srm@nsc.UUCP (Richard Mateosian) (10/16/84)
* The response to my offer to recount the history of Series 32000 part numbering was large, so I'm posting the story. Right from the beginning (1978) the design objective for National's new microprocessor project was to produce a 32- bit vax-like "mainframe in silicon" with floating point sup- port and demand paged virtual memory. Since the Z8000 and 68000 were only rumors at that point, some imaginative per- son dubbed it the 8096 to one-up the 8086. It was soon realized that the plan was much too ambitious to be undertaken without intermediate steps, and the present division of the architecture into CPU, MMU, and FPU was con- ceived. Partly to simplify design and packaging of the 3-chip set, partly to avoid outdistancing the market (as happened with National's unsuccessful 16-bit IMP), and partly out of con- cern for the problems of users migrating from 8080 designs, three versions of the CPU were planned: The NS16008 would have a 16-bit version of the archi- tecture (16-bit registers and addresses) and a mode in which it could execute 8080 code directly. Its data bus would be 8 bits. The NS16016 would also be a 16-bit version of the architecture. It would be the same as the NS16008, but it would have a 16-bit data bus. I do not remember whether it had a provision for 24-bit addressing. The NS16032 would have a full 32-bit architecture and implementation but with a 16-bit data path to memory. You can see that the part numbering became illogical here. The NS16008 and NS16016 followed the pattern to which we have reverted today with the Series 32000: aaxdd means aa-bit architecture, dd-bit data bus. Under that scheme, the NS16032 would have been called the NS32016, and that is what it is called now. The original plan was a good one for the time it was con- ceived, but delays combined with rapid market changes to force rethinking, and in May 1982 when the "NS16000 Family" was introduced it consisted of the NS16032 CPU, the NS16081 FPU and the NS16082 MMU. Only the "misnumbered" NS16032 had survived, while the NS16008 and NS16016 had been placed "on hold". The idea of three different bus widths survived, but in a new form. This time all three CPUs would have the same 32-bit internal architecture and implementation, but with 8-bit, 16-bit and 32-bit data paths to memory. With the new CPU lineup, the numbering confusion became worse. It was decided that the new 8-bit bus version would be called by the old number, NS16008, even though the new part had a full 32-bit architecture and internal implementa- tion and no 8080-code execution capabilities. The 32-bit bus version was called the NS32032. This confusion was reduced when the NS16008 was renamed NS08032 in 1983. Now the numbering scheme was consis- tent, but it had flipped over, so that aaxdd had become ddxaa. Of course, now there was no connection at all to the name NS16000 Family, and in 1984 we bit the bullet and adopted the name Series 32000. At that time, we reverted to the aaxdd ordering, so that the three CPUs became the NS32032, NS32016, and NS32008. All of the other NS16000 Family chip numbers had been of the form NS16yyy, and these were all changed to NS32yyy, giving us the NS32081 FPU, the NS32082 MMU, the NS32201 TCU (clock chip), the NS32202 ICU, and the NS32203 DMA. With the change to Series 32000, most of our support equip- ment was also renamed/numbered, so we now have the Sys32 development system. For that system, different /-suffixes (e.g., Sys32/10) will denote different models. The current models all run Genix on the NS32016. Our emulators are designated ISE-dd, where ISE-dd emulates an NS320dd. Our evaluation boards (except for the no longer produced DB16) all have similarly descriptive numbers. In addition to this renumbering of current parts, we took the opportunity to renumber future products that we had been giving out limited information on. When the NS16000 Family was introduced in 1982, we referred to a high-performance CMOS implementation of the entire architecture on a single piece of silicon as the NS32132. It was to be the next CPU released. Our current plans call for a succession of CPUs, all designated NS32x32. What we used to call the NS32132 corresponds most closely in this new scheme to what we are now calling the NS32532. The new NS32132 is much closer to the NS32032. The NS32332, which will be introduced next year, is a major reimplementation of the architecture. We'll be releasing information on that part fairly soon. You may be wondering what happened to the designations NS32232 and NS32432. The NS32232 never existed, except in concept. All of its proposed new features have been incor- porated into the NS32132 and NS32332. As for the NS32432, we foresaw a naming problem there. People tend to refer to the NS32032 as "the '032", and they have started calling the NS32332 "the '332". We felt that there would be no end of confusion caused by introducing a part that could not avoid being referred to as "the '432", since as some of you no doubt remember, there was a microprocessor architecture with that designation released a few years ago by another semi- conductor company. In fact, there is a net.micro.432! Speaking of net names, you may ask why net.micro.16k is not renamed net.micro.32k or something of the sort. I've asked the same question, and I've been told it's too much trouble. So that's the whole story. Aren't you glad you asked? -- Richard Mateosian {amd,decwrl,fortune,hplabs,ihnp4}!nsc!srm nsc!srm@decwrl.ARPA
edler@cmcl2.UUCP (10/17/84)
Isn't there a clash of naming conventions between the NS32000 series and the WE 32000 chip? I am talking about the chip in the 3b2 (or is it in the 3b5? or both?). Jan Edler New York University edler@nyu-cmcl2.arpa cmcl2!edler
henry@utzoo.UUCP (Henry Spencer) (10/19/84)
> Isn't there a clash of naming conventions between the NS32000 series > and the WE 32000 chip? In a word, yes. But this is nothing new. "4024" is not an unambiguous part number either; Motorola had a series of old bipolar parts with 4000-series part numbers, and most of them clash with CMOS numbers now. It's surprising that this problem isn't more frequent. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
wapd@houxj.UUCP (Bill Dietrich) (10/23/84)
I don't know if Bell Labs changed the name of our chips from BELLMAC-32A and BELLMAC-32B to WE 32000 and WE 32100 before or after National changed from NS 16... to NS 32... . We did have somebody design our naming convention to be expandable and so on, and since BTL has lots of spare lawyers I assume several of them were consulted. (free advertisement time) The WE 32000 Microsystem is a 5-chip or 6-chip module (chips are mounted on a mini-PC card with 214 pins). There are two MBI (microbus interface) chips, two MBC (microbus controller) chips, a CPU chip, and an optional MMU chip. Without MMU, the whole thing is called the WE 32001 Processor Module. With MMU, it is the WE 32002 Processor Module with Memory Management Unit. The 3B-2/300 uses the WE 32002 (it uses our MMU chip). The 3B-5 uses the WE 32001 (it uses a discrete MMU). Our second generation VLSI consists of individual chips, not all mounted on a module. The WE 32100 CPU is the CPU (naturally), and the MMU is the WE 32101. Bill Dietrich Bell Labs, Holmdel, NJ houxj!wapd
wapd@houxj.UUCP (Bill Dietrich) (10/23/84)
I forgot to mention one thing about the WE 32000 series of chips by AT&T (someone's mail reminded me) : These chips are not available commercially yet (as individual chips). We've had working chips in systems for about 3 years now, and we're perfectly willing to sell systems, but so far the decision has been that the way to make money is to sell systems, not chips. This may change, because the debate re-surfaces every 6 months or so. In fact, at present the only way to get our instruction set manuals and so forth is to wave a lot of money at us and sign a non-disclosure agreement. The only good published paper on the CPU architecture is : Berenbaum, Condry and Lu, "The Operating System and Language Spport Features of the BELLMAC-32 Microprocessor", Symposium on Architectural Support for Programming Languages and Operating Systems, March 1982. The only good published paper on the MMU architecture is : Dietrich, Fuccio and Lu, "Architecture of a VLSI Map for BELLMAC-32 Microprocessor", Spring COMPCON 1983. Bill Dietrich ATT Bell Labs, Holmdel NJ houxj!wapd
jss@sjuvax.UUCP (Jonathan Shapiro) (10/24/84)
[Aren't you hungry...?] There shouldn't be a part numbering clash. Since when do microprocessor n8umbers conflict with electronic switching circuit numbers? -- disgusted with 3b2. Jon Shapiro
crandell@ut-sally.UUCP (Jim Crandell) (10/25/84)
> I forgot to mention one thing about the WE 32000 series > of chips by AT&T (someone's mail reminded me) : > > These chips are not available commercially > yet (as individual chips). We've had working chips > in systems for about 3 years now, and we're perfectly > willing to sell systems, but so far the decision > has been that the way to make money is to sell systems, > not chips.... Oh boy, another DEC. Just what this industry REALLY needs. -- Jim Crandell, C. S. Dept., The University of Texas at Austin {ihnp4,seismo,ctvax}!ut-sally!crandell
srm@nsc.UUCP (Richard Mateosian) (10/27/84)
* I believe that our legal department and theirs are in contact in an attempt to resolve any conflicts. -- Richard Mateosian {amd,decwrl,fortune,hplabs,ihnp4}!nsc!srm nsc!srm@decwrl.ARPA
steveg@hammer.UUCP (Steve Glaser) (10/29/84)
In article <cmcl2.24200001> Jan Edler <edler@nyu-cmcl2.arpa> writes: >Isn't there a clash of naming conventions between the NS32000 series >and the WE 32000 chip? I am talking about the chip in the 3b2 >(or is it in the 3b5? or both?). That's not the only one, there is also the NCR 32000 series processors. Steve Glaser tektronix!steveg or steveg.tektronix@csnet-relay
srm@nsc.UUCP (Richard Mateosian) (11/02/84)
> >Isn't there a clash of naming conventions between the NS32000 series > >and the WE 32000 chip? > > That's not the only one, there is also the NCR 32000 series processors. NCR part numbers have a discretely placed / (e.g. NCR 32/000). -- Richard Mateosian {amd,decwrl,fortune,hplabs,ihnp4}!nsc!srm nsc!srm@decwrl.ARPA