pcg@cs.aber.ac.uk (Piercarlo Grandi) (10/29/90)
[ ... Tek is out of 88K based systems -- poor 88k ... ]
Note that I am redirecting followups to comp.arch, because the alleged
demise of the 88k is not longer the sole subject of this thread.
On 23 Oct 90 02:22:26 GMT, mash@mips.COM (John Mashey) said:
mash> In article <2176@lupine.NCD.COM> rfg@NCD.COM (Ron Guilmette)
mash> writes:
rfg> I think this is pretty close to the mark. While I would not call
rfg> the 88k doomed, any more than I would call, say, the Intergraph
rfg> Clipper doomed, or the AMD29000, I think it has missed the chance
rfg> to gain any sizeable market penetration. I think you'll find a
rfg> trend in the industry toward clustering around 2 or 3 architectures
rfg> (as we have really always done: 8080/z80 and 6502, 80x86 and 680x0,
rfg> etc.). Unless something really strange happens, you'll see the
rfg> SPARC and MIPS chips float to the top of the heap. The rest of the
rfg> pack is left to niche markets or oblivion.
mash> I don't know what the true metric is: chip unit volume, system
mash> unit volume, or total value.
Well, if we are comparing chips it is probably total value of chips
shipped. Number of chips shipped is also another good metric, and
probably we want to see both numbers, because they say different things.
Given that CPU&support chips are a small fraction of system cost, it
seems silly to decide the popularity of a chip architecture on the unit
or sales volume of the systems it goes into.
rfg> The last time I heard, MIPS was still a little (basically one
rfg> product) company that was bleeding red ink. Also, the last thing I
rfg> heard about DEC's push in the (MIPS-based) RISC business was that
rfg> they were really not selling very many DECstations at all.
Well, MIPS has got the DEC account. That is by itself interesting;
actually *very* interesting; and DEC is selling fairly well, even if
other parts of the company are doing not so well. After all, as some
business weekly suggested some time ago, MIPS' business is basically
technology licensing, like Adobe, not products.
mash> Of course, some facts would help here: 1) MIPS sells machines
mash> ranging from $9K desktops thru $150K servers, and has cranked out
mash> quite a few system products in a few years.
Incidentally, why not make a MIPS PC/AT compatible? I mean, a machine
that has an R3000 chip set instead of a 386 chip set, and is otherwise
identical (can use the same peripherals, boards, cages, etc...). I think
that 386 compatibility would easily be done with a 386 plug in board
(instead of doing the opposite, like Everex and others, who put with
good success an 88k or 29k onto a plug in board) and VP/ix or DOSmerge.
It would be extraodinarily inexpesive -- I guess that an R3000 chip set
would be cheaper than a 386/486 chip set. I would believe that going for
the 8088/80286/80386 motherboard replacement market would be nice, and
could provide the needed volume for MIPS, or AMD or Motorola, or SPARC,
or the ARM, or anybody else.
If somebody says that the PC/AT motherboard technology is not well
suited to running high speed RISC chips, please tell me why the 486
seems competitive with such high speed RISC chips when running in PC/AT
type motherboards (ISA or EISA). Also, please tell me in which way it is
different from motherboard technology in the new Sun SPARC machines,
except that the letter do not have that many slots :-).
mash> We do have 700+ people, and have done >$100M so far in 1990.
mash> (This is NOT big, of course, but it's not a little 1-product
mash> company.)
This is *miniscule*. Many regional car dealerships, or Coca Cola
distributors, or McDonald's licensees, have substantially higher
turnovers and profits in the USA. Naturally it is interesting that MIPS
are doing 100M/700 == 150K dollars per employee on average. That they
are not posting huge profits is simply a miracle (I know better of
course).
Also consider Dell, or CompuAdd, or many of the players in the IBM PC
clone market. By comparison with them, MIPS is pretty small, and does
not have comparable growth rates or profits; it has substantially higher
turnover per employee though.
mash> 2) MIPS has designed both CMOS and ECL chipsets of various kinds,
mash> and has all kinds of technology-license products, as well as lots
mash> of software products, both of its own, and third-party.
Ok, ok, John Mashey, we know you are darn good. We know, OK? :-).
Not only that, you are the only ones, apart from SPARC, that have
licensed your thing to many second sources. You are also "open"!
mash> 3) We just announced results for last quarter, and I'd hardly call
mash> it bleeding red ink (slight profit), although life is certainly
mash> not easy out there right now for almost everybody in this
mash> business.
Including even the above mentioned clone makers.
mash> As is well-known, we are working hard with our colleagues at
mash> B.I.T. to improve yields on the ECL chips.
In the latest issue of Byte I read that BIT have *dumped* the R6000
development. Am I hallucinating? The justification given was that it is
expected that CMOS chips will soon reach the same speeds, and the window
of opportunity for an ECL based CPU has been reestimated from half a
dozen years to one or two, which is too little. Too bad, because I love
ECL.
mash> It doesn't take much arithmetic to see what happens when you have,
mash> for instance, 20 $150K computers you'd like to ship, and each is
mash> missing one chip..... $3M takes a big byte from a quarter, at our
mash> size.
That is why MIPS had better be a technology company. No matter how much
liquidity they have, big oligopolists can make their customers pay for
their suppliers' mistakes better.
mash> Well, do note that MIPS also has plenty of money in the bank,
mash> although hardly in this league. As I think another poster noted,
mash> the web of partnerships, relationships, investors, etc, around
mash> MIPS is much bigger than MIPS itself.
Both would melt in ten seconds if MIPS looked like being a loser.
Fortunately MIPS looks like being a winner.
--
Piercarlo "Peter" Grandi | ARPA: pcg%uk.ac.aber.cs@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.ukaxaris@acsu.buffalo.edu (vassilios e axaris) (10/29/90)
There is a company in Pennsylvania (Mars (?)) which manufactures an AT based machine which has a SPARC processor and is manufactured by Tatung. It has the AT bus, which accepts a 386 card for DOS, and a high speed bus for the sparc's memory and video. DOS output is redirected to the main video memory (using if I remember correctly, a shared memory) and runs SunOS (licsenced). It is fairly cheap (starting prices ~$4999). The info is from the Sun Systems journal (or something like that). Vassilios E. Axaris
peter@ficc.ferranti.com (Peter da Silva) (10/29/90)
In article <43029@eerie.acsu.Buffalo.EDU> axaris@acsu.buffalo.edu (vassilios e axaris) writes: > There is a company in Pennsylvania (Mars (?)) which manufactures an AT based > machine which has a SPARC processor and is manufactured by Tatung. [runs > sunOS] Sounds decent. > It is fairly cheap (starting prices ~$4999). Sorry. That's not "fairly cheap". That's not even "moderate". AT bus machines with 386es in them start under $2000. The main difference in manufacturing cost and components I can see is that you're running SunOS instead of DOS... is the SunOS license $3000? Is that with or without disk? -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
axaris@acsu.buffalo.edu (vassilios e axaris) (10/30/90)
I think the $4999 (why don't they make it an even $5000 ?) was without mass storage devices. I can check into it and post all details. Let me go now and salivate over the new Dec Station 5000 that we' ve got... :-) Vassilios E. Axaris
mash@mips.COM (John Mashey) (10/30/90)
In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: >Well, MIPS has got the DEC account. That is by itself interesting; >actually *very* interesting; and DEC is selling fairly well, even if >other parts of the company are doing not so well. After all, as some >business weekly suggested some time ago, MIPS' business is basically >technology licensing, like Adobe, not products. About 70% of our business comes from product, about 30% from technology. This is public knowledge, and I say it all the time in public talks. It would REALLY BE GOOD, if people who don't understand the business stopped making definitive claims about it.... > > >mash> We do have 700+ people, and have done >$100M so far in 1990. >mash> (This is NOT big, of course, but it's not a little 1-product >mash> company.) > >This is *miniscule*. Many regional car dealerships, or Coca Cola >distributors, or McDonald's licensees, have substantially higher >turnovers and profits in the USA. Naturally it is interesting that MIPS >are doing 100M/700 == 150K dollars per employee on average. That they >are not posting huge profits is simply a miracle (I know better of >course). Actually, the $100M/700 is NOT the way anyone computes $/employee. You have to compute $(year)/(average number of employees), which of course gave something like $200K or more per employee last year. >mash> As is well-known, we are working hard with our colleagues at >mash> B.I.T. to improve yields on the ECL chips. > >In the latest issue of Byte I read that BIT have *dumped* the R6000 >development. Am I hallucinating? ..... Yes, or Byte is. -- -john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc> UUCP: mash@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash DDD: 408-524-7015, 524-8253 or (main number) 408-720-1700 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086
crisp@mips.COM (Richard Crisp) (10/30/90)
In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: >In the latest issue of Byte I read that BIT have *dumped* the R6000 >development. Am I hallucinating? The justification given was that it is I also remember back in 1978 or 1979 that Byte stated that Motorola would offer a version of the 68000 that had the microcode in EPROM so that users could re-microprogram it for whatever instruction set they wanted! Caveat Emptor! -- Richard Crisp crisp@mips.com MIPS Computer Systems !decwrl!mips!crisp 928 Arques MS 2-02 (408) 524-8177 Sunnyvale, Ca 94086
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/30/90)
In article <P0R6UO2@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: | > It is fairly cheap (starting prices ~$4999). | | Sorry. That's not "fairly cheap". That's not even "moderate". AT bus machines | with 386es in them start under $2000. Not with 4MB, FPU, ethernet, and 1152x900 displays, they don't. These are aimed for a totally separate market. I haven't tested one yet, but on paper they will produce better cost/performance than the Sun SS+, at least at list price. 25MHz SPARC, FPU, MMU, 64 bit memory path, 64k cache. Color, harddisk, 386, and thinnet are options. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) VMS is a text-only adventure game. If you win you can use unix.
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/30/90)
In article <42488@mips.mips.COM> crisp@mips.COM (Richard Crisp) writes: | I also remember back in 1978 or 1979 that Byte stated that Motorola would | offer a version of the 68000 that had the microcode in EPROM so that | users could re-microprogram it for whatever instruction set they wanted! That actually may have been a plan at the time. I believe that IBM PC/370 system was based on a 68k with modified microcode, and I heard that the original idea was to allow anyone to modifiy the engine to custom applications. Not to defend BYTE, but it has been known that a company will plan a product and then not follow thru ;-) -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) VMS is a text-only adventure game. If you win you can use unix.
brett@cayman.amd.com (Brett Stewart) (10/31/90)
In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: > >Incidentally, why not make a MIPS PC/AT compatible? I mean, a machine >that has an R3000 chip set instead of a 386 chip set, and is otherwise >identical (can use the same peripherals, boards, cages, etc...). I think >that 386 compatibility would easily be done with a 386 plug in board >(instead of doing the opposite, like Everex and others, who put with >good success an 88k or 29k onto a plug in board) and VP/ix or DOSmerge. Deskstation Technology has done exactly this with the Am29000. I havent seen their actual product, but I have seen a photo. Their phone number is 913-599-1900. Best Regards; Brett Stewart Advanced Micro Devices, Inc. 1-512-462-5321 FAX 5900 E. Ben White Blvd MS561 1-512-462-4336 Telephone Austin, Texas 78741 USA brett@cayman.amd.com
brett@cayman.amd.com (Brett Stewart) (10/31/90)
In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: > >mash> In article <2176@lupine.NCD.COM> rfg@NCD.COM (Ron Guilmette) >mash> writes: > >rfg> I think this is pretty close to the mark. While I would not call >rfg> the 88k doomed, any more than I would call, say, the Intergraph >rfg> Clipper doomed, or the AMD29000, I think it has missed the chance >rfg> to gain any sizeable market penetration. I think you'll find a >rfg> trend in the industry toward clustering around 2 or 3 architectures >rfg> (as we have really always done: 8080/z80 and 6502, 80x86 and 680x0, >rfg> etc.). Unless something really strange happens, you'll see the >rfg> SPARC and MIPS chips float to the top of the heap. The rest of the >rfg> pack is left to niche markets or oblivion. > >mash> I don't know what the true metric is: chip unit volume, system >mash> unit volume, or total value. > >Well, if we are comparing chips it is probably total value of chips >shipped. Number of chips shipped is also another good metric, and >probably we want to see both numbers, because they say different things. >Given that CPU&support chips are a small fraction of system cost, it >seems silly to decide the popularity of a chip architecture on the unit >or sales volume of the systems it goes into. ^^^^^^^^^^^^^^^^^^^^^^^^^^^ Or even sillier to decide the popularity of a chip on ONLY ONE OR TWO systems that it goes into, rather than the collection of all the systems it goes into. It may not be well understood that the architecture contest is far from over for chip makers. I have sat in on discussions where we say things like 'only 50,000 29000's? It's not worth the trouble.' Economics for us, selling a sub $100 product with a few dollars of margin is very different from economics for other commentators in this forum. Our data says that most of the prospective consumers of 32-bit RISC's are as yet uncommitted, but will start to commit over the next, say, 18 months. Best Regards; Brett Stewart Advanced Micro Devices, Inc. 1-512-462-5321 FAX 5900 E. Ben White Blvd MS561 1-512-462-4336 Telephone Austin, Texas 78741 USA brett@cayman.amd.com
crisp@mips.COM (Richard Crisp) (10/31/90)
In article <2806@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes: >In article <42488@mips.mips.COM> crisp@mips.COM (Richard Crisp) writes: > >| I also remember back in 1978 or 1979 that Byte stated that Motorola would >| offer a version of the 68000 that had the microcode in EPROM so that >| users could re-microprogram it for whatever instruction set they wanted! > > That actually may have been a plan at the time. I believe that IBM >PC/370 system was based on a 68k with modified microcode, and I heard >that the original idea was to allow anyone to modifiy the engine to >custom applications. > The PC/370 did use 68K chips with modified microcode. They also had several modifications made to the "core" as well. When we (the 68K design team) read the Byte article I mentioned above, we all laughed. It was the most ludricrous thing we had heard to date regarding our chip. There never was a plan to offer user microprogrammable 68K's that any of us has heard of before or after that date. Byte simply stuck their neck out and printed some unfounded rumor as they have done recently involving BIT and the R6000 chipset. It happens all the time in the press. You just can't believe everything you read. -- Richard Crisp crisp@mips.com MIPS Computer Systems !decwrl!mips!crisp 928 Arques MS 2-02 (408) 524-8177 Sunnyvale, Ca 94086
astevens@acorn.co.uk (Ashley Stevens) (10/31/90)
In article <42483@mips.mips.COM> mash@mips.COM (John Mashey) writes:
->>mash> As is well-known, we are working hard with our colleagues at
->>mash> B.I.T. to improve yields on the ECL chips.
->>
->>In the latest issue of Byte I read that BIT have *dumped* the R6000
->>development. Am I hallucinating? .....
->Yes, or Byte is.
Somewhere (I think it was either the latest issue of PCW, or the Computer
Guardian) I read that BIT had dropped the development of a special version
of the R6000, for DEC, and that the normal version was delayed.
--------------------------------------------------------------------------
Ashley Stevens (disclaimer: I speak for myself etc)
astevens@acorn.co.uk
Acorn Computers, 645 NewmarketRd, Cambridge, UK. Tel.(0223) 214411pcg@cs.aber.ac.uk (Piercarlo Grandi) (10/31/90)
On 29 Oct 90 21:00:54 GMT, crisp@mips.COM (Richard Crisp) said:
crisp> In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk>
crisp> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes:
pcg> In the latest issue of Byte I read that BIT have *dumped* the R6000
pcg> development. Am I hallucinating? The justification given was that
pcg> it is
Incidentally: that piece of news I may have read on EDS actually. It
seemed pretty definite -- apaprently the guys were quoting from some BIT
source. On the other hand if John Mashey says this is not true, he
probably knows better than me and whoever was the source.
crisp> I also remember back in 1978 or 1979 that Byte stated that
crisp> Motorola would offer a version of the 68000 that had the
crisp> microcode in EPROM so that users could re-microprogram it for
crisp> whatever instruction set they wanted!
Haha. This piece of news at least was factually true. If you ordered
enough of them you could ask Motorola to change the onchip microcode ROM
mask for the 68k. I seem to remember that there were other cases, but
the well known one was IBM, who actually reprogrammed the 68k to emulate
a subset of the 370 architecture, and was able to run VM/370 on it (with
some effort). The oddest thing was that the 370/68k was actually a
coprocessor in a revamped IBM PC, alongside an 8088. Ah wonders of
technology.
As to other rumours, I read recently, again on Byte or on the net, that
the 88k has enough spare die to also be able to store some interesting
chunk of microprogram. This to me makes much less sense, but who knows?
--
Piercarlo "Peter" Grandi | ARPA: pcg%uk.ac.aber.cs@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uksritacco@hpdmd48.boi.hp.com (Steve Ritacco) (10/31/90)
I would guess that the supposed Byte article was refering to DEC's R6000 effort which was canceled, not BIT's !!!!!
mash@mips.COM (John Mashey) (10/31/90)
In article <3886@acorn.co.uk> astevens@acorn.co.uk (Ashley Stevens) writes: >In article <42483@mips.mips.COM> mash@mips.COM (John Mashey) writes: >->>mash> As is well-known, we are working hard with our colleagues at >->>mash> B.I.T. to improve yields on the ECL chips. >> >->>In the latest issue of Byte I read that BIT have *dumped* the R6000 >->>development. Am I hallucinating? ..... >->Yes, or Byte is. >Somewhere (I think it was either the latest issue of PCW, or the Computer >Guardian) I read that BIT had dropped the development of a special version >of the R6000, for DEC, and that the normal version was delayed. ARGH! I'm sure my polemic on being careful about the press passed this one coming the other way. Here's a great example of the warp factor in action: Basically, most of what was described is wrong.... 1) DEC DID say publicly that they'd cancelled an R6000-based product, although not necessarily future ECL ones, and certainly NOT MIPS-chip based things in general. (I mention this, because the press got that widely confused, in every conceivable way). 2) As far as I know, special version == normal version. I never heard of any special version of the R6000 for DEC, but if anyone KNOWS of one, please correct me. 3) Regarding "delay of the normal version", if that's supposed to mean that it wouldn't be available for a while, that's wrong. If it means that it's taken longer to get better yields than we'd like, or that its clock rate is not as high as we'd like, then that might make sense. Note that a lot of faster micros have been described in the literature, in detail .... but haven't shipped yet, or really have been cancelled... existence is a virtue. BTW, since it gets tiring fixing wrong stuff, for context: *commercial message on* the R6000 remains the fastest general-purpose microprocessor being shipped today. Although a few others may be in the same league on floating-point, depending on benchmark mix, it is 1.7X - 2X faster on integer than anything else currently shipping (according to SPEC data). We have an interesting internal example of why somebody might care: on an RC6280, running a gate-level simulator, simulating an approximately-1M-transistor chip, it takes 8 DAYS to boot UNIX from reset to single-user prompt . Note that such things are pure integer, don't vectorize, and are nontrivial to split up. It would be fun to run, for example, the SPEC benchmarks this way, except that it would take 1-2 years. *commercial off* -- -john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc> UUCP: mash@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash DDD: 408-524-7015, 524-8253 or (main number) 408-720-1700 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086
peter@ficc.ferranti.com (Peter da Silva) (10/31/90)
In article <2804@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes: > In article <P0R6UO2@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: > | > It is fairly cheap (starting prices ~$4999). > | Sorry. That's not "fairly cheap". That's not even "moderate". AT bus machines > | with 386es in them start under $2000. > Not with 4MB, FPU, ethernet, and 1152x900 displays, they don't. The 1152x900 display is the only part of that that's a significant cost component... the memory, ethernet, and FPU are in the few-hundred range. Buddy of mine just bought 1024x768 VGA (within spitting distance... besides, it's color) for under $1000 with the card and monitor. You're still well over 386 box prices. Sure, yours is faster... but the CPU is cheaper. Oh yeh, I'm including a disk. Yours is diskless, right? Basically, it should be possible to sell the thing for the $3000 to $4000 range and still make a profit. THAT would compete with 386 UNIX boxes in price and kill them in performance. The only thing missing would be 386 ABI compatibility. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
crisp@mips.COM (Richard Crisp) (10/31/90)
In article <PCG.90Oct30191923@teachk.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: > >crisp> I also remember back in 1978 or 1979 that Byte stated that >crisp> Motorola would offer a version of the 68000 that had the >crisp> microcode in EPROM so that users could re-microprogram it for >crisp> whatever instruction set they wanted! > >Haha. This piece of news at least was factually true. If you ordered >enough of them you could ask Motorola to change the onchip microcode ROM >mask for the 68k. I seem to remember that there were other cases, but >the well known one was IBM, who actually reprogrammed the 68k to emulate >a subset of the 370 architecture, and was able to run VM/370 on it (with >some effort). The oddest thing was that the 370/68k was actually a >coprocessor in a revamped IBM PC, alongside an 8088. Ah wonders of >technology. > Perhaps I missed something, but I see no evidence of the factuality of Byte's assertions in the above statement. I mentioned EPROM implemented microcode store, Mr. Grandi has said nothing to support his assertion that Byte's claims were true. All that has been offered as evidence was the fact that IBM had a custom derivative of the 68K built by Motorola. As I mentioned earlier, more than the microcode was changed, the "core" or execution unit was also changed in order to support IBM's requirements. In fact there were two chips that were part of this program, one called the "Cascadilla Minor" and the other called the "Cascadilla Major". The suffixes reflected the level of deviation from the original 68K design. I seem to remember that it took two "minors" and one "major" to implement the module mentioned above. Let's remember what the original thread was about: The trade rags oftentimes hear of rumors and publish factually incorrect stories. Just like the errors made in newspapers, magazines and television (remember the recent case where "20/20" featured a story about the man who played "Buckwheat" in the Our Gang show? Well it turned out that Buckwheat had been dead for 10 years and the man featured on the show was an imposter!), our lovable trade rags aren't infallible. The prudent reader will take rumors as just that. -- Richard Crisp crisp@mips.com MIPS Computer Systems !decwrl!mips!crisp 928 Arques MS 2-02 (408) 524-8177 Sunnyvale, Ca 94086
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/31/90)
In article <PCG.90Oct30191923@teachk.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: | As to other rumours, I read recently, again on Byte or on the net, that | the 88k has enough spare die to also be able to store some interesting | chunk of microprogram. This to me makes much less sense, but who knows? Loadable control store is a great idea, and can really improve the performance of a program. It becomes awkward in a shared CPU environment, where a bad implementation could mean a context switch includes a change of instruction set! If it were managed like a non-sharable device, it would be doable, but I have doubts about the number of people who could make use of the ability, and it sort of implies access to kernel mode or a way to lock certain status bits (makes things truely complex). In all the years we have had VAXen, and have had hackers of both the brilliant and demented types, and only one (as far as I know) person actually added iseful and working instructions. The FFT1 and FFT2 instructions do give a 40-50% reduction in CPU time for certain problems, but they represent such a small percentage of use that I doubt anyone ever loads them. They are hopefully preserved in his thesis for future need. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Haloween costume: stark naked in a body bag
pcg@cs.aber.ac.uk (Piercarlo Grandi) (10/31/90)
mash> As is well-known, we are working hard with our colleagues at
mash> B.I.T. to improve yields on the ECL chips.
On 29 Oct 90 19:44:09 GMT, mash@mips.COM (John Mashey) said:
mash> In article <PCG.90Oct28162504@teachh.cs.aber.ac.uk>
mash> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes:
pcg> In the latest issue of Byte I read that BIT have *dumped* the R6000
pcg> development. Am I hallucinating? .....
mash> Yes, or Byte is.
Well, I must post a correction; my recollection was inexact. The source
is PCW (Byte clone in the UK) and it quotes some USA newsletter, and DEC
sources.
What is reported is that in light of what is called a 'disaster' (John
Mashey himself was not happy at having several expensive machines that
cannot be delivered) at BIT on the R6000, DEC (but not MIPS or BIT) have
pulled out of the R6000 development, citing as reason that they reckon
they will have equivalent CMOS technology not much later than BIT will
have the R6000 ready, thus reducing the ECL opportunity window to one or
at most two years.
This is sad, because I like ECL (for affective reasons, let's say).
--
Piercarlo "Peter" Grandi | ARPA: pcg%uk.ac.aber.cs@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.ukhenry@zoo.toronto.edu (Henry Spencer) (11/01/90)
In article <1990Oct30.164155.10708@mozart.amd.com> brett@cayman.amd.com (Brett Stewart) writes: >>Incidentally, why not make a MIPS PC/AT compatible? I mean, a machine >>that has an R3000 chip set instead of a 386 chip set, and is otherwise >>identical (can use the same peripherals, boards, cages, etc...)... > >Deskstation Technology has done exactly this with the Am29000... One should note, however, that the usefulness of this is greatly exaggerated. You say you want to use all those nifty AT peripherals? Well, just how do you plan to write device drivers for them? You don't get hardware docs for them. In fact, more often than not there *are* no hardware docs to be had, no matter how cozy you are with the manufacturer: there are a couple of gnomes chained in a back room who hack the 8088 assembler code in the on-board ROM until it works -- hack as in "hacked to pieces with an axe" -- and that ROM code is the only documentation on how the board really works. Rotsa ruck trying to write a driver from that. (I haven't tried this myself, but people who have tell me that it really is that bad.) -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
my@dtg.nsc.com (Michael Yip) (11/01/90)
After all this price arguments, did anyone include the cost of the
software (eg the UNIX OS and util) into the comparsion?
-- Mike
PS: By the way, the 1024x768 monitor + video card is probably
an interlaced display. And that is A LOT cheaper than the
1120x1024 display that is non-interlace (I assume).mash@mips.COM (John Mashey) (11/01/90)
In article <1095@dg.dg.com> quirk@dg-rtp.dg.com (Peter Quirk) writes: >The 88K is not going away. In fact, Motorola has announced (leaked?) >news of the 88110 >recently and published a broad brush roadmap for the 88K cpu for the >next ten years. >They predict they will be delivering 400 MIPS on a CMOS chip before the >end of the decade. >Almost as interesting was the announcement of the 88300 - a combination >of the 88110 and >68030 I/O for embedded controller applications. All those 68030-based >controllers will >start screaming when fitted with an 88300. Sigh. There is a discussion theme going on, here, and in comp.arch, about the unreliability of second/third/fourth-hand information. The discussion takes the form of: A: I hear that magazine X printed that Y said such and such, or that company Q is doing or has done Z. B: I was there, Y didn't say that, X got it wrong OR: I work for Q, and we didn't do Z; here are the true facts I thought that more of the details of this particular issue (88110) were covered in <42311.mips.mips.com>, posted October 23. If thatr got lost somewhere, I'll repost it: it quoted the interesting parts of the presentation, a copy of which is on my bookshelf. The number given was 4000 mips/chip, and the presentation at Microprocessor Forum was explicitly NOT an announcement; no dates were given, other than to say (under pressure from audience) it would be announced next year. A single foil covered the 88300 family: "Family of integrated processor products 88000 Architecture compatible Emphasis: High integration Low cost Low power Modular design Compatible with 68300 family I/O modules Leverages 88110 technology, tools, and software" *opinion* that is not an announcement, and if you can tell me from that what they'll look like and whether or not they'll be competitive with other parts on the market whenever they come out, I am imnpressed. Just for calibration: 1) The 68040 was described (not announced) at Hot Chips, 6/89. 2) In 1986 Motorola presentations, the 78000 (previous number of 88000) had the following schedule: Alpha parts July 1987 Beta Oct 87 Production (200-500 sets) April 88 MC Production July 88 The foils also say (exact quote): "1987 - 20 MHZ CMOS 78000 MPU and 78200 CMMU CHIP SET (13 MIPS AND 6 MFLOPS) 1989 - 30 MHZ 78000 CHIP SET (20 MIPS AND 10 MFLOPS) 1989 - A/I ORIENTED DERIVATIVE PROCESSOR AND CMMU CHIP SET (>2 MLIP) * TAG PROCESSING IN PROCESSOR * GARBAGE COLLECTION IN CMMU 1990 - VECTORIZED FLOATING POINT 1991 - GAAS INTEGER UNIT (>50 MIPS)" You may recall that the 88K was announced 2Q88, but it was about 3Q89 before many production chips were shipped, given the FP bugs. Now, this is NOT to say that Moto is bad and evil, and says things that don't happen. Vendors often have plans they believe in, and things just don't work that way, and t his happens to almost everybody, especially since customers DEMAND a 10-year roadmap, when NOBODY really knows exactly what they're going to do in 5 years, much less 10. However, the point is: the industry right now is undergoing a terrific "futures war" in which everyone outpredicts everybody else, and wonderful bubble charts are drawn to show futures. In addition, at conferences, people describe one chip set after another, each faster. The only problem is that many of them NEVER come to pass. Of things described in Hot CHips and Microprocessor Forum, within last 18 months, at least 3-4 CPUs described to eager audiences have already been cancelled before they were ever shipped, and many more had better have some Good Luck if they're going to make it soon enough to be interesting. -- -john mashey DISCLAIMER: <generic disclaimer, I speak for me only, etc> UUCP: mash@mips.com OR {ames,decwrl,prls,pyramid}!mips!mash DDD: 408-524-7015, 524-8253 or (main number) 408-720-1700 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086
richard@aiai.ed.ac.uk (Richard Tobin) (11/02/90)
In article <1481@frapper.nsc.com> my@frapper.UUCP (Michael Yip) writes: >After all this price arguments, did anyone include the cost of the >software (eg the UNIX OS and util) into the comparsion? With luck, within a year or so, they may be two industrial-strength free Unixes available - GNU, from the Free Software Foundation, and 4.4-detox (ie BSD "detoxified" - with the AT&T code removed). If this happens (and of course, we've all been looking forward to it for some time now...) it could greatly open up the market for unix workstations. -- Richard -- Richard Tobin, JANET: R.Tobin@uk.ac.ed AI Applications Institute, ARPA: R.Tobin%uk.ac.ed@nsfnet-relay.ac.uk Edinburgh University. UUCP: ...!ukc!ed.ac.uk!R.Tobin
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (11/02/90)
In article <3686@skye.ed.ac.uk> richard@aiai.UUCP (Richard Tobin) writes: | If this happens (and of course, we've all been looking forward to it | for some time now...) it could greatly open up the market for unix | workstations. Right. Just like MINIX did, Coherent did, etc. You can get UNIX SysV.3 for <$1k now, today, from several people. You can't get a workstation size performance (ie. enough resolution and CPU to run X) for less than $4k, and you have to build it yourself to do it. The cost of the extra disk to hold the source is more than the cost of UNIX from a commercial vendor. If and when these versions come out and get as reliable as the commercial products they will be effective in opening some new markets, people who want to play with the o/s, but buying from a vendor is cost effective for a business. Note that companies pay Cyngus $100k/yr each to not have to patch, upgrade, install, etc, the GNU software. And I suspect that the cost for a whole o/s would be at least double that. There are just not enough competent people to do compiler or o/s work to change the market much. If you doubt the lack of good people just hink about the quality of the products turned out by some highly successful vendors ;-) -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Halloween costume: stark naked in a body bag
pcg@cs.aber.ac.uk (Piercarlo Grandi) (11/03/90)
On 31 Oct 90 23:10:23 GMT, mash@mips.COM (John Mashey) said: mash> Sigh. There is a discussion theme going on, here, and in comp.arch, mash> about the unreliability of second/third/fourth-hand information. mash> The discussion takes the form of: mash> A: I hear that magazine X printed that Y said such and such, mash> or that company Q is doing or has done Z. mash> B: I was there, Y didn't say that, X got it wrong mash> OR: I work for Q, and we didn't do Z; here are the true facts Yup, but gossip is not just for old ladies :-(. On this subject, I'd like to have any news about two interesting developments, one of which is somehow related to foils on Motorla plans that you quote with a smile: mash> The foils also say (exact quote): mash> [ ... ] mash> 1991 - GAAS INTEGER UNIT (>50 MIPS)" As to GaAs wonders, I seem to recall (hallucinating again?) reading, to my great surprise, that the 68030 had been GaAs'ed, and the military architecture derived from the MIPS one had too. I seem to recall that both chip sets (part of the VHSIC programme? -- how it is going, if it is still going?) had really impressive frequencies etc... The news were as of last year (1989), at least for the prototypes. Has anybody got any news on thsi interesting subject? -- Piercarlo "Peter" Grandi | ARPA: pcg%uk.ac.aber.cs@nsfnet-relay.ac.uk Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk
henry@zoo.toronto.edu (Henry Spencer) (11/03/90)
In article <3686@skye.ed.ac.uk> richard@aiai.UUCP (Richard Tobin) writes: >With luck, within a year or so, they may be two industrial-strength >free Unixes available - GNU, from the Free Software Foundation... I would be really surprised to see a GNU kernel (never mind a *system*, which requires dozens of utilities to be even marginally usable) within a year. I'd be even more surprised if it resembled Unix much. -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
meissner@osf.org (Michael Meissner) (11/05/90)
In article <PCG.90Oct30191923@teachk.cs.aber.ac.uk> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: | As to other rumours, I read recently, again on Byte or on the net, that | the 88k has enough spare die to also be able to store some interesting | chunk of microprogram. This to me makes much less sense, but who knows? If they had that much die space, they should have fixed some of the 'features' of the 88k, such as having a real signed divide instruction, and true IEEE Nan/infinity/denorm support in the hardware. -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Do apple growers tell their kids money doesn't grow on bushes?