davidsen@steinmetz.steinmetz.UUCP (William E. Davidsen Jr) (07/15/87)
80386 user's mailing list Volume 1 #4 June 30, 1986 In this issue: 80386 test programs benchmarks the multiply bug 80386 smalltalk realtime BIOS request (with $$) ---------------------------------------------------------------- 80386 mailing list contributions should be made to 386users@udel.edu /or/ ...!uunet!seismo!udel.edu!386users Requests to join the list: 386users-request /or/ ...!uunet!seismo!udel.edu!386users-request ---------------------------------------------------------------- From: davidsen@crdos1.UUCP Subject: 80386 multiply test I should add to my last posting that, because the problem *may* be dependent on temperature, power supply voltage, etc, the test is probably not definitive. However, if your chip fails you should assume that it's bad. I have run it on several and none have failed yet. Is that a good sign? NB: the test runs many hours, if you haven't tried it yet! The prices on 80386 machines keep dropping. I see that AMT is now selling their machines for $2000 and $2400. The $2400 machine is the AMI motherboard, which is what I have in my GV386, and the AMI BIOS. It comes without hard disk or display. This *seems* to be the only machine with 64k cache. The $2000 machine is a Compaq clone, with column static memory on a 32 bit bus, a socket for an 80387, etc. This machine has a Pheonix BIOS. No memory on the mother board, but 1MB come with it, and sockets for another MB. A populated 4MB board is $1500. Bottom dollar clone: my wife, who runs a system house, got a dealer blurb on a Korean made 386 for $2995 with 30MB hard disk and monographics but no monitor. This has an 80287 socket. I did a limited check on the company and they seem for real (have been selling to C'land for 2+ years). I'll keep you posted, I think I'm going to get one and see if it's any good. Has anyone benchmarked the 80387 vs. the 80287. I have a 10MHz 80287 currently, and don't want to upgrade to a 16MHz 80387 if it's only 60% faster. If someone has some personal measurement they can share, particularly on multiply and divide, I would love to have it. The programs which take too long on what I have now need a lot more than 60% to make them practical. ---------------------------------------------------------------- From: hughM@sixhub.UUCP To: 386users@crdos1.UUCP Subject: more on 80386 tests **************** why couldn't they get it right? **************** >From infoworld: 80386 Test Software Inadequate, Intel Says By Tom Moran Intel begin advising its customers last week that software said to test 80386 chips for 32-bit multiply problems may not be able to do so accurately. "The multiply problem is not just a function of the data being multiplied," an Intel spokeswoman said about the problem that occurs in a small percentage of the chips. (See "Firms Downplay Impact of Intel 386 Chip Flaw," April 20.) "It's also subject to the frequency at which the part is being operated, the voltage being applied to the processor, and the ambient temperature." If a diagnostic program shows that a chip is bad, "that's fine and correct if the program is a good program. But if it tells you the chip is good, that's not necessarily true because of those environmental factors. It's a caveat for people who want to test their 386s," the spokeswoman said. Intel uses testers that cost around $3 million and are designed to duplicate a complete range of environmental conditions. "Taht's the only way to fully excercise the chip," the spokeswoman said. The company recommends that owners of untested 386s work with the manufacturer or dealer they bought their systems from to determine if they have a bad chip and if it should be replaced. In spite of shortages caused by the multiplying bug, Intel said last week that it is still on schedule to meet its goal of shipping 500,000 80386 chips this year. The revamped version of the 386, said to solve the multiply problem, is due to ship in July, Intel said. ---------------------------------------------------------------- To: 386users@UDEL.EDU Subject: questions about 386 "bug" in MUL instruction Date: Mon, 22 Jun 87 01:56:00 -0400 From: Steve Dyer <steinmetz!uunet!seismo!spdcc.com!dyer> Dave Farber recently submitted a C/MASM combination (386bug.c and 386bug.asm) to the INFO-IBMPC lending library at ISI. I snarfed them today and built a version to run under XENIX/286 on my Intel Inboard. I guess I'd like to understand better the failure mode the programs are supposed to test, and whether a result of receiving no errors for the duration of the test (it takes 7-8 hours to run) is a reliable indicator that I needn't worry about this in the future. Right now it's been running for several hours, and is at iteration 51700*65535 without finding any errors. Does this bug manifest itself as an error in 1/1000000 operations, or would a chip which exhibited the problem be expected to fail rather quickly? Running this (or any code which used 32-bit opcode prefixes) under XENIX 286 brought up an interesting issue; isn't it correct that register saving between tasks would save only the 286 register images (i.e., none of the extended registers)? Trying to run two copies of 386bug caused an error after a few iterations (but not immediately!--weird.) I haven't looked at this carefully enought to understand exactly what happened. --- Steve Dyer dyer@harvard.harvard.edu dyer@spdcc.COM aka {ihnp4,harvard,linus,ima,bbn,m2c}!spdcc!dyer ---------------------------------------------------------------- To: 386users@UDEL.EDU Subject: 386 bug--comments from Intel person Date: Mon, 22 Jun 87 20:15:51 -0400 From: Steve Dyer <steinmetz!uunet!seismo!spdcc.com!dyer> On Compuserve, I asked for opinions on how reliable the 386BUG test program was, and I received a comment from a person at Intel. He said that no test could predict that a part would or would not fail, that all chips being shipped for the past month have been screened (wonder what test they used?), and that a screened 386 has its package marked with two Greek sigmas. He then said that there was a phone number to call if your chip wasn't one of the screened ones-- 1-800-538-3373. I'll open mine up later this evening... ---------------------------------------------------------------- From: markb@mitisft.Convergent.COM (Mark Beyer) Subject: Re: IOCALL Benchmark Results Summary and Source Summary: Convergent Server/PC Date: 22 Jun 87 20:31:52 GMT Organization: Convergent Technologies, San Jose, CA In article <1594@ncr-sd.SanDiego.NCR.COM>, stubbs@ncr-sd.SanDiego.NCR.COM (Jan Stubbs) writes: > Send your results to me directly. Jan, My email couldn't get to you so I'm posting this instead. Here's the results for our new machine: Machine: Convergent Server/PC CPU: 20MHZ 80386 OS: Unix System V.3 The times are: real: 20.41 user: 1.21 sys: 19.0 Mark Beyer ---------------------------------------------------------------- From: dyer@spdcc.COM (Steve Dyer) Subject: 32-bit memory benchmark results for Inboard 386/AT card Summary: reposting & 32-bit memory results; Sun country... Date: 21 Jun 87 07:17:15 GMT Organization: S.P. Dyer Computer Consulting, Cambridge MA This is a reposting of benchmarks of the Intel Inboard 386/AT, this time with new results on the speed of the board fully loaded with 32-bit memory. Earlier results simply reported results using 16-bit memory on the AT bus. As you can see by the results below, the difference is quite startling! Earlier results indicated a ratio of 1.7/1 in average CPU speed -- with the 32-bit memory expansion, this has increased to better than 2.6/1 compared with a stock 8mhz AT. At least from the Dhrystone benchmarks reported below, we're beginning to edge into Sun 3 territory... --------------------------------------- All tests were run on the same hardware and software environment, with the exception of the replacement of the 286 for the 386 card, under SCO XENIX 286 OS v. 2.1.3 with Development System v. 2.1.4. Note that this is a test of the execution speed of 286 16-bit instructions. The Inboard used in this test has no memory of its own other than its cache. IBM PC/AT 8mhz IBM PC/AT with Intel Inboard 386/AT at 16mhz, cache enabled 16-bit mem 32-bit mem Drystone 1.0 no reg reg no reg reg no reg reg 1278 1292 2293 2304 3429 3405 Drystone 1.1 1084 1094 1957 1963 2906 2893 Buchholz (sum of user & sys times in sec) short cpu 0.3 0.2 0.1 medium cpu 3.3 1.9 1.2 long cpu *** *** (values out of range) short I/O 0.9 0.6 0.4 I/O bound 3.1 1.9 1.4 long mixed 56.9 33.8 21.8 -- Steve Dyer dyer@harvard.harvard.edu dyer@spdcc.COM aka {ihnp4,harvard,linus,ima,bbn,m2c}!spdcc!dyer ---------------------------------------------------------------- From: m5@bobkat.UUCP (Mike McNally ) Subject: Re: 32-bit memory benchmark results for Inboard 386/AT card Date: 24 Jun 87 14:05:42 GMT Organization: Digital Lynx, Inc; Dallas, TX In article <122@spdcc.COM> dyer@spdcc.COM (Steve Dyer) writes: > IBM PC/AT 8mhz IBM PC/AT with Intel Inboard 386/AT > at 16mhz, cache enabled > > 16-bit mem 32-bit mem > >Drystone 1.0 no reg reg no reg reg no reg reg > 1278 1292 2293 2304 3429 3405 >Drystone 1.1 1084 1094 1957 1963 2906 2893 Why is the with-register-variables time for the 32-bit version *slower* than the without-register-variables time? Isn't that sort-of strange? === Mike McNally, mercifully employed at Digital Lynx --- Where Plano Road the Mighty Flood of Forest Lane doth meet, And Garland fair, whose perfumed air flows soft about my feet... uucp: {texsun,killer,infotel}!pollux!bobkat!m5 (214) 238-7474 ---------------------------------------------------------------- From: dyer@athena.mit.edu (Steve Dyer) Subject: Re: 32-bit memory benchmark results for Inboard 386/AT card Date: 29 Jun 87 21:37:04 GMT Organization: Massachusetts Institute of Technology >Why is the with-register-variables time for the 32-bit version *slower* >than the without-register-variables time? Isn't that sort-of strange? Sure is, but that's what I got. I'm not sure why; perhaps it has to do with alignment of the variables (note that the results came from the SAME objects--the only difference was whether the 286 program was running in 16-bit memory or 32-bit memory). I'll be posting XENIX 386 results in a day or two; stay tuned... Steve Dyer dyer@harvard.HARVARD.EDU dyer@spdcc.COM aka {harvard,wanginst,ima,ihnp4,bbn,halleys}!spdcc!dyer ---------------------------------------------------------------- To: 386users@UDEL.EDU Subject: Desperately Seeking Swapping File-Id: Munck.Office-AT.386.1859 Date: Tue, 30 Jun 87 10:55:08 EDT From: Bob Munck <steinmetz!uunet!seismo!mitre-bedford.arpa!munck> I am in need of a 386 BIOS that supports demand paging. That is, support for the Directory and Page Tables, an interrupt handler for the page absent interrupt that reads in the needed page and restarts the task, and support for writing out the "least-recently used" pages and marking them as absent. At minimum, I'd be happy with a BIOS that allows my code to set up and activate the paging mechanism for user tasks. It would need to allow my code to handle the page absent interrupt and swapping out which would use it to read and write pages. I've written several paging systems for other machines, notably the 360/67 - 370, but I don't want to get bogged down writing disk I/O support. My project is funded by the DoD's Ada Joint Program Office. It is a prototype of the standard OS (MIL-STD-1838a) intended to support an APSE (Ada Programmer Support Environment). Because MITRE is a federal R&D facility, your BIOS can be kept under a non-disclosure or company proprietary agreement; because the project is a prototype, the code I write will be available with no strings to whoever helps me. In other words, you can help me at no risk of disclosure of your company secrets, and you'll get back a working demand paging system. It may even eventually pass the security certification process. Also, it's a neat OS, combining principles from Smalltalk, Ada, and relational data bases; it makes UNIX look very old and tired. -- Bob Munck, MITRE Corporation Munck@MITRE-Bedford.ARPA ...{ihnp4,utzoo,philabs,decvax}!linus!munck.UUCP "The MITRE Corporation, MS A-156, Bedford, MA 01730".USPS 617/271-3671.BELL ---------------------------------------------------------------- To: 386users@UDEL.EDU Subject: Submission for 386users Date: Thu, 25 Jun 87 08:58:17 -0400 Message-Id: <25790.551624297@udel.edu> From: BBoards Support (agent: James M Galvin) <steinmetz!uunet!seismo!UDEL.EDU!bboards> Bill, I have taken care of the addition. Here is a submission. Jim ------- Forwarded Message From: "David Reisner @favorite" <sdcsvax!telesoft!dar@ucbvax.berkeley.edu> To: ucbvax!udel.edu!386users-request@ucbvax.berkeley.edu Date: Wed, 24 Jun 87 18:29:26 PST Subject: Add to list + first submission Hi. Please add me to the list David Reisner sdcsvax!telesoft!dar Also, a reply to a previous submission (I've forgotten the appropriate address) - --------- Subject: Re: 386 Smalltalks The most recent (new) version of Softsmarts Smalltalk - a Xerox Version 2 image for AT compatibles under MSDOS - runs on the Compaq 386. I don't know of any implementation under {U,Xe}nix. Softsmarts (415)327-8100. - -David sdcsvax!telesoft!dar ------- End of Forwarded Message ---------------------------------------------------------------- End of 80386 users mailing list - volume 1 #4 -- bill davidsen (wedu@ge-crd.arpa) {chinet | philabs | sesimo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me