cwebster@rodan.acs.syr.edu (Chris Webster) (11/08/89)
First of all, about the problem with 486 chips: I recall reading precisely what the problem was, but I can't remember where. Since it evidently wasn't on here, it must've been in a magazine. Since the only technical magazine (only because it's the only computing magazine that I read :) is BYTE, I assume that it was mentioned in the latest issue. The problem occurs when a certain instruction is executed after another certain instruction. If there isn't enough time or other instructions executed between them, it doesn't work right. (Sorry, but that's as technical as I can be.) In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of yesterday's Compaq press conference. (Today is Nov 7th, for those of you who don't get this on the same day.) They mentioned the two new Compaq i486 machines, but unfortunately in not great detail. Just hype; they even quoted someone on the floor as saying "poor ibm". Evidently the higher end model sits on the floor (sounded to me like a tower configuration AT) and can have 2 i486's or 2 386's, or combinations of both combos in it. For $15grand and up, it outperforms minicomputers. (Any one have reliable technical info?) Evidently, the problems with the chip have been worked out, since suppossively (according to the Net) Compaq found the problems. Disclaimer: I'm only an Undergrad, so my opinions do not reflect those of Syracuse University. They make too much sense for the administration to agree with. :-)
timsb@adspp.UUCP (Tim S. Boshart) (11/08/89)
In article <1263@rodan.acs.syr.edu> cwebster@rodan.acs.syr.edu (Chris Webster) writes: } } First of all, about the problem with 486 chips: I recall reading precisely } what the problem was, but I can't remember where. Since it evidently wasn't } on here, it must've been in a magazine. Since the only technical magazine } (only because it's the only computing magazine that I read :) is BYTE, I } assume that it was mentioned in the latest issue. The problem occurs } when a certain instruction is executed after another certain instruction. } If there isn't enough time or other instructions executed between them, it } doesn't work right. (Sorry, but that's as technical as I can be.) } } In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of yesterday's Compaq press conference. (Today } is Nov 7th, for those of you who don't get this on the same day.) They } mentioned the two new Compaq i486 machines, but unfortunately in not great } detail. Just hype; they even quoted someone on the floor as saying "poor } ibm". Evidently the higher end model sits on the floor (sounded to me } like a tower configuration AT) and can have 2 i486's or 2 386's, or } combinations of both combos in it. For $15grand and up, it outperforms } minicomputers. (Any one have reliable technical info?) Evidently, the } problems with the chip have been worked out, since suppossively (according } to the Net) Compaq found the problems. I was in Chicago yesterday to look at the new machines. Here is some more info: The high end model is the new Compaq SystemPro. It is currently available with a 33MHz 386 chip, and will be available with a 33MHz 486 chip when it is available. The system can be configured with either one or two processors. Here are some of the specifications from some of the literature I picked up: System Processor Flexible support for 33-MHz 386 and 486 processor Design: technology; support for two system processors. 386/33 System 32-bit 33-MHz 386 microprocessor; 33-MHz Processor on All 82385 Cashe Memory Controller with 64 Kbytes of 386-based high-speed static RAM; separate sockets for optional Models: 33-MHz Intel 387 and 33-MHz Weitek 3167 numeric coprocessors. Multiprocessor COMPAQ 386/33 System Processor Board; Options: compatible with Novell NetWare 386, SCO UNIX System V/386 and COMPAQ LAN MANAGER 386/486. System COMPAQ Flexible Advanced Systems Architecture Architecture: with Multiprocessing Support integrates multiple system processors, a separate 32-bit processor/memory bus and the 32-bit EISA I/O bus for highest possible system throughput. Expansion Slots: Total of 11 expansion slots; seven 8/16/32-bit EISA and four 32-bit processor/memory; eight slots available on all models. Memory: 4 megabytes of 32-bit memory standard, expandable to 256 megabytes. Shared system memory design with 80-ns enhanced page memory. Disk Storage Base unit (386-240) comes with a 240MB 2-Drive Devices: Array, Model 386-420 with a 420MB 2-drive array, Model 386-840 comes with 840MB 4 drive array. System can be expanded to 1.5GB internal storage, and a total of 4.28GB combined internal and external. -- Tim S. Boshart | uunet!adspp!timsb | "After a time, you may find that Abacus Data Systems | timsb@adspp.UUCP |'having' is not so pleasing a thing 2707 Middlebury St. | 219-295-4290 (voice)| as 'wanting.' It is not logical, Elkhart, IN 46516 | 219-522-2964 (fax) | but it is often true." -- Spock
jerry@starfish.Convergent.COM (Gerald Hawkins) (11/09/89)
cwebster@rodan.acs.syr.edu (Chris Webster) writes: > First of all, about the problem with 486 chips: I recall reading precisely > what the problem was, but I can't remember where. - Our newspaper here (San Jose Mercury) carried an article detailing a problem with multiplies and divides. Intel says the problem will occur when processing floating point multiplies and divides. The implication was that one could work around it by emulating. Obviously, that defeats half the purpose of using the '486. > In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of Compaqs 11/6 press conference. They > mentioned the two new Compaq i486 machines, but unfortunately in not great > detail. - Even having worked at a place where it was possible to release a computer with a major bug known, I still don't see how you can do it! Why would any customer buy a PC with a known bug? They'll have to release some dog version of BIOS which looks for the killer instructions, slowing down all FP operations. Comment? Anybody from Compaq out there? Anybody going to buy a new Compaq '486 machine even knowing the '486 has a floating point bug? " I dreamed I saw Joe Hill last night, alive as you and me. Said I, 'But Joe, you're ten years dead.' 'I never died' said he. " Jerry ( jerry@starfish.Convergent.COM ) -----
cs4g6ag@maccs.dcss.mcmaster.ca (Stephen M. Dunn) (11/09/89)
In article <1263@rodan.acs.syr.edu> cwebster@rodan.acs.syr.edu (Chris Webster) writes:
$ First of all, about the problem with 486 chips: I recall reading precisely
$ what the problem was, but I can't remember where. Since it evidently wasn't
There are actually two problems ... one has to do with executing the FSINCOS
and FTAN (I think those are the correct mnemonics) instructions when your
data meets certain conditions ... the other has to do with trapping errors
in virtual memory mode or something like that.
For more information, I suggest reading through comp.sys.intel or comp.arch.
There has been a fair bit of discussion of these bugs in those two groups.
$ minicomputers. (Any one have reliable technical info?) Evidently, the
$ problems with the chip have been worked out, since suppossively (according
$ to the Net) Compaq found the problems.
According to Intel, the problems have been found and corrected; all
chips shipped from now on are correct, and old chips can be returned for
replacement.
--
Stephen M. Dunn cs4g6ag@maccs.dcss.mcmaster.ca
<std_disclaimer.h> = "\nI'm only an undergraduate!!!\n";
****************************************************************************
They say the best in life is free // but if you don't pay then you don't eat
mcdonald@uxe.cso.uiuc.edu (11/09/89)
- >Even having worked at a place where it was possible to release a computer >with a major bug known, I still don't see how you can do it! Why would >any customer buy a PC with a known bug? They'll have to release some dog >version of BIOS which looks for the killer instructions, slowing down all >FP operations. Simple. Some buyers don't know about the bugs. Remember, IBM sells a lot to corporate accounts. The buyers of these don't know beans about computers. They just buy what IBM tells them to. I point out that I had no idea that my Model 80 was defective when I got it - I only found out when some programs bombed. I admit I hadn't checked carefully about it before getting it, as it was free. Look inside any IBM 386 machine - for well over a year the machines they sold were ALL defective - the 386 itself is bad. Look in your older 386 box at the 386. IF the chip doesn't say 80386DX, it is defective (not counting 386SX's, of course). Just TRY to get IBM to fix it! It CAN be done, but it isn't easy. It took us months of bitching. Just remember IBM's motto "We don't care!". Doug McDonald
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (11/10/89)
In article <1418@starfish.Convergent.COM>, jerry@starfish.Convergent.COM (Gerald Hawkins) writes: | Comment? Anybody from Compaq out there? Anybody going to buy a new | Compaq '486 machine even knowing the '486 has a floating point bug? Why not? The bug occurs infrequently and Intel has already said it will replace the chips on request. I'm waiting for the 33MHz part mayself, the problems which are too big for a 386 are going to need at least 2:1 improvement to be practical. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) "The world is filled with fools. They blindly follow their so-called 'reason' in the face of the church and common sense. Any fool can see that the world is flat!" - anon
slin@cory.Berkeley.EDU (Steven Philip Lin) (11/11/89)
In article <1418@starfish.Convergent.COM> jerry@starfish.Convergent.COM (Gerald Hawkins) writes: >Even having worked at a place where it was possible to release a computer >with a major bug known, I still don't see how you can do it! Why would >any customer buy a PC with a known bug? They'll have to release some dog >version of BIOS which looks for the killer instructions, slowing down all >FP operations. > >Comment? Anybody from Compaq out there? Anybody going to buy a new >Compaq '486 machine even knowing the '486 has a floating point bug? Well, I'm not from Compaq but I did work at Micronics a few summers ago when 386 computers were just coming out. When the 32-bit multiplication error was discovered in the 386 there was a big scare that everybody would stop buying 386's. That never happened; the purchase orders kept on coming in. This was before Intel had announced their exchange policy (i.e. if you have a defective chip you can trade it in when we get bug free chips), and Intel waited two or three months after discovery of the bug before announcing an exchange back policy. Apparently the fascination for wanting a new product overcame people's reservations for buying a defective product. The 386 was still fancy enough to warrant buying, bug or no bug. Plus there was virtually no software out there that used 32-bit multiplication so the bug wasn't really all too serious. From what I hear at Micronics, the same thing is happening with the 486. People still want to buy the new 486 boards, even with the buggy chip because they know they can just exchange the 486 for a new one when Intel ships the fixed chips.
poffen@chomolungma (Russ Poffenberger) (11/12/89)
In article <1263@rodan.acs.syr.edu> cwebster@rodan.acs.syr.edu (Chris Webster) writes: > > First of all, about the problem with 486 chips: I recall reading precisely > what the problem was, but I can't remember where. Since it evidently wasn't > on here, it must've been in a magazine. Since the only technical magazine > (only because it's the only computing magazine that I read :) is BYTE, I The Oct. 31 issue of EE times had a good description of the bug. Russ Poffenberger DOMAIN: poffen@sj.ate.slb.com Schlumberger Technologies UUCP: {uunet,decwrl,amdahl}!sjsca4!poffen 1601 Technology Drive CIS: 72401,276 San Jose, Ca. 95110 (408)437-5254 ------------------------- In a dictatorship, people suffer without complaining. In a democracy, people complain without suffering.