Booth - My ReplyTo: address is valid, other addresses are for tracking UCE ----------------------------- markbSig@oxim.co.uk ----------------------------- OxIM Ltd. Registered Office 12 Kings Meadow, Ferry Hinksey Road, Oxford OX2 0DP Voice +44(0)1865 204881, Fax +44(0)1865 204882 Registered in England 3619439 ###### From: Mark Booth Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: Fri, 9 Jul 1999 21:46:32 +0100 Organization: OxIM Ltd. Message-ID: References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> Reply-To: Mark Booth NNTP-Posting-Host: oxim.demon.co.uk X-NNTP-Posting-Host: oxim.demon.co.uk:158.152.26.40 X-Trace: news.demon.co.uk 931553345 nnrp-09:237 NO-IDENT oxim.demon.co.uk:158.152.26.40 X-Complaints-To: abuse@demon.net MIME-Version: 1.0 X-Newsreader: Turnpike Integrated Version 4.02 M Lines: 106 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!isdnet!newsfeed.icl.net!dispose.news.demon.net!demon!news.demon.co.uk!demon!oxim.demon.co.uk!markbFrom2 In article <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk>, Richard Tobin writes >In article , >Trent Gamblin wrote: >>What the heck is a T800-20 Transputer card? Sounds like something >>you'd make up, when arguing about who's computer is better :) To be honest, there was a time when the worlds fastest computers were massively parallel systems based on hundreds/thousands of T800 Transputers. At the time they were outstanding. They were a fully 32bit processor which came with built in communications links[1], built in memory[2], fast memory interface[3], simple interfacing[4], scheduler and communications primitives built into the microcode, and a unique stack based huffman encoded RISC instruction set[5] which led to extremely compact code. [1] Which allowed them to communicate with 4 neighbours or peripherals at 10 or 20Mbps each. [2] 4KB SRAM. [3] For static and dynamic rams. [4] 5v power & single 5MHz clock. [5] The most common instructions were 8bit, less commonly used instructions were 2 or more bytes. >The Transputer was a processor made by Inmos during the 80s (and >possibly still made), intended for use in parallel. The classic Transputers were discontinued finally in December of last year. They can still be found, and indeed if you have a set top box for digital TV then you probably have a great-great-grandson of a Transputer in it. They were originally intended to be used in embedded systems or as the processors in application specific peripherals like graphics boards. But when the supercomputer guys found out about them they just started bolting hundreds together into huge systems. It took Inmos quite buy surprise. >You could get >cards with a bunch of transputers to plug in to PCs and other hosts. The best thing is, you could string as many as you wanted/needed/could afford together. I have worked with systems with over one hundred T8's, and at the time they blew away a whole network of DEC Stations and Suns when doing distributed ray tracing. In 1989, when my poor 386 was ages hours to render a single screen of a mandelbrot set, I could connect to the MEiKO computing surface and have it calculate it faster than my RS232 connection could transport the data back. >Like many "novel architectures" from that period - promising efficient >parallelism through some paradigm promoted with ideological zeal (in >this case CSP, Communicating Sequential Processes) - it failed to live >up to its promise. It may have failed to live up to some peoples hype, but it certainly performed. What was disappointing was that the software techniques and tools were not available to fully exploit the power. People had difficulty relating to the idea of running more than one program at once, let alone multiple threads within a single program or distributing those threads of execution over multiple processors. Now days 2 to 8 way SMP systems are common place, and 2 way systems are almost household. As an individual processor the Transputer family was better bang for the buck than anything else at the time, a T400 20 MIPS processor for $20! and a T800-30 could pump out 4.3 MFLOPS! As a processor for massively parallel systems it performed far better on embarrassingly parallel problems (ray tracing, mandlebrot etc.) than many real world problems, but then so do all massively parallel machines. And as an embedded processor it was great too. You could boot it down one of its communications links, debug it that way, it had an exceptionally low latency context switch and interrupt latency, and all you needed for a usable processor was 5v power and 5MHz clock. In the end it is a shame that Inmos screwed up with the T9000, and ST screwed up rest. What would be really nice, would be a 'New Transputer'. How about 0.18um die, 300-600MHz 64bit superscalar core, 4MB SRAM/cache, 4x1Gbps links, single cycle floating point operations (or SIMD), built in support for SDram or multi-channel Rambus memory interface (no support chips). The SHARC gets pretty close. >I'm sure a web search would have turned up this information. Too true, but any excuse......... Take care, Mark......... -- Mark Booth - My ReplyTo: address is valid, other addresses are for tracking UCE ----------------------------- markbSig@oxim.co.uk ----------------------------- OxIM Ltd. Registered Office 12 Kings Meadow, Ferry Hinksey Road, Oxford OX2 0DP Voice +44(0)1865 204881, Fax +44(0)1865 204882 Registered in England 3619439 ###### From: ard@p850ug1.demon.co.uk (Tony Duell) Subject: Re: 286-386 transition Newsgroups: alt.folklore.computers Date: 10 Jul 1999 22:16:43 +0100 Organization: P850 User Group Message-ID: <7m8d7r$ip@p850ug1.demon.co.uk> References: <7m48bv$ia6$1@nntp1.u.washington.edu> <7m6spg$7mi$1@nntp1.u.washington.edu> NNTP-Posting-Host: p850ug1.demon.co.uk X-NNTP-Posting-Host: p850ug1.demon.co.uk:158.152.97.199 X-Trace: news.demon.co.uk 931645579 nnrp-07:26917 NO-IDENT p850ug1.demon.co.uk:158.152.97.199 X-Complaints-To: abuse@demon.net X-Newsreader: TIN [version 1.2 PL2] Lines: 109 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!dispose.news.demon.net!demon!news.demon.co.uk!demon!p850ug1.demon.co.uk!not-for-mail Derek Peschel (dpeschel@u.washington.edu) wrote: : And Tony Duell, if you're out there, now would be a good time to jump in and : help this nice person out. :) Why me? Because I've got an ITEM (Inmos Transputer Evaluation Machine) sitting 2m from me? Or a pile of TRAMs (TRAnsputer Modules) on the shelf? Or because I've homebrewed transputer boards? OK, much of this has already been said, but I'll pull it all together... A transputer is closer to a microcontroller than a microprocessor in some ways. It's a single chips with : A CPU core (16 or 32 bit, maybe with floating point). Some SRAM (4K was typical on the chips I worked with) A boot ROM (see below). 4 (at least on the ones I worked with) high-speed (10mbs, 20mbps) serial links. A simple hardware scheduler. The idea was to make a system designed for parallel processing. If you have a number of parallel tasks, which communicated through 'channels of bytes', you could either put them on separate transputers communicating by the links, or on one transputer (letting the scheduler run the tasks as appropriate), or spilit them between transputers, with several tasks per transputer. The links are point-to-point. A link (2 connections, one for each direction + grounds) was connected between a serial 'port' (confusingly called a 'link' in the data sheet) on one device to a similar 'port' on another. You couldn't connect 3 devices using a single link, or anything like that. To link them to a host PC (or whatever) you use a chip called a C011 or a C012. This was a special UART-like thing which linked to a standard microprocessor bus on one side and had a transputer link on the other. The boot ROM on the transputer chip allowed you to boot a transputer over one of the links. Basically, after reseting the transputer, you could download a program over the link (and thus from a PC fitted with a C011 chip) into the transputer's memory. This transputer could similarly boot other transputers connected to its links, etc. The host software on the PC normally booted a suitable program on the transputers and then acted as a 'host server'. The transputer system could send message packets over the link to the PC to (for example) display a string on the screen, read/write a file on the PC's disk, etc. You didn't _have_ to do it that way. A transputer has conventional address and data buses. You can give it an external boot ROM. It can also have conventional peripheral chips linked up to it (from which you can boot using code in the external ROM). Or external RAM (there is an on-chip DRAM controller and refresh generator). To connect the links you either hardwired them (often with plug-in patchleads) if you knew what you wanted, or used a thing called a C004. This chip was essentially a switch for up to 32 links. You configured it to connect the links together as you wanted. You guessed it - the configuration info was passed over yet another link (!). There are things called 'TRAMs'. These are PCB modules with a transputer and some other chips on them. In general the only external connections are transputer links, the 'system services' (reset, error, analyse control lines), etc. The address and data buses from the transputer are not brought off the TRAM. TRAMs plug into a suitable motherboard (which may be a standard ISA card...) which often contains a C004, 'system services' support circuitry, host interface, etc. There are 'compute' TRAMs -- just a transputer and memory, which you use to add processing power. I/O TRAMs have peripheral functions, like GPIB ports, Ethernet, SCSI, etc. So you can use 'normal' peripherals on your transputer system. Then there are the strange ones, like a EEPROM TRAM (to boot a standalone transputer netowrk from code you stuck in the EEPROM), VecTRAMs (transputer + array processor chip ;-)) etc. You can also get PC ISA cards with a single transputer, RAM, and a host adapter on them. The other links are available on the back panel, so you can link an external network off them. One of the strangest transputer ISA cards I've seen is a B020 (commonly called a bozo ;-)). This is a trnasputer based graphics card. It drives a VGA monitor, and has a trnasputer + RAM + VRAM + video chips on it. Not a lot supports it, though. The software was either written in occam -- a language designed to be easy to compile for trnasputers -- or in something like parallel C or parallel fortran. You could try assembler, and the instruction set manual does exist, but it's a little odd. Occam is a strange language. Not too bad when you get used to it, but odd. The worst feature IMHO is that identation is critical. You signal the end of a logical block by not indenting as far. You spend most of your time counting spaces ;-) It wasn't common in the systems I was working with to try to run an OS on the transputer (I was using them for real-time control). But there is a thing called Helios, I think. I have never looked at it - one day I will. AFAIK, transputers are no longer being made :-(. There are still a fair few around (probably more in the UK than the States), and ISA transputer cards do appear at radio rallies (hamfests) from time to time over here. : -- Derek -tony ###### From: tmg@BAIT.bc.sympatico.ca (Trent Gamblin) Subject: Re: 286-386 transition Newsgroups: alt.folklore.computers Date: Sun, 11 Jul 1999 01:55:24 GMT Reply-To: tmg@BAIT.bc.sympatico.ca References: <7m48bv$ia6$1@nntp1.u.washington.edu> <7m6spg$7mi$1@nntp1.u.washington.edu> Message-ID: User-Agent: slrn/0.9.5.6 (UNIX) Lines: 9 NNTP-Posting-Host: 209.52.208.120 X-Complaints-To: news@bctel.net X-Trace: news.bctel.net 931658124 209.52.208.120 (Sat, 10 Jul 1999 18:55:24 PDT) NNTP-Posting-Date: Sat, 10 Jul 1999 18:55:24 PDT Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.direct.ca!newsfeed.bctel.net!news.bctel.net!tmg Derek Peschel wrote: > The regular PC is only used to handle the I/O. I think the connection [big honkin snip] Thanks for all the information. Sure wish I was around in the old days.. But I think it's time I hop over to comp.sys.transputer. Bye. ###### From: dpeschel@u.washington.edu (Derek Peschel) Subject: Re: 286-386 transition Newsgroups: alt.folklore.computers Date: 11 Jul 1999 03:53:27 GMT Organization: University of Washington, Seattle Lines: 16 Message-ID: <7m94fn$a32$1@nntp1.u.washington.edu> References: <7m6spg$7mi$1@nntp1.u.washington.edu> <7m8d7r$ip@p850ug1.demon.co.uk> NNTP-Posting-Host: saul6.u.washington.edu X-Trace: nntp1.u.washington.edu 931665207 10338 (None) 140.142.17.39 X-Complaints-To: help@cac.washington.edu NNTP-Posting-User: dpeschel Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.belnet.be!news-spur1.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.berkeley.edu!news.u.washington.edu!dpeschel In article <7m8d7r$ip@p850ug1.demon.co.uk>, Tony Duell wrote: >Derek Peschel (dpeschel@u.washington.edu) wrote: >: And Tony Duell, if you're out there, now would be a good time to jump in and >: help this nice person out. :) > >Why me? Because I've got an ITEM (Inmos Transputer Evaluation Machine) >sitting 2m from me? Or a pile of TRAMs (TRAnsputer Modules) on the shelf? >Or because I've homebrewed transputer boards? Those are good reasons. I was also hoping you'd have some URLs handy for the people who want them -- I thought I'd seen you mention them in the past. -- Derek ###### From: "Rob Nicholson" Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: Sat, 10 Jul 1999 19:54:31 +0100 Organization: Customer of Planet Online Lines: 11 Message-ID: <7md5nr$9e0$1@news4.svr.pol.co.uk> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> NNTP-Posting-Host: modem-112.glucophage.dialup.pol.co.uk X-Trace: news4.svr.pol.co.uk 931797563 9664 62.136.69.240 (12 Jul 1999 16:39:23 GMT) NNTP-Posting-Date: 12 Jul 1999 16:39:23 GMT X-Complaints-To: abuse@theplanet.net X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2014.211 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2014.211 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.mathworks.com!nntp-out.monmouth.com!newspeer.monmouth.com!diablo.theplanet.net!news.theplanet.net!newspost.theplanet.net!not-for-mail > Like many "novel architectures" from that period - promising efficient > parallelism through some paradigm promoted with ideological zeal (in > this case CSP, Communicating Sequential Processes) - it failed to live > up to its promise. Shame really as it was a very exciting technology. Didn't it have a special language for it - Occam? Rob. ###### From: "S. L. Wellborn" Newsgroups: alt.folklore.computers Subject: Re: 286-386 transition Date: Sun, 11 Jul 1999 21:56:21 -0700 Organization: Oregon Public Networking Lines: 117 Message-ID: <7mbsn8$blv$1@news.efn.org> References: <7m48bv$ia6$1@nntp1.u.washington.edu> <7m6spg$7mi$1@nntp1.u.washington.edu> <7m8d7r$ip@p850ug1.demon.co.uk> NNTP-Posting-Host: dynip24.oak.efn.org Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2014.211 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2014.211 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!kanja.arnes.si!news-hub.siol.net!news-out.cwix.com!newsfeed.cwix.com!128.223.220.30!logbridge.uoregon.edu!news.efn.org!not-for-mail Were these used in hypercube designs? I seem to remember some boards with Inmos chips near the carcass of a nCube machine that was laying around for a while in a surplus store in Portland,OR until someone snatched them up. Tony Duell wrote in message news:7m8d7r$ip@p850ug1.demon.co.uk... > Derek Peschel (dpeschel@u.washington.edu) wrote: > : And Tony Duell, if you're out there, now would be a good time to jump in and > : help this nice person out. :) > > Why me? Because I've got an ITEM (Inmos Transputer Evaluation Machine) > sitting 2m from me? Or a pile of TRAMs (TRAnsputer Modules) on the shelf? > Or because I've homebrewed transputer boards? > > OK, much of this has already been said, but I'll pull it all together... > > A transputer is closer to a microcontroller than a microprocessor in some > ways. It's a single chips with : > > A CPU core (16 or 32 bit, maybe with floating point). > Some SRAM (4K was typical on the chips I worked with) > A boot ROM (see below). > 4 (at least on the ones I worked with) high-speed (10mbs, 20mbps) serial > links. > A simple hardware scheduler. > > The idea was to make a system designed for parallel processing. If you > have a number of parallel tasks, which communicated through 'channels of > bytes', you could either put them on separate transputers communicating > by the links, or on one transputer (letting the scheduler run the tasks > as appropriate), or spilit them between transputers, with several tasks > per transputer. > > The links are point-to-point. A link (2 connections, one for each > direction + grounds) was connected between a serial 'port' (confusingly > called a 'link' in the data sheet) on one device to a similar 'port' on > another. You couldn't connect 3 devices using a single link, or anything > like that. > > To link them to a host PC (or whatever) you use a chip called a C011 or a > C012. This was a special UART-like thing which linked to a standard > microprocessor bus on one side and had a transputer link on the other. > > The boot ROM on the transputer chip allowed you to boot a transputer over > one of the links. Basically, after reseting the transputer, you could > download a program over the link (and thus from a PC fitted with a C011 > chip) into the transputer's memory. This transputer could similarly boot > other transputers connected to its links, etc. > > The host software on the PC normally booted a suitable program on the > transputers and then acted as a 'host server'. The transputer system > could send message packets over the link to the PC to (for example) > display a string on the screen, read/write a file on the PC's disk, etc. > > You didn't _have_ to do it that way. A transputer has conventional > address and data buses. You can give it an external boot ROM. It can also > have conventional peripheral chips linked up to it (from which you can > boot using code in the external ROM). Or external RAM (there is an > on-chip DRAM controller and refresh generator). > > To connect the links you either hardwired them (often with plug-in > patchleads) if you knew what you wanted, or used a thing called a C004. > This chip was essentially a switch for up to 32 links. You configured it > to connect the links together as you wanted. You guessed it - the > configuration info was passed over yet another link (!). > > There are things called 'TRAMs'. These are PCB modules with a transputer > and some other chips on them. In general the only external connections > are transputer links, the 'system services' (reset, error, analyse > control lines), etc. The address and data buses from the transputer are > not brought off the TRAM. TRAMs plug into a suitable motherboard (which > may be a standard ISA card...) which often contains a C004, 'system > services' support circuitry, host interface, etc. > > There are 'compute' TRAMs -- just a transputer and memory, which you use > to add processing power. > > I/O TRAMs have peripheral functions, like GPIB ports, Ethernet, SCSI, > etc. So you can use 'normal' peripherals on your transputer system. > > Then there are the strange ones, like a EEPROM TRAM (to boot a standalone > transputer netowrk from code you stuck in the EEPROM), VecTRAMs > (transputer + array processor chip ;-)) etc. > > You can also get PC ISA cards with a single transputer, RAM, and a host > adapter on them. The other links are available on the back panel, so you > can link an external network off them. > > One of the strangest transputer ISA cards I've seen is a B020 (commonly > called a bozo ;-)). This is a trnasputer based graphics card. It drives a > VGA monitor, and has a trnasputer + RAM + VRAM + video chips on it. Not a > lot supports it, though. > > The software was either written in occam -- a language designed to be > easy to compile for trnasputers -- or in something like parallel C or > parallel fortran. You could try assembler, and the instruction set manual > does exist, but it's a little odd. > > Occam is a strange language. Not too bad when you get used to it, but > odd. The worst feature IMHO is that identation is critical. You signal > the end of a logical block by not indenting as far. You spend most of > your time counting spaces ;-) > > It wasn't common in the systems I was working with to try to run an OS on > the transputer (I was using them for real-time control). But there is a > thing called Helios, I think. I have never looked at it - one day I will. > > AFAIK, transputers are no longer being made :-(. There are still a fair > few around (probably more in the UK than the States), and ISA transputer > cards do appear at radio rallies (hamfests) from time to time over here. > > : -- Derek > > -tony > ###### From: cscwyw@leonis.nus.edu.sg (Wong Yeow Whye) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Followup-To: comp.arch,alt.folklore.computers Date: 12 Jul 1999 05:16:39 GMT Organization: National University of Singapore Lines: 25 Message-ID: <7mbtnn$pjb$1@nuscc.nus.edu.sg> References: <7m1a36$d3f$1@rena.mat.uc.pt> NNTP-Posting-Host: leonis.nus.edu.sg X-Newsreader: TIN [version 1.2 PL2] Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.rhein-neckar.de!news-kar1.dfn.de!news-fra1.dfn.de!news0.de.colt.net!colt.net!news-lond.gip.net!news.gsl.net!gip.net!tank.news.pipex.net!pipex!do.de.uu.net!uunet!ams.uu.net!ffx.uu.net!newsfeed.attap.net!mango.singnet.com.sg!dahlia.singnet.com.sg!ocean.singnet.com.sg!id4.nus.edu.sg!nuscc.nus.edu.sg!cscwyw Mark Booth (markbFrom2@oxim.co.uk) wrote: : What would be really nice, would be a 'New Transputer'. How about 0.18um I believe it's called the 21364 :-) : die, 300-600MHz 64bit superscalar core, 4MB SRAM/cache, 4x1Gbps links, : single cycle floating point operations (or SIMD), built in support for : SDram or multi-channel Rambus memory interface (no support chips). The : SHARC gets pretty close. I have designed boards for both the transputer (late 80s to mid 90s) and the SHARC (mid 90s onwards). Compared with the transputer, the SHARC is a pain to work with. Among other things, the transputer's 5MHz clock was a breeze, unlike the SHARC's 40MHz clock and its (relatively) short clock transition times. I have not seen the latest SHARC tools (been out of touch for about 2 years) but I found INMOS's DOS based tools much easier to work with. (I have even wire-wrapped a few transputer boards and they worked well with absolutely no problems whatever - yes, no solder, no PCB traces, just them skinny wires enclosed in plastic.) Stephen Wong Department of Computational Science National University of Singapore ###### From: Mark Booth Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: Mon, 12 Jul 1999 22:16:21 +0100 Organization: OxIM Ltd. Message-ID: <1ar7xTAlsli30Ayo@oxim.demon.co.uk> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> <7md5nr$9e0$1@news4.svr.pol.co.uk> Reply-To: Mark Booth NNTP-Posting-Host: oxim.demon.co.uk X-NNTP-Posting-Host: oxim.demon.co.uk:158.152.26.40 X-Trace: news.demon.co.uk 931819081 nnrp-10:13499 NO-IDENT oxim.demon.co.uk:158.152.26.40 X-Complaints-To: abuse@demon.net MIME-Version: 1.0 X-Newsreader: Turnpike Integrated Version 4.02 M Lines: 51 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.mathworks.com!dispose.news.demon.net!demon!news.demon.co.uk!demon!oxim.demon.co.uk!markbFrom2 In article <7md5nr$9e0$1@news4.svr.pol.co.uk>, Rob Nicholson writes >Didn't it have a special language for it - Occam? Yup, but Occam never was just for Transputers. Indeed the first time I used Occam was an Occam 1 interpreter on a VAX. The move from interpreted Occam 1 on a 0.7 mips VAX being shared with 50 other people, to a real Transputer running compiled Occam 2 on a farm of multiple 20mips T800's was pretty eye opening. *8') SPOC and KROC are two more recent implementations of Occam. Certainly KROC is still under development, and can be downloaded in a binary distribution for Linux amongst others. Have a hunt around the Internet Parallel Computing Archive, if you are interested. Occam can represent parallel constructs and communication much more clearly than any other language I have seen. Many people hated it for niggling little things like indentation being syntactically significant, having upper case keywords or not having recursion. The only thing I really missed was the lack on C style unions. The difference in efficiency and clarity between the same algorithm implemented in C or Occam could be astounding. In fact, when we were being taught CSP, we were taught it using a modified Occam syntax rather than the tersely mathematical original CSP syntax. That made it somewhat easier to get to grips with. The amazing thing is that while Occam is a high level language, more often than not a line of Occam would be compiled down to a single Transputer instruction. Indeed Occam was often called the assembly language of Transputers. This might make it sound low level, but only if you don't know that the instruction set included instructions to create a process and initiate communication. Effectively there was an operating system micro-kernel built in microcode, and very efficient it was too. Most of the people who were into Occam, now seem to be getting into Java, where they are having to re-teach lessons learnt and forgotten over the last 20 years. Anyway, better get off this hobby horse, *8') Take care, Mark......... -- Mark Booth - My ReplyTo: address is valid, other addresses are for tracking UCE ----------------------------- markbSig@oxim.co.uk ----------------------------- OxIM Ltd. Registered Office 12 Kings Meadow, Ferry Hinksey Road, Oxford OX2 0DP Voice +44(0)1865 204881, Fax +44(0)1865 204882 Registered in England 3619439 ###### From: Jan Vorbrueggen Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: 13 Jul 1999 10:01:42 +0200 Organization: Institut fuer Neuroinformatik, Ruhr-Universitaet Bochum, Germany Lines: 13 Message-ID: References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> <7md5nr$9e0$1@news4.svr.pol.co.uk> <1ar7xTAlsli30Ayo@oxim.demon.co.uk> NNTP-Posting-Host: thalia.neuroinformatik.ruhr-uni-bochum.de X-Trace: sunu789.rz.ruhr-uni-bochum.de 931852745 13799 134.147.176.76 (13 Jul 1999 07:59:05 GMT) NNTP-Posting-Date: 13 Jul 1999 07:59:05 GMT X-Newsreader: Gnus v5.3/Emacs 19.33 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!rz.uni-karlsruhe.de!news.uni-stuttgart.de!news.ruhr-uni-bochum.de!not-for-mail Mark Booth writes: > Occam can represent parallel constructs and communication much more > clearly than any other language I have seen. Many people hated it for > niggling little things like indentation being syntactically significant, > having upper case keywords or not having recursion. The only thing I > really missed was the lack on C style unions. The difference in > efficiency and clarity between the same algorithm implemented in C or > Occam could be astounding. Completely agreed. Thanks for saving me the thinking and typing... Jan ###### From: prs@gol.com (Peter Stephenson) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Message-ID: <379335f6.7574118@nnrp.gol.com> References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> <7md5nr$9e0$1@news4.svr.pol.co.uk> <1ar7xTAlsli30Ayo@oxim.demon.co.uk> X-Newsreader: Forte Agent 1.5/32.451 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 14 Date: Tue, 13 Jul 1999 12:50:59 GMT NNTP-Posting-Host: 203.216.42.154 X-Complaints-To: abuse@gol.com X-Trace: nnrp.gol.com 931870259 203.216.42.154 (Tue, 13 Jul 1999 21:50:59 JST) NNTP-Posting-Date: Tue, 13 Jul 1999 21:50:59 JST Organization: Global Online Japan Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!newsfeed.nyu.edu!newsfeed.gol.com!203.216.70.8.MISMATCH!POSTED.nnrp.gol.com!not-for-mail On Mon, 12 Jul 1999 22:16:21 +0100, Mark Booth wrote: > >Occam can represent parallel constructs and communication much more >clearly than any other language I have seen. Many people hated it for >niggling little things like indentation being syntactically significant, >having upper case keywords or not having recursion. The only thing I >really missed was the lack on C style unions. The difference in >efficiency and clarity between the same algorithm implemented in C or >Occam could be astounding. As I recall, it was also relatively easy to mathematically prove that a certain piece of Occam did what it was supposed to do - and that quite a lot of Occam code had been proved in that way. Remember anything about that? ###### From: Dan.Pop@cern.ch (Dan Pop) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: 13 Jul 99 14:01:02 GMT Organization: CERN European Lab for Particle Physics Lines: 27 Message-ID: References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> <7md5nr$9e0$1@news4.svr.pol.co.uk> NNTP-Posting-Host: ues5.cern.ch X-Trace: sunnews.cern.ch 931876018 16112 (None) 137.138.32.79 X-Complaints-To: news@sunnews.cern.ch X-Newsreader: NN version 6.5.0 #18 (NOV) Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!EU.net!cern.ch!news.cern.ch!danpop In <7md5nr$9e0$1@news4.svr.pol.co.uk> "Rob Nicholson" writes: >> Like many "novel architectures" from that period - promising efficient >> parallelism through some paradigm promoted with ideological zeal (in >> this case CSP, Communicating Sequential Processes) - it failed to live >> up to its promise. > >Shame really as it was a very exciting technology. Didn't it have a special >language for it - Occam? This was the main problem with the system I had to work with: to take full advantage of the thing you had to use Occam (a language with a few adepts and ignored by the rest of the world), but I had to program in Transputer FORTRAN, which was a pain to use for parallel programming. If they provided decent extensions for parallel programming in the mainstream languages of the time, the transputer systems might have had a chance as computation accelerators for the PCs and workstations of the time. Or, who knows, even as the main processors of some workstation line. Dan -- Dan Pop CERN, IT Division Email: Dan.Pop@cern.ch Mail: CERN - IT, Bat. 31 1-014, CH-1211 Geneve 23, Switzerland ###### From: Jan Vorbrueggen Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: 13 Jul 1999 16:45:11 +0200 Organization: Institut fuer Neuroinformatik, Ruhr-Universitaet Bochum, Germany Lines: 11 Message-ID: References: <7m1a36$d3f$1@rena.mat.uc.pt> <6ulncr2cyq.fsf@chonsp.franklin.ch> <7m4mfo$29of$1@pc-news.cogsci.ed.ac.uk> <7md5nr$9e0$1@news4.svr.pol.co.uk> <1ar7xTAlsli30Ayo@oxim.demon.co.uk> <379335f6.7574118@nnrp.gol.com> NNTP-Posting-Host: thalia.neuroinformatik.ruhr-uni-bochum.de X-Trace: sunu789.rz.ruhr-uni-bochum.de 931876951 6522 134.147.176.76 (13 Jul 1999 14:42:31 GMT) NNTP-Posting-Date: 13 Jul 1999 14:42:31 GMT X-Newsreader: Gnus v5.3/Emacs 19.33 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!rz.uni-karlsruhe.de!news.uni-stuttgart.de!news.ruhr-uni-bochum.de!not-for-mail prs@gol.com (Peter Stephenson) writes: > As I recall, it was also relatively easy to mathematically prove that > a certain piece of Occam did what it was supposed to do - and that > quite a lot of Occam code had been proved in that way. Remember > anything about that? Occam programs (with some 8-| restrictions) also have a normal form, so that you can prove the equivalence of two pieces of code. Jan ###### From: Mike Tomlinson Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Fri, 16 Jul 1999 12:10:49 +0100 Organization: The Pub Lines: 12 Message-ID: References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> NNTP-Posting-Host: modem-59.benadryl.dialup.pol.co.uk Mime-Version: 1.0 X-Trace: news5.svr.pol.co.uk 932194927 5692 62.136.54.187 (17 Jul 1999 07:02:07 GMT) NNTP-Posting-Date: 17 Jul 1999 07:02:07 GMT X-Complaints-To: abuse@theplanet.net X-Newsreader: Turnpike (32) Version 4.01 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!news0.de.colt.net!colt.net!newspeer.clara.net!news.clara.net!diablo.theplanet.net!news.theplanet.net!newspost.theplanet.net!jasper.freeserve.co.uk!mike In article <7m1a36$d3f$1@rena.mat.uc.pt>, Rui Pedro Mendes Salgueiro writes >Yes, I think the 386 had a performance advantage since day one >(IIRC, there was also a 20 MHz and later a 25 Mhz). Intel made 386s up to 33MHz. AMD came out with a 40MHz variant which was roughly comparable in performance to a 486-25. -- Mike Tomlinson ###### From: "Jack Peacock" Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Sat, 17 Jul 1999 14:17:35 -0700 Organization: Simco Lines: 23 Message-ID: <932246155.349.86@news.remarQ.com> References: <3770CB69.18467660@strategypartners.com><7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net><377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner><7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com><7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com><7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de><377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> NNTP-Posting-Host: 207.168.124.145 NNTP-Posting-Date: Sat, 17 Jul 1999 21:15:55 GMT X-Trace: 932246155.349.86 S8PRG1HYA7C91CFA8C qube-02.us-ca.remarq.com X-Complaints-To: newsabuse@remarQ.com MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2314.1300 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2314.1300 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.berkeley.edu!remarQ73!supernews.com!remarQ.com!remarQ69!corp.remarQ.com!not-for-mail Mike Tomlinson wrote in message news:w75AkJA5Mxj3EwoQ@jasper.freeserve.co.uk... > In article <7m1a36$d3f$1@rena.mat.uc.pt>, Rui Pedro Mendes Salgueiro > writes > > >Yes, I think the 386 had a performance advantage since day one > >(IIRC, there was also a 20 MHz and later a 25 Mhz). > > Intel made 386s up to 33MHz. AMD came out with a 40MHz variant which > was roughly comparable in performance to a 486-25. > Cyrix made clock doubled 486DRx2 parts up to 66Mhz (33x2) which fit in the 386 socket. TI also made clock doubled 486DLC parts up to 50Mhz, based mostly on the Cyrix core. The Cyrix 66Mhz parts had real heat problems but the 50Mhz parts with 1K cache compared well with a low end 486 system. The problem was Cyrix charged dearly for the parts until TI came out with a copy and brought the price down. Surplus Cyrix parts made nice cheap 486 systems out of old 386 boards, in fact I still have an original 8Mhz IBM AT with a JET 386 adapter, containing a 16Mhz clock doubled Cyrix 486DRx2, pushing the outer limits of what you can get out of an original AT motherboard. Jack Peacock ###### From: Mike Tomlinson Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Sun, 18 Jul 1999 06:39:02 +0100 Organization: The Pub Lines: 19 Message-ID: References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <932246155.349.86@news.remarQ.com> NNTP-Posting-Host: modem-107.buproprione.dialup.pol.co.uk Mime-Version: 1.0 X-Trace: news6.svr.pol.co.uk 932278286 8304 62.136.55.235 (18 Jul 1999 06:11:26 GMT) NNTP-Posting-Date: 18 Jul 1999 06:11:26 GMT X-Complaints-To: abuse@theplanet.net X-Newsreader: Turnpike (32) Version 4.01 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news.ifi.unizh.ch!news.rhein-neckar.de!fu-berlin.de!newsfeed.nacamar.de!ayres.ftech.net!news.ftech.net!newspeer.clara.net!news.clara.net!diablo.theplanet.net!news.theplanet.net!newspost.theplanet.net!jasper.freeserve.co.uk!mike In article <932246155.349.86@news.remarQ.com>, Jack Peacock writes >Cyrix made clock doubled 486DRx2 parts up to 66Mhz (33x2) which fit in >the 386 socket. They also produced a 486SRx2 upgrade for the 386SX. This clipped over the original CPU, disabling it, and took over. It wouldn't work with some early 386SX-16s that had no disable pin. You had to run a small driver in config.sys to enable the cache. These upgrades were produced with a lifetime warranty; I had to smile when I returned a dead one to Cyrix last year and received a brand new replacement, in the shrinkwrapped box with all the accessories. I'd fully expected to be told replacements were no longer available. -- Mike Tomlinson ###### From: "Michael D." Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Sun, 18 Jul 1999 06:46:17 -0500 Organization: UUNET Lines: 33 Message-ID: <7mseog$3ap$1@ffx2nh3.news.uu.net> References: <3770CB69.18467660@strategypartners.com><7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net><377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner><7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com><7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com><7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de><377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <932246155.349.86@news.remarQ.com> NNTP-Posting-Host: channel86.1s.net X-Trace: ffx2nh3.news.uu.net 932298320 3417 208.244.232.140 (18 Jul 1999 11:45:20 GMT) X-Complaints-To: news@ffx2nh3.news.uu.net NNTP-Posting-Date: 18 Jul 1999 11:45:20 GMT X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 5.00.2314.1300 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2314.1300 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!blackbush.xlink.net!newsfeed.tli.de!newsfeed.amsterdam.nl.net!sun4nl!uunet!ams.uu.net!ffx.uu.net!ffx2nh3!not-for-mail Mike Tomlinson wrote in message news:rwz5oHA2hWk3EwDI@jasper.freeserve.co.uk... > In article <932246155.349.86@news.remarQ.com>, Jack Peacock > writes > > >Cyrix made clock doubled 486DRx2 parts up to 66Mhz (33x2) which fit in > >the 386 socket. > > They also produced a 486SRx2 upgrade for the 386SX. This clipped over > the original CPU, disabling it, and took over. It wouldn't work with > some early 386SX-16s that had no disable pin. > > You had to run a small driver in config.sys to enable the cache. These > upgrades were produced with a lifetime warranty; I had to smile when I > returned a dead one to Cyrix last year and received a brand new > replacement, in the shrinkwrapped box with all the accessories. I'd > fully expected to be told replacements were no longer available. > > -- > Mike Tomlinson Trust me, they have many thousands of obsolete parts in stock. With progress being what it is, they could keep the old machines going for many years to come. Please bear in mind, that we did the research, while they reaped the profits. Now that they think they don't need us anymore, we have been discarded, just like the old parts. What a surprise I hope they are in for! Sorry to have barged in. Michael D. kc5yfl@1s.net ###### Message-ID: <3791752D.579E@ozemail.com.au> From: Phil Herring Reply-To: revdoc@ozemail.com.au Organization: Church of the Many-tentacled Beast of Morth X-Mailer: Mozilla 3.02 (Win95; I) MIME-Version: 1.0 Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) References: <3770CB69.18467660@strategypartners.com> <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <932246155.349.86@news.remarQ.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 11 NNTP-Posting-Host: slsyd91p46.ozemail.com.au X-Trace: ozemail.com.au 932279553 203.108.214.110 (Sun, 18 Jul 1999 16:32:33 EST) NNTP-Posting-Date: Sun, 18 Jul 1999 16:32:33 EST Distribution: world Date: Sun, 18 Jul 1999 16:33:17 +1000 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!news.maxwell.syr.edu!news.mel.connect.com.au!news.mel.aone.net.au!newsfeed.aone.net.au!ozemail.com.au!not-for-mail Mike Tomlinson wrote: > > [...] These > upgrades were produced with a lifetime warranty; I had to smile when I > returned a dead one to Cyrix last year and received a brand new > replacement, in the shrinkwrapped box with all the accessories. I'd > fully expected to be told replacements were no longer available. Given the rate at which CPUs obsolesce, I'm sure that the vendors have warehouses full of old parts going nowhere, or perhaps being sold off, in odd lots at very low prices, to companies making embedded systems. ###### From: mj@isy.liu.se (Michael Josefsson) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition Date: Tue, 20 Jul 1999 10:03:37 CET Organization: LiTH / Dept of EE Lines: 81 Message-ID: References: <7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> <377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> <7ku740$971$2@vixen.cso.ui NNTP-Posting-Host: lagrange.isy.liu.se Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Newsreader: Trumpet for Windows [Version 1.0 Rev B] Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news1.sunrise.ch!news.imp.ch!sunqbc.risq.qc.ca!feed2.news.luth.se!luth.se!newnews.hk-r.se!news.ifm.liu.se!lagrange.isy.liu.se!mj In article <7m20u3$4kp$1@rena.mat.uc.pt> Rui Pedro Mendes Salgueiro writes: >From: Rui Pedro Mendes Salgueiro >Newsgroups: comp.arch,alt.folklore.computers >Subject: Re: 286-386 transition >Followup-To: alt.folklore.computers >Date: 8 Jul 1999 12:09:55 +0100 >Organization: Universidade de Coimbra >Lines: 78 >Message-ID: <7m20u3$4kp$1@rena.mat.uc.pt> >References: ><7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net> ><377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner> > <7ku740$971$2@vixen.cso.ui >uc.edu> <7kuk3j$m1q$1@nnrp1.deja.com> <7l0j1l$pgt$1@pyrite.mv.net> ><7l5ql9$sc4$1@nnrp1.deja.com> <7l8fhk$g28$1@pyrite.mv.net> ><377E8223.EF657BE@gmx.de> <377fbed6.81114078@news.ultranet.com> ><7m1a36$d3f$1@rena.mat.uc.pt> <37844CDF.88A9BA4D@hda.hydro.com> >NNTP-Posting-Host: localhost >X-Trace: rena.mat.uc.pt 931432199 4762 127.0.0.1 (8 Jul 1999 11:09:58 GMT) >X-Complaints-To: usenet@rena.mat.uc.pt >NNTP-Posting-Date: 8 Jul 1999 11:09:58 GMT >User-Agent: tin/pre-1.4-981114 ("The Watchman") (UNIX) (FreeBSD/2.2.7-RELEASE >(i386)) >Path: >news.ifm.liu.se!news.lth.se!feed2.news.luth.se!luth.se!news-ge.switch.ch!rccn.ne >t!rena.mat.uc.pt!not-for-mail >Xref: news.ifm.liu.se comp.arch:93837 alt.folklore.computers:201578 >In alt.folklore.computers Terje Mathisen wrote: >> Rui Pedro Mendes Salgueiro wrote: >>> * around 1988 or 1989 our departamental computer lab bought the >>> first PC for use by students (there were other PCs in use in >According to our records, it was in 1989. Just now, I took a look >at him (at its remains, more exactly. The power supply, the 20 MB >hard disk and its controller, the drives, were all removed for use >in other machines) and it is a 8088-2 ! The socket for the 8087 is >(as usual) empty. >>> other parts of the department, including a number of ATs (IBMs >>> with EGA ! A big thing at the time) >These were from 1986. >>> I don't remember for sure if the 386 was already available, but >> Compaq delivered the first 386 machines around 1986, I remember getting >> one of the first available here in Norway, running at a blazing 16 MHz. >Around November 1986, was when the Amstrad 1512 was arriving here >(I remember a lot of people being quite worried about the computers >they had in stock, having suddendly got very hard to sell, since >the Amstrad was half their price). The Amstrads (the 1512, first >and later the 1640) were the computers which made PCs common in >Portugal. Although at the same time, the IBM ATs were already >available (maybe even not uncommon), the Amstrads (and other clones) >were more significant in installed base. >> By 88-89 the 286 was pretty much gone. >Hmm, I think it was the summer of 1990 when I borrowed (from a shop >which belonged to some friends) a computer to do some schoolwork at >home. It was a 286 (brand new), which I think was still the common >PC at the time. Maybe the situation in richer countries was different. Not really. (if sweden was a richer country back then...). When I wrote my final exam paper in 1989 I used a brand new (and expensive as I was told) 286. But later that year I got my first own PC 24 MHz 386 1 Meg ram and 40 Meg HD, Hercules BW graphics. Still got it - running FreeBSD but with 8 Meg RAM, apart from its leaking battery works fine! /Micke >-- >http://www.mat.uc.pt/~rps/photos/ an ex-tifoso since 95/11/13 > Mark Sandman - Morphine, RIP (July 3th, 1999, Italy) >.pt is Portugal| `Whom the gods love die young'-Menander (342-292 BC) > Europe | Villeneuve 50-82, Toivonen 56-86, Senna 60-94 ###### From: nobody@nowhere.not (who?) Newsgroups: comp.arch,alt.folklore.computers Subject: Re: 286-386 transition (was: EV67 cruises, PA-8500 snoozes) Date: Mon, 19 Jul 1999 13:10:01 GMT Organization: Smartnet Internet Services [via news] Lines: 8 Message-ID: <37932325.233388585@news.smart.net> References: <3770CB69.18467660@strategypartners.com><7kqs0e$31k$1@server05.icaen.uiowa.edu> <37711045@News.Destek.net><377150CD.AA36ACAD@strategypartners.com> <1999Jun23.225731.1@eisner><7ku740$971$2@vixen.cso.uiuc.edu> <7kuk3j$m1q$1@nnrp1.deja.com><7l0j1l$pgt$1@pyrite.mv.net> <7l5ql9$sc4$1@nnrp1.deja.com><7l8fhk$g28$1@pyrite.mv.net> <377E8223.EF657BE@gmx.de><377fbed6.81114078@news.ultranet.com> <7m1a36$d3f$1@rena.mat.uc.pt> <932246155.349.86@news.remarQ.com> NNTP-Posting-Host: ketrontw.smart.net X-Newsreader: Forte Free Agent 1.11/32.235 Path: chonsp.franklin.ch!pfaff.ethz.ch!news-zh.switch.ch!news-ge.switch.ch!newsfeed.mathworks.com!news-out.cwix.com!newsfeed.cwix.com!207.176.80.103!news.smart.net!not-for-mail Mike Tomlinson wrote in message news:w75AkJA5Mxj3EwoQ@jasper.freeserve.co.uk... > Intel made 386s up to 33MHz. AMD came out with a 40MHz variant which > was roughly comparable in performance to a 486-25. My recollection is that the 386-33s were comparable to 486-25s, while the 386-40s were actually faster (for typical sorts of things done with PCs at that time).