brian@udcps1.UUCP (Brian R. Haug) (06/05/87)
----------------------------------------------------------------
32K NEWS April 1987
News about the NS32000 processor - Hardware and Software
Not affiliated with National Semiconductor Corporation
----------------------------------------------------------------
Greetings!
This is the first edition of what we hope will become a forum for
the exchange of NSC320xx news, hardware ideas and software.
We'll print code fragments in their entirety, duplicate
schematics, and distribute code on diskette. All we ask is that
no restrictions be placed on the use of information you share.
The, 32032 is the best 32-bit processor available. The 68000
series is its closest competition, but suffers from n
architecture that was not designed to support demand-paged,
virtual memory. Motorola has also been slow in comingnout with
the other members of its chip set. The result should have been a
smashing success for the NS32032. What happened? As usual,
marketing mistakes have snatched defeat from the jaws of victory.
(There have been rumors of a 32-bit processor from Intel, but
after carefully examining product announcements as well as real
products, it is apparent that their new processor is simply an 8-
bit processor with a 32-bit bus.)
We all have a stake in the future of the 320xx. Howh Because
we're the programmers that'll have to write code for inferior
processors if it fails. We're the engineer that'll have to
sweat over incomplete chip sets if it fails. We're the users who
will ultimately suffer from incomplete bug-ridden computers if
it fails. And we're also the people whose lives will be made so
much more unconstrained and productive WHEN it succeeds! So send
in your news, your schematics, your code. We'll share it all.
And we'll be sending dtems to other publications as we go. We'll
do it for free, for as long as we can afford to do so.
32K's In the News
-----------------
The Nov. 13, 1986 issue of EDN had a story about Owl Computers,
640 Crest Drive, Encinitas CA. Its president, Gene Bartsch,
produces VME- and Multibus-based 32032 computer boards.
BYTE, Jan. 1987, reports that several laser printer manufacturers
are designing in the 320xx instead of the 68000. They credit the
full 32-bit ALU and powerful bit operations, among other factors.
Hardware Items
--------------
Symmetric Computer Systems has a portable computer called the 375
which includes a Series 32000 processor with 2MB of RAM, floppy
and 50MB winchester, running 4.2BSD Unix, for $4,995. It comes
with C, Fortran, Pascal, Basic, APL, assembler, Lisp and Prolog,
plus Emacs, Ingres, Tex and Spice. A terminal is all you need to
add. Symmetrics Computer Systems, 1620 Oakland Road, Suite D200,
San Jose CA 95131.
Zaiaz Corporation, Huntsville AL, manufactures a PC coprocessor
board which can support up to 16MB of RAM. With 2MB and Unix,
the board costs $2,695. They also have a Clipper coprocessor for
a cool $6,000. Zaiaz Corp., 2225 Drake Avenue, Huntsville AL
358s5.
(What is a PC coprocessor? It is a board which you insert
into a standard PC, and uses the PC to perform its video and
disk I/O. Unencumbered from those duties, the CPU can run
at full speed, although the PC can be a limiting factor in
I/O-bound tasks.)
Definicon Systems has available a parts kit for the PD32 board
described in issue 32 of Micro Cornucopia. This kit includes a
board, all parts including 32016 and 1MB of RAM, and Unix on 30
floppies for $850. Assembled and tested, the system is
considerably higher. This is a PC coprocessor also. There is
some discussion in the Micro C. article about using a different
kind of computer for the host; however, PC-clones are about the
cheapest kind of computer available today.
There is a user's group for the PD32. For information, contact
Dan Efron, 8910 Westmoreland Lane, Minneapolis MN 55426.
Definicon Systems also has a 32032 coprocessor board with up to
8nB and Unix starting around $1,500. Definicon Systems Inc.,
31324 Via Colinas, Suite 108, Westlake Village CA 91362.
Compupro produced an S-100 board known as the CPU-32016, using a
32016 processor. It has been stated that these boards can go
10MHz.
Jameco advertises the National Semiconductor Designer's Kits,
$59.95 for 32016 set and $74.95 for a 32032 set. These sets
include the CPU, MMU, FPU, TCU and ICU, plus monitor ROMs and a
schematic for how to wire them up.2 The monitor (known as the
Tiny Development System) includes a debugger and simple
assembler. Jameco, 1355 Shoreway Road, Belmont CA 94002.
No printed circuit board is available for the Designer's Kits.
The 32032 requires a 68 pin LCC socket - anyone know where to get
one? (The same socket is used for an obscure Intel processor
known as the 80186.) Are wire-wrap models available?
To order back issues of Micro Cornucopia (#32 has a circuit
diagram for the P.32, especially recommended), write Micro
Cornucopia, P.O. Box 223, Bend OR 97709-0223. Include $3 U.S.
per back issue.
Software Items
--------------
Not a lot of immediately useful software for the 32000 seems to
be available.
The only operating system that everyone agrees to be available is
Unix, and not cheap, and no source or configurability. But it
comes with everything in the world.
Multi Solutions, Inc. claims that a version of their S1
operating system ( Unix is a dinosaur, CP/M and MS-DOS are toys")
will be available for the 32000, for $950 ("OEM Configurable"
plus "media charge"), plus $400 to $600 for each language
compiler. Despite the grandiosity of their ads, it appears that
they may actually have a real product out there. Is anyone using
it? Let us know. Multi Solutions, Inc., Suite 207, 123 Frankyin
Corner Road, Lawrenceville NJ 08648.
Library volumes available
-------------------------
The following volumes are available in 8" CP/M, 5-1/4" PC or 3-
1/2" Atari ST formats, for $8 apiece, or free if you submit a
disk.
Z32 - Cross assembler running on Z-80 CP/M, including Z-80
assembler, Lunar Lander-type game, 32K disassembler and sample
file , written by Neil Koozer.
A32000 - Cross assembler written in C, written by Richard Rodman,
described in Dec. 1986 Dr. Dob,'s Journal.
Please Write!
-------------
Is there something you need? Send out the call! People that
have written me have expressed an interest in the following:
- Schematic for 32032 Designer's Kit
- Printed circuit board for Designer's Kit
- 32032 sockets
- Single-user operating system
- Multi-user operating system like VMS
I personally don't know much about VMS. I have used Unix, and
there are some things it does quite well. At this stage of the
lame, we need to open up the doors that divide us into little
communities of specific products, because the 32000 users (and
would-be-users) are too few to be splintered. Send specs, send
source, send stuff we can all use and enjoy.
Let us all know about your successes - and failures (if any) - in
working with the Series 32000. Write to Richard Rodman, 1923
Anderson Road, Falls Church VA 22043. And enjoy computing in the
32-bit dimension!
LET'S WRITE AN OPERATING SYSTEM!
It seems to me that one of the principal lacks in the "hacker" or
computer hobbyist world today is a portable, simple, easily
configurable operating system. Naturally, the dream of every
programmer is to create his own operating system, but operating
systems don't sell well in the marketplace... so let's give it
away for free!
Commercial operating systems are too bloated with features. This
is actually true of all commercial software. Why? Because
features sell. ("What? Your word processor doesn't have automatic
hyphenation? Ours does - and it also has Cyro-Phoenician
punctuation support!" "Oh yeah? Well ours has automatic currency
justification AND in-line computation of hyoerbolic tangents!")
So I have been collecting ideas. Most of them center around the
scheduler and the user interface. (Obviously, any true operating
system is multitasking.) The virtual-memory feature of the 32000
should be a core mechanism of the kernel, rather than an oh-by-
the-way simplification to an overburdened swapping mechanism.
Simplicity in software means writing code to avoid writing code -
the generality and quality of an operating system should be
measured by all of the utilities it DOESN'T have - because it
doesn't need them.
The kernel should be simple enough that people using the 32000 in
firmware (ROM) systems, such as laser printer controllers or
telecommunications equipment, could make use of it. But its
simplicity should not impose limitations upon it.
Initial stages: Scheduler (multitasking kernel)
Memory manager
Debugger
Second stages: File manager
Command line interface
Compiler(s)
Text editor
Simple utilties
Third stages: Graphical user interface
Networking
Should the operating system should be written in a high-level
language? It could be hand-compiled, if necessary, into 32000
assembly language.
OK! So this is a project that might require man-years of work and
we might not finish everything! At least we'll all learn
something from it. Besides, since we'll be sharing all of the
source, everyone, even the people who already have Unix, should be
able to get something useful out of it!
Do you have any ideas? Let's work together on this.lm@cottage.UUCP (06/08/87)
In article <266@udcps1.UUCP> brian@udcps1.UUCP (Brian R. Haug) writes: >The, 32032 is the best 32-bit processor available. The 68000 Ahem. I know of few statements that could have caused me to question the content of this article more than one that contains "XXXXX is the best" . Sweeping generalizations never work. Larry McVoy lm@cottage.wisc.edu or uwvax!mcvoy
amos@instable.UUCP (06/08/87)
In article <266@udcps1.UUCP> brian@udcps1.UUCP (Brian R. Haug) writes: > LET'S WRITE AN OPERATING SYSTEM! Such an operating system already exists - it's called OS-9; it was written originally for 6800, and now is running on 680x0, but by the descriptions I have heard, it could only benefit from the added features of the 32k. I haven't seen one yet - is there anybody out there who knows more? -- Amos Shapir National Semiconductor (Israel) 6 Maskit st. P.O.B. 3007, Herzlia 46104, Israel Tel. (972)52-522261 amos%nsta@nsc.com @{hplabs,pyramid,sun,decwrl} 34 48 E / 32 10 N
rhealey@ub.D.UMN.EDU.UUCP (06/09/87)
Great blurb on the 32k, why'd you leave us big endder's out tho? 32032's power two different parallel processing powerhouses: Encore Mutimax series Sequent (I think) Wonder why the 32k was chosen to be the base unit in multi processor computer's? Probably because the other's couldn't handle the job! This message is comming to you from an Encore Multimax. 8 CPU's functioning as one fast one. 4.2 BSD, networking, and a whole lot more. Encore and Sequent show the awsome power that can be unleashed from just a 32032. The mind reels at the thought of 32532's in place of 32032's!! Sorry if this sounds like a plug for you know who but their machine is great in the bang/buck catagory although your average Joe on the street can't buy one. Don't rule out us bigger 32k systems, we can come up with alot of idea's too!!! Any of you Encore people care to comment on 32k flag waving (yeaaaa!!) -Rob Healey rhealey@ub.d.umn.edu
martin@iris.ucdavis.edu (Bruce K. Martin Jr.) (06/09/87)
In article <642@umnd-cs.D.UMN.EDU> rhealey@ub.UUCP (Rob Healey) writes: > > Great blurb on the 32k, why'd you leave us big endder's out tho? > 32032's power two different parallel processing powerhouses: > > Encore Mutimax series > Sequent (I think) > > Wonder why the 32k was chosen to be the base unit in multi processor > computer's? Probably because the other's couldn't handle the job! > According to a talk I attended last week by Dr. Chen of Encore (Director of Software R&D -- if memory serves), one of the main reasons that the NS32K series was choosen was the great instruction set (Test and Set mentioned explicitly). > > -Rob Healey > > rhealey@ub.d.umn.edu ...bruce ps. We have a great name for our Encore... Hydra :-) Bruce K. Martin martin@iris.ucdavis.edu
cyrus@hi.UUCP (Tait Cyrus) (06/09/87)
In article <266@udcps1.UUCP> brian@udcps1.UUCP (Brian R. Haug) writes: >---------------------------------------------------------------- >32K NEWS April 1987 > News about the NS32000 processor - Hardware and Software > Not affiliated with National Semiconductor Corporation >---------------------------------------------------------------- > >This is the first edition of what we hope will become a forum for >the exchange of NSC320xx news, hardware ideas and software. Well, I think that this is a good idea, so to get things rolling... Here at the University of New Mexico, we are building a 5th order hypercube (32 node) using the 32032 and 32016. Each node will contain the following: The items marked by a * are things that are different from anybody elses hypercube. 1) Compute engine a) 32032, (cpu - 32 bit) b) 32201, (tcu - timing/control unit) c) 32202, (icu - interrupt controller) d) 32082, (mmu - memory management unit) e) 32081, (fpu - floating point unit) * f) 8 Megabytes of DRAM g) 128 kilobytes of SRAM * h) running GENIX (Nationals version of 4.2 UNIX) 2) I/O engine a) 32016, (cpu - 16 bit) b) 32201, (tcu - timing/controll unit) c) 32202, (icu - interrupt controller) d) 32082, (mmu - memory management unit) e) 32081, (fpu - floating point unit) f) 128 kilobytes of SRAM * g) 2 RS232 ports * h) ethernet controller (AMD's lance) * i) SMD disk controller (8466) * k) 5 ~10 megabytes/sec (1 direction) links The 8Mbytes of DRAM is sudo-dual ported; that is both processors can access it. The 128 kilobytes of SRAM is accessable only by the local cpu and is used primarily as buffer space for both the ethernet and disk controllers on the I/O engine and stack space on the compute engine. Currently available hypercubes use serial transfers of 10 MegaBITS/sec whereas ours will be 16 bit parallel (at ~10 begaBYTES/sec). Current status: We have both of the above wirewrapped on a vme card and are about to start testing out our hardware design. After that we will start to port GENIX. We hope to have this done in 1-2 months. -- @__________@ W. Tait Cyrus (505) 277-0806 /| /| University of New Mexico / | / | Dept of EECE - Hypercube Project @__|_______@ | Albuquerque, New Mexico 87131 | | | | | | hc | | e-mail: | @.......|..@ cyrus@hc.dspo.gov or cyrus@hc.arpa or | / | / {gatech|ucbvax|convex}!unmvax!hi!cyrus @/_________@/
morrison@hc.DSPO.GOV (Bill Morrison) (06/09/87)
> Here at the University of New Mexico, we are building a 5th order > hypercube (32 node) using the 32032 and 32016. Each node will contain > the following: > Here at Los Alamos National Lab we are proud to be the parts supplier for above project, whenever they need anything they call us up and we take care of it for them (glad we are of use). Some famous quotes from UNM. "But we don't have any money." "Do you have any 26 gauge wire wrap wire." "My disk broke could you bring another down." "Did you order * yet?" How are you all at UNM anyway, I will be down later with the stuff you needed.
ram@nucsrl.UUCP (Renu Raman) (06/09/87)
>Great blurb on the 32k, why'd you leave us big endder's out tho? >32032's power two different parallel processing powerhouses: >Encore Mutimax series >Sequent (I think) >Wonder why the 32k was chosen to be the base unit in multi processor >computer's? Probably because the other's couldn't handle the job! True. Until recently only NS 32Ks were used in shared memory multis but times have changed. Sequent's recent Multiprocessor is based on 80386(yuk!) and claim a 81 MIPS (24 processor Balance 21000 is 21 Mips) rating (If at all you beleive in MIPS rating - I don't) They use 32 386s in a shared memory environment. Here comes my question. Like it or not shared memory multis are here to stay (uniprocessor power is saturating). Why don't the big three, like the transputer provide extra communication channels. Like-wise semaphore and instructions to support access of shared data will be an important feature if these processors are to be used in shared memort multis. >rhealey@ub.d.umn.edu >---------- Renu Raman UUCP:...ihnp4!nucsrl!ram 1410 Chicago Ave., #505 ARPA:ram@eecs.nwu.edu Evanston IL 60201 AT&T:(312)-869-4276
cyrus@hi.UUCP (Tait Cyrus) (06/09/87)
In article <4517@hc.DSPO.GOV> morrison@hc.DSPO.GOV (Bill Morrison) writes: >> Here at the University of New Mexico, we are building a 5th order >> hypercube (32 node) using the 32032 and 32016. Each node will contain >> the following: >> >Here at Los Alamos National Lab we are proud to be the parts supplier >for above project, whenever they need anything they call us up and we >take care of it for them (glad we are of use). Well, I have to apologize for not supplying the FULL facts. The hypercube that is being built is a joint project between Los Alamos National Laboratories and the University of New Mexico. Although a fair amount of the current work is being done at the U., a lot of the initial development was done at LANL. Currently a lot of help, in the way of parts and man power, has been supplied, and is still being supplied, by LANL. Again, I apologize for not giving 'everyone' credit for this very interesting and revolutionary project :^). Hopefully, full credit has been awarded to ALL involved. up Again, sorry for the slip -- @__________@ W. Tait Cyrus (505) 277-0806 /| /| University of New Mexico / | / | Dept of EECE - Hypercube Project @__|_______@ | Albuquerque, New Mexico 87131 | | | | | | hc | | e-mail: | @.......|..@ cyrus@hc.dspo.gov or cyrus@hc.arpa or | / | / {gatech|ucbvax|convex}!unmvax!hi!cyrus @/_________@Li
clif@intelca.UUCP (Clif Purkiser) (06/10/87)
> > Great blurb on the 32k, why'd you leave us big endder's out tho? > 32032's power two different parallel processing powerhouses: > > Encore Mutimax series > Sequent (I think) > > Wonder why the 32k was chosen to be the base unit in multi processor > computer's? Probably because the other's couldn't handle the job! It seems that Sequent has seen the light and switched processors for their next generation product. The reason they switched wasn't because the 386 good run MS-DOS it was because it was faster. -- Clif Purkiser, Intel, Santa Clara, Ca. {pur-ee,hplabs,amd,scgvaxd,dual,idi,omsvax}!intelca!clif These views are my own property. However anyone who wants them can have them for a nominal fee.
david@ms.uky.csnet (David Herron -- Resident E-mail Hack) (06/10/87)
In article <642@umnd-cs.D.UMN.EDU> rhealey@ub.UUCP (Rob Healey) writes: > Encore Mutimax series > Sequent (I think) Sort-of-true. It does power the Encore Multi-Max series. And it does power the old Sequent Balance. HOWEVER, Sequent has (for some reason) switched over to 80386's for their new machine(s). I do not know if they are totally abandoning the 32k or what (I'm not involved, except peripherally, with the purchase of this system, so don't know all the issues). > Wonder why the 32k was chosen to be the base unit in multi processor > computer's? Probably because the other's couldn't handle the job! I wonder why Sequent decided to switch? I have some vague memory about a PC co-processor mode in the Sequent, but that may have been a bad dream... :-) Doesn't seem like a very rational thing to do to me. > This message is comming to you from an Encore Multimax. This message is coming to you from a Unix PC being used as a terminal to a Sequent with an rlogin session over to a uVaxII. It's a long story and I'm not sure if anybody would believe me anyway.... > 8 CPU's > functioning as one fast one. 4.2 BSD, networking, and a whole lot > more. Encore and Sequent show the awsome power that can be unleashed > from just a 32032. The mind reels at the thought of 32532's in place > of 32032's!! Sorry if this sounds like a plug for you know who but > their machine is great in the bang/buck catagory although your > average Joe on the street can't buy one. If I remember correctly, the current MultiMax has 32332's in it. And according to the sales guy(s) who visited us a couple weeks ago, they had a large hand in the development of the 32332 and 32532. We are currently evaluating machines in order to buy a machine of this sort. We have a 23 processor 32032 Sequent in our machine room which Sequent has lent to us. It's a nice machine. In comparing it with the MultiMax (description only, we've never used one) it seems they're very very similar machines. Sequent has an advantage in that they started earlier and have more fully developed parallel support. Encore has a possible advantage in that they aren't in the process of switching processors. (I suppose that Sequent will be having some problems because of this great switch-over). We don't particularly care what processor is in the machine we finally get so long as it's very nicely parallel. -- ----- David Herron, cbosgd!ukma!david, david@UKMA.BITNET, david@ms.uky.csnet ----- (also "postmaster", "news", and the Usenet map maintainer for Kentucky.) ----- bsmtp-users@ms.uky.csnet for bsmtp discussion ----- bsmtp-users-request@ms.uky.csnet for administrivia
grenley@nsc.UUCP (06/11/87)
Oh boy, a debate about which CPU is better! I didn't realize any Intel guys read this group...envy, I suppose. In article <3479@intelca.UUCP> clif@intelca.UUCP (Clif Purkiser) writes: >> Wonder why the 32k was chosen to be the base unit in multi processor >> computer's? Probably because the other's couldn't handle the job! > > It seems that Sequent has seen the light and switched processors >for their next generation product. The reason they switched wasn't >because the 386 good run MS-DOS it was because it was faster. >Clif Purkiser, Intel, Santa Clara, Ca. >{pur-ee,hplabs,amd,scgvaxd,dual,idi,omsvax}!intelca!clif No Cliff, they switched because they fely their market DEMANDED ms-dos compatibility - the idea is that every sequent user can run IBM-PC software, I guess. Oh, boy. (I own a macintosh, so you can probably guess where my preferences are) The 386 is not as fast as the 32532, period. Furthermore, it is, I am told, even slower when running in 8086 compatibilty mode. If you believe otherwise, Cliff, then I'm sure you will be happy to participate in my proposed multi-vendor CPU shoot-out. See the comp.sys.arch for details, I haven't figured out how to cross-post yet. If you and the other Intel fans want to tell yourselves how pretty you are, do it in your own group. Regards, george grenley
ken@rochester.arpa (Ken Yap) (06/11/87)
|If you and the other Intel fans want to tell yourselves how pretty |you are, do it in your own group. I'm no fan of the Intel architecture but it seems to me that Cliff has just as much right as anybody else to (mis)inform in this newsgroup. If their mistakes get exposed, well so much the worse for them and so much the better for NSC. Ken
rhealey@ub.D.UMN.EDU (Rob Healey) (06/11/87)
In article <3479@intelca.UUCP> clif@intelca.UUCP (Clif Purkiser) writes: >> Wonder why the 32k was chosen to be the base unit in multi processor >> computer's? Probably because the other's couldn't handle the job! > > It seems that Sequent has seen the light and switched processors >for their next generation product. The reason they switched wasn't >because the 386 good run MS-DOS it was because it was faster. >-- >Clif Purkiser, Intel, Santa Clara, Ca. > At the risk of starting a war, are you comparing a 32032 to a 386 or a 32532 to a 386? 32032 came out in 83ish or so, and I don't remember the 386 being around in quantity at that time. -Rob rhealey@ub.d.umn.edu
rhealey@ub.D.UMN.EDU (Rob Healey) (06/11/87)
In article <6779@g.ms.uky.csnet> david@ms.uky.csnet (David Herron -- Resident E-mail Hack) writes: >In article <642@umnd-cs.D.UMN.EDU> rhealey@ub.UUCP (Rob Healey) writes: >> Encore Mutimax series >> Sequent (I think) >> This message is comming to you from an Encore Multimax. I put this in to show that the system isn't what is to be but an actual functioning system. >> 8 CPU's >> functioning as one fast one. 4.2 BSD, networking, and a whole lot >> more. Encore and Sequent show the awsome power that can be unleashed >> from just a 32032. The mind reels at the thought of 32532's in place >> of 32032's!! >If I remember correctly, the current MultiMax has 32332's in it. And >according to the sales guy(s) who visited us a couple weeks ago, they >had a large hand in the development of the 32332 and 32532. We have one of the first Max's, that's why we have 032's instead of 332's. The Max we have handles student programming and handles 64 users with very little slow down. The worst load average we've seen without TRYING to overload the machine is about 6 or so. When we ran parrellel make we tweeaked it up to 13.9x. The new CPU boards are supposed to offer a nocticable improvement. We've beta tested alot of their new kernels and Encore has ALWAYS responded quickly to problems we've found, usually by having our crash dump federal expressed to them overnight. For a commercial site UMAX 4.2 should be VERY stable, we mix equipment horribly here which has caused most of the problems. If you stick with Encore equipment ALL the way you won't have very many problems. > >We are currently evaluating machines in order to buy a machine of this >sort. We have a 23 processor 32032 Sequent in our machine room which >Sequent has lent to us. It's a nice machine. In comparing it with the >MultiMax (description only, we've never used one) it seems they're very >very similar machines. >----- David Herron, cbosgd!ukma!david, david@UKMA.BITNET, david@ms.uky.csnet We've only dabbled in actual parallel code but as far as the CPU's co-operating with each other and staying sain the Max does the job. Pascal, fortran and C all allow access to the parallel routines. If you have any specific pointed questions you'd like to ask let me know via email and I'll answer them as best I can. -Rob Healey rhealey@ub.d.umn.edu #include <disclaimers.h> Nobody ever listens to me anyways, why should they share my opinion? It's mine, ALL mine and you can't have it... Well, maybe if you ask REEEEEEAALL nice.
shiv@mas1.UUCP (Shiv Haris) (06/13/87)
I have done quite some assembly level programming on the 32000 and found it very simple. The specific project I worked on was porting VRTX kernel to the 32000. We adopted it from the 68000 version of VRTX. It was much easier to handle registers in a symetrical architecture. Thanks to the extra bit in the PSR for independence of user/supervisior mode and user/system stack we could do with just one kernel stack. Code compactness is also important when speed is important. And of course 32000 wins. Speed and compactness in such an envirionment is priority one. The only problem that National Semi has is not being lucky enough to get the blessings of the Big Brother (IBM). Other than that National came to the scene of 32 bit chips long ago. It had no hangups of upward compatibilty and designed a decent chip. I hate to see perfection being lost to upward compatibilty. Whoever said MS-DOS was a good operating system. But gained popularity because of IBM. There were other OS's good for the 8086 but were not popular. - Shiv Haris (ex-National Semi engineer who ported a 68000 VRTX realtime system to 32000 and now working with the 80386 and yet to find a more decent microprocessor than 32000) Most of you deal with higher level languages and don't see the architecture at all. You have the right to your point of view. Perfection Timing and Popularity may not always be in Sync. All flames welR sdon
schoff@nic.nyser.net.UUCP (06/15/87)
Sequent has now punted the 32X32 line for good old 386's. Are there any serious EE masochists thinking about a daughter board upgrade to the current processor boards? Marty Schoffstall schoff@nic.nyser.net
gnu@hoptoad.UUCP (06/15/87)
In article <542@mas1.UUCP>, shiv@mas1.UUCP (Shiv Haris) writes: > The only problem that National Semi has is not being lucky enough to > get the blessings of the Big Brother (IBM). Other than that > National came to the scene of 32 bit chips long ago. It had > no hangups of upward compatibilty and designed a decent chip. My recollection is that the 32000s weren't passed by because IBM disapproved but because they could not get their act together to ship working chips. Companies and individuals who were used to getting functioning chips after 3 revs (less if they weren't alpha test sites) gave up after 15 revs without working chips. My personal experience at Sun was that the first chips Motorola let out to us always had bugs, but were good enough for hardware testing, and that the first chips, or the next rev, were good enough to run Unix under, with compilers that used the whole instruction set. There were nits that might persist for 3 or 4 mask revisions, possibly requiring kludges in the kernel, but nothing a compiler or application program had to worry about. I also recall that the "full 32 bit" 32000 chip was only 30% faster than the 16-bit version, while Motorola had a part in design that ended up 300% to 500% as fast. This tends to support my conclusion that it was lack of expertise at chip design and debugging that derailed the 32000. -- Copyright 1987 John Gilmore; you may redistribute only if your recipients may. (This is an effort to bend Stargate to work with Usenet, not against it.) {sun,ptsfa,lll-crg,ihnp4,ucbvax}!hoptoad!gnu gnu@ingres.berkeley.edu
janm@runx.ips.oz (Jan Mikkelsen) (06/15/87)
irly sure that the entire kernel is pure assembly. Most of the
commands are 'C'. I don't know how different the Motorola series is
from National Semi's, but somehow I don't think recoding an operating
system would be a minor task. If I had the source, I'd give it a go though! :-)
Oh, a minor point :- OS9 was first written for the 6809, not the 6800.
Jan Mikkelsen.
ACSnet: janm@runx.ips.oz JANET: runx.ips.oz!janm@ukc
ARPA: janm%runx.ips.oz@seismo.css.gov CSNET: janm@runx.ips.oz
UUCP: {enea,hplabs,mcvax,prlb2,seismo,ubc-vision,ukc}!munnari!runx.ips.oz!janm
"He's dead, Jim."roger@nsc.nsc.com (Roger Thompson) (06/16/87)
In article <2295@hoptoad.uucp>, gnu@hoptoad.uucp (John Gilmore) writes: > > My recollection is that the 32000s weren't passed by because IBM > disapproved but because they could not get their act together to ship > working chips. Companies and individuals who were used to getting When IBM was out searching for a micro, our CPU was stable. What was the real decider, your guess is as good as mine. But it probably had more to do with application software being available without cumbersome AT&T licenses. > functioning chips after 3 revs (less if they weren't alpha test sites) > gave up after 15 revs without working chips. My personal experience at We had *NIX up and running in 1983. The first micro with demand paged virtual memory. We are working on our third generation MMU and I'm still waiting for your Mot to come up with one customers like SUN will use. > > I also recall that the "full 32 bit" 32000 chip was only 30% faster > than the 16-bit version, while Motorola had a part in design that ended > up 300% to 500% as fast. This tends to support my conclusion that > it was lack of expertise at chip design and debugging that derailed > the 32000. Yes our 32032 was only 30% faster than the 32016, but there is a good reason. The 32016 was a 32-bit micro to start with. Not Like the 68000. The internals of the 32016 and the 32032 are the same. Only the external data bus is different. The 32332 ( at the same frequencies) is any where from 50 to 80% faster than the 32032. All three have the same 32 bit internal register architecture and ALU and ALL are 100% software compatible. Can you say that for all the various Sun systems. Roger Thompson ---- Series 32000 group
roger@nsc.nsc.com (Roger Thompson) (06/16/87)
In article <175@nic.nyser.net>, schoff@nic.nyser.net (Martin Lee Schoffstall) writes: > Sequent has now punted the 32X32 line for good old 386's. > Are there any serious EE masochists thinking about a daughter > board upgrade to the current processor boards? I would like to say that Sequent has not punted the 32x32 line. They have chosen to offer different products, one line based on Nationals processors and one on Intels. In fact, if you read some of Sequents press articles, none even hint that they are ------ national. In fact, their president was very complimentary with respect to the recently announced 32532. Sequent has every intention of using it. As it relates to daughter board updates, I'm sorry but I can't help you there. Roger Thompson ------ Series 32000 Group
terryl@tekcrl.UUCP (06/16/87)
In article <4399@nsc.nsc.com> tekcrl!tektronix!cae780!amdcad!amd!intelca!oliveb!pyramid!nsc!roger roger@nsc.nsc.com (Roger Thompson) writes: +In article <2295@hoptoad.uucp>, gnu@hoptoad.uucp (John Gilmore) writes: +> +> My recollection is that the 32000s weren't passed by because IBM +> disapproved but because they could not get their act together to ship +> working chips. Companies and individuals who were used to getting + +When IBM was out searching for a micro, our CPU was stable. What was +the real decider, your guess is as good as mine. But it probably had +more to do with application software being available without +cumbersome AT&T licenses. Ho, boy, I can't really believe this one. Please, Mr Thompson, define "stable". I know for a fact that we (Tektronix) had to do quite a bit of software "workarounds" for bugs in the chip (most notably the memory-manage- ment/TLB hardware). When we discussed these problems with National, the answer was "Oh, of course we know about these problems; they're fixed in the next mask revision" (or words to that effect). If memory serves me correctly (about a 50-50 chance), we were using revision J chips. Naturally, National never told us about these bugs until we discussed it with them. Just my opinion, mind you, but I think Mr. Gilmore was more on the mark than you were. +> functioning chips after 3 revs (less if they weren't alpha test sites) +> gave up after 15 revs without working chips. My personal experience at + +We had *NIX up and running in 1983. The first micro with demand paged +virtual memory. We are working on our third generation MMU and I'm +still waiting for your Mot to come up with one customers like SUN +will use. Well, we had real 4.2 BSD Unix running on a 68010-based system running in late 1983-early 1984, so I'm not going to claim we were the first, but I'm wondering how companies like Sun were able to have one around the time frame you're mentioning (not to mention quite a few vendors who used the multiple 68000 trick to do demand paging; I'd have to say they were the REAL first micros with demand paging). Third generation MMU??? Sounds like you finally have all of the bugs worked out (I know, cheap shot, but as I said before, it was the MMU that was the cause of a lot of headaches for us). Does that also mean the MMU will finally work at the same speed as the processor?? (Another cheap shot, but we couldn't run systems at more than 8 Mhz because of the MMU chip). +> +> I also recall that the "full 32 bit" 32000 chip was only 30% faster +> than the 16-bit version, while Motorola had a part in design that ended +> up 300% to 500% as fast. This tends to support my conclusion that +> it was lack of expertise at chip design and debugging that derailed +> the 32000. + +Yes our 32032 was only 30% faster than the 32016, but there is a good +reason. The 32016 was a 32-bit micro to start with. Not Like the 68000. +The internals of the 32016 and the 32032 are the same. Only the external +data bus is different. The 32332 ( at the same frequencies) is any +where from 50 to 80% faster than the 32032. All three have the same +32 bit internal register architecture and ALU and ALL are 100% +software compatible. Can you say that for all the various Sun systems. Well, I have to concede this one to Mr. Thompson. No, I can't say that the 68000, 68010, and 68020 are 100% software compatible. What I can say is, that if code were to run on a 68000, it could run *unmodified* on a 68010 or a 68020 (I'm talking user code here, not system code). So the various flavors of 680X0 are software compatible UPWARDS (except for one instruction on the 68000). As for the internals of the chips, I'll concede this one to Mr. Thompson, also. Alas, the 68000 and 68010 are really only 16 bits internally, (though the register set is still the same). Terry Laskodi of Tektronix
mats@forbrk.UUCP (Mats Wichmann) (06/17/87)
(I am not really sure why Roger's reply bugs me so much, but it just sticks in my craw to where I have to stick my foot in where it will probably not be welcome....) In article <4399@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes: >When IBM was out searching for a micro, our CPU was stable. What was >the real decider, your guess is as good as mine. But it probably had >more to do with application software being available without >cumbersome AT&T licenses. > My experience was that individual CPU chips were all that was stable. It was quite possible to build *a* machine that worked very well. However, National sure raised doubts as to being able to ship a stable product in volume. At the company I was with, we had decided to go with a combination of a different CPU and a National MMU and FP chips. After working with this combination for months, we had to give up and throw the National chips out of the design. Why? We couldn't get precise specs out of National, and the chips did a number of unexplained things. Turns out there were two factors at work - the first one was that for whatever reason, National wasn't getting consistent yeilds, and they liked to give you three-chip sets (CPU, MMU, FP) that had been hand-picked to make sure they worked together. Any given CPU chip might not work with any given MMU chip, etc. This is enough of a deterrent to anybody who has to manufacture equipment, and was a killer for us, because we didn't want the National CPU chip at all. The second factor, the unavailability of precise timing specs, finally became clear at the end of the cycle - National finally admitted that they *couldn't* give us specs because they were still changing the CPU to make it work right, and were going to need to change the other two chips to match, so they wouldn't commit to any specs on the support chips. Motorola sure never pulled anything like that on us!!! >We had *NIX up and running in 1983. The first micro with demand paged >virtual memory. We are working on our third generation MMU and I'm >still waiting for your Mot to come up with one customers like SUN >will use. > First? That will get you lots of debate; I suggest you talk to people at Sun, old-timers at UniSoft, etc. Mats Wichmann
schoff@a.nyser.net.UUCP (06/17/87)
I realize that this is a question for my sequent salesman but reading between the lines you hint that I'll be able to buy an upgrade to my B21K that uses the 32532? Truly Astonishing. The only story that I have heard is that to upgrade my sequent machine I swap my 32032 based boards for 386 based boards. Marty Schoffstall schoff@nic.nyser.net
roger@nsc.UUCP (06/18/87)
In article <1751@tekcrl.TEK.COM>, terryl@tekcrl.TEK.COM writes: > > Ho, boy, I can't really believe this one. Please, Mr Thompson, define > "stable". I know for a fact that we (Tektronix) had to do quite a bit of > software "workarounds" for bugs in the chip (most notably the memory-manage- > ment/TLB hardware). When we discussed these problems with National, the answer > was "Oh, of course we know about these problems; they're fixed in the next > mask revision" (or words to that effect). If memory serves me correctly (about > a 50-50 chance), we were using revision J chips. Naturally, National never > told us about these bugs until we discussed it with them. Just my opinion, > mind you, but I think Mr. Gilmore was more on the mark than you were. > > The point I was making obviously hit a tender spot. As it related to IBM which was the original question, they were not looking for an MMU based system( I haven't seen one in the PC) and as you have said the CPUs worked real well. So our only area of contention in Teks case was in your mind the MMU and the incident of matched chip sets. My lab which is NOT in engineering has several running systems with the 1983/84 vintage parts. All run very well. In fact they are used for our interoffice mail. Uptime is constantly over 50 days. We in fact had for over a year one of Teks very own systems ---- stayed up real well. Yes we had delivery problems and yield problems and communication problems but as it related to the hardware aspects of the Tek system ----- better judgement tells me to leave well enough alone, but the blame does not always fall on the supplier. We have however as a result of the early days of Series 32000 been trained to believe that the customer is always right. We now screen parts over wider voltages and at HIGHER frequencies to meet special needs of various customers. One of which is Teks. The 32382 MMU which is used with the 32332 is fully production released and being shipped in our customers products ( OPUS, Encore and others) all on the 1 st mask rev. We have no changes in the pipe. Teks inputs have helped. > Well, we had real 4.2 BSD Unix running on a 68010-based system running > in late 1983-early 1984, so I'm not going to claim we were the first, but > I'm wondering how companies like Sun were able to have one around the time > frame you're mentioning (not to mention quite a few vendors who used the > multiple 68000 trick to do demand paging; I'd have to say they were the REAL > first micros with demand paging). > Yes ----- but was the MMU a discrete solution. Had to be. That was my point. > was the cause of a lot of headaches for us). Does that also mean the MMU > will finally work at the same speed as the processor?? (Another cheap shot, > but we couldn't run systems at more than 8 Mhz because of the MMU chip). > the older 082 is available in 6, 10 and 12.5 MHZ and the 32382 is at 15 MHZ. We aren't marketing lower frequencies of that one simply because all our yield runs at the 15 MHZ. We are in an inventory position. Roger
bzs@bu-cs.BU.EDU (Barry Shein) (06/20/87)
Posting-Front-End: GNU Emacs 18.41.4 of Mon Mar 23 1987 on bu-cs (berkeley-unix) > Any of you Encore people care to comment on 32k flag waving (yeaaaa!!) > > -Rob Healey > > rhealey@ub.d.umn.edu > One of Boston University's Encore Multimax's (we have four) was just upgraded to 6 NS32332 CPUs. As part of the upgrade someone here trashed our finger so I went and rebuilt it using their parallel make. It compiled and linked the 5000+ lines of code (46 modules) in about 42 seconds. Yes, there are advantages to such systems. -Barry Shein, Boston University
loverso@sunybcs.UUCP (John Robert LoVerso) (06/20/87)
In article <179@nic.nyser.net> schoff@a.nyser.net (Martin Lee Schoffstall) writes: > The only story that I have heard is that to upgrade my > sequent machine I swap my 32032 based boards for 386 > based boards. You also have to swap your memory boards (the current ones aren't fast enough). And then you have to recompile everything (actually, but up a new release of the O/S with 386 support - I understand it will compile to produce an 032-based system). But, don't forget to have all your user's recompile things in their ~/bin! John
elg@killer.UUCP (Eric Green) (06/22/87)
in article <334@forbrk.UUCP>, mats@forbrk.UUCP (Mats Wichmann) says: > In article <4399@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes: >>When IBM was out searching for a micro, our CPU was stable. What was >>the real decider, your guess is as good as mine. But it probably had >>more to do with application software being available without >>cumbersome AT&T licenses. Really? Did NSC go into a timewarp? If I recall right, the IBM PC was designed in 1980-81, and the 16032 was introduced in '81.... the PC was already designed! As for application software, I remember when people lambasted the IBM PC because it didn't HAVE any... Said one magazine writer at the time, "The only decent word processor available on the IBM PC is Wordstar, and it actually runs FASTER under CP/M!". >>We had *NIX up and running in 1983. The first micro with demand paged >>virtual memory. We are working on our third generation MMU and I'm >>still waiting for your Mot to come up with one customers like SUN >>will use. >> > First? That will get you lots of debate; I suggest you talk to people > at Sun, old-timers at UniSoft, etc. Just modify the statement to say "The first microprocessor with a single-chip MMU of decent design". I never did understand why 16032-based Unix machines never became popular... apparently Unix for a long time was too expensive for low-end machines, and the 16032 (cum 32016) didn't have the power to compete in the higher-end markets. Still, such a pretty architecture, perfectly suited for Unix..... Eric Green {ihnp4,cbosgd}!killer!elg elg@usl.CSNET -- * I don't care if Unix is trademark of GOD, I'm not going to fill up my letters with little asterisks whenever I write a bulletin, that's RIDICULOUS!
lm@cottage.WISC.EDU (Larry McVoy) (06/28/87)
In article <8877@bu-cs.BU.EDU> bzs@bu-cs.BU.EDU (Barry Shein) writes:
$ trashed our finger so I went and rebuilt it using their parallel make.
$ It compiled and linked the 5000+ lines of code (46 modules) in about
$ 42 seconds.
5000+ lines for finger?!?!?! Seriously? Jeez. :-)
Larry McVoy lm@cottage.wisc.edu or uwvax!mcvoysoper@encore.UUCP (Pete Soper) (06/29/87)
>In article <3761@spool.WISC.EDU> lm@cottage.WISC.EDU (Larry McVoy) writes: >>In article <8877@bu-cs.BU.EDU> bzs@bu-cs.BU.EDU (Barry Shein) writes: >>$ trashed our finger so I went and rebuilt it using their parallel make. >>$ It compiled and linked the 5000+ lines of code (46 modules) in about >>$ 42 seconds. >> >>5000+ lines for finger?!?!?! Seriously? Jeez. :-) Surely this was gnu-finger. Seriously, the third party Modula-2 compiler being prepared for Encore customers reproduces itself in 25 seconds on a Multimax 320, using 4 of a possible 20 cpus. On the slower model 120 it takes 45 seconds. If source files are arranged on a 320 just right to balance the work involved, the recompile takes 9 seconds with 6 processors. This involves 12000 lines of Modula-2 and about a dozen modules. Pete Soper - Encore Computer, Languages and Tools Group -- Pete Soper, Encore Computer Corp {talcott,ihnp4,decvax,allegra}!encore!soper
chongo@amdahl.amdahl.com (Landon Curt Noll) (07/15/87)
>In article <4399@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes: >When IBM was out searching for a micro, our CPU was stable. What was >the real decider, your guess is as good as mine. But it probably had >more to do with application software being available without ... Roger, I have confused by this. Perhaps you can explain a few things: I seem to recall a LONG LONG road from the Rev E 16032 (that could almost keep a Un*x kernel running) to a Rev R (that is almost bug free). Am I wrong or does this conflict with your statement of ``our CPU was stable''? It seems that both Mot and Intel have done very very well even with the MMU problem you talk about. The vast majority of Un*x boxes contain Mot or Intel CPUs. Maybe the market place doesn't see the lack of a complete chip set as a big problem, or maybe there is something about the NSC chip set that negates this advantage? chongo <> /\oo/\ -- [views above shouldn't be viewed as Amdahl views, or as views from Amdahl, or as Amdahl views views, or as views by Mr. Amdahl, or as views from his house]
roger@nsc.nsc.com (Roger Thompson) (07/16/87)
In article <10192@amdahl.amdahl.com>, chongo@amdahl.amdahl.com (Landon Curt Noll) writes: > Roger, I have confused by this. Perhaps you can explain a few things: > > I seem to recall a LONG LONG road from the Rev E 16032 (that could almost > keep a Un*x kernel running) to a Rev R (that is almost bug free). Am I > wrong or does this conflict with your statement of ``our CPU was stable''? > The point I was trying to make is that IBMs decission was software based not hardware. The MMU would/and does not make any difference to IBM. Oh yes, the 32532 is still on schedule. In fact it has been born and it does a whole lot more than wiggle. A whole lot more than towers of hanoi and we see no reason of not being able to sample on schedule. Roger
grenley@nsc.nsc.com (George Grenley) (07/16/87)
In <10192@amdahl.amdahl.com> chongo@amdahl.UUCP (Landon Curt Noll) writes: >>In article <4399@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes: >>When IBM was out searching for a micro, our CPU was stable. What was >> (paraphrase, to the effect that 16032 was availble way back when) > >Roger, I have (sic) confused by this. Perhaps you can explain a few things: > >I seem to recall a LONG LONG road from the Rev E 16032 (that could almost >keep a Un*x kernel running) to a Rev R (that is almost bug free). Am I >wrong or does this conflict with your statement of ``our CPU was stable''? > >It seems that both Mot and Intel have done very very well even with the MMU >problem you talk about. The vast majority of Un*x boxes contain Mot >or Intel CPUs. Maybe the market place doesn't see the lack of a complete >chip set as a big problem, or maybe there is something about the NSC chip set >that negates this advantage? Landon, Iguess we all know by now that you're not too fond of NSC. So be it. Nevertheless you should know that CPU architecture elegance is NOT the primary reason to pick a CPU. Look at the number of people who buy Amdahls - surely it doesn't represent the optimum 32 bit architecture.... But seriously, folks, CPU architecture isn't the bottom line, no matter how much us CPU types might wanna think so. Look at the number of design wins Intel got with the 8086 (hammered dog shit architecture) AFTER the 68000 was in volume production - I know, I was an FAE for Intel at the time. The 8086 family STILL sucks - but Intel's marketing whores don't - they're the best in the business. Still, with a good architecture AND good marketing, a chip like the '532 could surprise people. I know, I've seen it run (tee hee hee) Love, George
jans@tekchips.TEK.COM (Jan Steinman) (07/17/87)
grenley@nsc.nsc.com (George Grenley) writes, quotes: >In <10192@amdahl.amdahl.com> chongo@amdahl.UUCP (Landon Curt Noll) writes: >>>In article <4399@nsc.nsc.com> roger@nsc.nsc.com (Roger Thompson) writes: >>>When IBM was out searching for a micro, our CPU was stable... >>I seem to recall a LONG LONG road from the Rev E 16032 (that could almost >>keep a Un*x kernel running) to a Rev R (that is almost bug free)... >Landon, I guess we all know by now that you're not too fond of NSC... I think either Landon has too much time on his hands, or he's just a National-baiter. Why else would someone spend valuable working hours reading a newsgroup devoted to a processor they hate? It would be nice to see some serious discussion for a change.
chongo@amdahl.amdahl.com (Landon Curt Noll) (07/20/87)
In article <4497@nsc.nsc.com> grenley@nsc.UUCP (George Grenley) writes: >Landon, Iguess we all know by now that you're not too fond of NSC. So be it. >Nevertheless you should know that CPU architecture elegance is NOT the primary >reason to pick a CPU. Look at the number of people who buy Amdahls - surely >it doesn't represent the optimum 32 bit architecture.... I wish to add to your comment on a few points: * I AM FOND OF THE 32000. The machine in my home is a 32000 based Symmetric 375 which was bought after I left NSC. * I wouldn't simply dismiss my comments as a just flames from an x-NSC employee. I was involved in both the architecture and Genix (Unix on the NS32000) for a number of years. I still maintain contact with NSC people and the state of the chip set. My objectives are to bring items up for discussion, regardless if such items make marketing people happy. I don't simply fault NSC for ``chip-maker'' problems. I turned down job offers from Mot, Intel and TI because they were on the whole, worse than NSC in this regard. * Regarding Amdahls: (allow me to substitute mainframes to avoid discussions of ``my mainframe is better then your mainframe'') Mainframes often trail the high end state of the art by a number of years. Their target is NOT people who want ``Superconducting Nitrogen Cooled Optical connected thingy-ma-gigs''. Factors such as MTBF, Price/performance, Compat-ness with other equipment, environmental factors, etc. are important. Mainframes won't spout the state-of-the-art in hardware parts. Even so, mainframes do represent the state-of-the-art in performance, MTBF, and price/performance for a number of situations. chongo <my other CPU is an Amdahl 5890-300E> /\oo/\ -- [views above shouldn't be viewed as Amdahl views, or as views from Amdahl, or as Amdahl views views, or as views by Mr. Amdahl, or as views from his house]
grenley@nsc.nsc.com (George Grenley) (07/22/87)
In article <1467@tekchips.TEK.COM> jans@tekchips.TEK.COM (Jan Steinman) writes: >grenley@nsc.nsc.com (George Grenley) writes, quotes: >>Landon, I guess we all know by now that you're not too fond of NSC... >I think either Landon has too much time on his hands, or he's just a >National-baiter. Why else would someone spend valuable working hours reading a >newsgroup devoted to a processor they hate? >It would be nice to see some serious discussion for a change. Okay, Jan. Here's something: Our new '532 works. We announced it a few months ago, and, as usual w/ semi companies, we announced before we had silicon. Now, we have silicon. We're actually a bit ahead of schedule. For official info, contact your NSC sales person. Officially I don't know anything. There are those who maintain that that is true on an unofficial basis as well. 8-) But seriously, folks, it works pretty good. I'll post something when we're ready to demo it publicly. SO GO DESIGN IT IN! WE NEED THE BUCKS! BTW, Landon, if you're reading this, don't post replies. Don't mail 'em, either. Regards, George
aeusesef@csun.UUCP (Sean Eric Fagan) (07/25/87)
In article <10427@amdahl.amdahl.com> chongo@amdahl.UUCP (Landon Curt Noll) writes: >In article <4497@nsc.nsc.com> grenley@nsc.UUCP (George Grenley) writes: > >Landon, Iguess we all know by now that you're not too fond of NSC. So be it. > >Nevertheless you should know that CPU architecture elegance is NOT the primary > >reason to pick a CPU. Look at the number of people who buy Amdahls - surely > >it doesn't represent the optimum 32 bit architecture.... [Lot's and lot's of stuff deleted because of stupd requirements] > > * Regarding Amdahls: (allow me to substitute mainframes to avoid discussions > of ``my mainframe is better then your mainframe'') > > Mainframes often trail the high end state of the art by a number > of years. Their target is NOT people who want ``Superconducting > Nitrogen Cooled Optical connected thingy-ma-gigs''. Factors such > as MTBF, Price/performance, Compat-ness with other equipment, > environmental factors, etc. are important. Mainframes won't spout > the state-of-the-art in hardware parts. Even so, mainframes > do represent the state-of-the-art in performance, MTBF, and > price/performance for a number of situations. I've got to agree here. The 32k is a nice chip, allowing me (who likes to program in assembly language -- I'm seeking treatment though 8-)) to choose among a wide variety of instructions. Let me ammend that: a wide variety of *SLOW* instructions. At work, I work on Control Data Cybers, preferrably the 170/760, the fastenst machine (other than the Cray) I've ever worked on. (For those who don't know, the Cybers were designed by Seymore (sp?) Cray while he worked for CDC; they are similar to the Cray's except for lack of vectors.) Besides being a RISC machine, the 170/760 has wires on the inside. Lot's of them. Almost nothing else on the back, in fact, and there isn't a single silicon chip in the entire thing. (It's enough to give a repair technician nightmares.) But, even though using old technology, this thing will outperform 90% of the machines in existance today, and all of the machines when it was new. (But I hate RISC!) > >chongo <my other CPU is an Amdahl 5890-300E> /\oo/\ ----- Sean Eric Fagan Office of Computing/Communications Resources (213) 852 5742 Suite 2600 1GTLSEF@CALSTATE.BITNET 5670 Wilshire Boulevard Los Angeles, CA 90036 {litvax, rdlvax, psivax, hplabs, ihnp4}!csun!{aeusesef,titan!eectrsef} -------------------------------------------------------------------------------- My employers do not endorse my | "I may be slow, but I'm not stupid. opinions, and, at least in my | I can count up to five *real* good." preference of Unix, heartily | The Great Skeeve disagree. | (Robert Asprin)