kiki@uhunix1.uhcc.Hawaii.Edu (04/01/91)
[someone speculates about evolution of Atari TT models with Motorola 68040 cpu]
I guess it would be safe to assume that Atari is considering the '040 cpu for
their next generation along the TT product line. But the microprocessor spec-
trum has shifted from the dominance of the 80x86 and 68000 families from Intel
and Motorola. One alternative is mentioned by the author of the following art-
icle. And although he seems to disdain the specifications of the chip, I think
it might be of significance and worthy of serious consideration.
-------------------------[start of article]------------------------------------
>From: neideck@kaputt.enet.dec.com (Burkhard Neidecker-Lutz)
Subject: PgC 7600 (was Re: Second-generation RISC)
Organization: CEC Karlsruhe
I'm typing what I've read in a local computer magazine here in Germany.
The magazine isn't exactly known for overwhelming accuracy, so beware.
Summary: I'm underwhelmed.
Done by small British company called PgC, Ltd. (Yawn). Funded by Clive
Sinclair (Oha !). RISC machine with a couple of onboard systems to make cheap
machine implementation possible. The components:
- Timer (connects to IFU)
- Serial control unit (SCU) connects to IFU and external world
- Instruction Fetch Unit (IFU) connects to ECU and external world
- Execution Control Unit (ECU) connects to a ROM and QCache
- CPU 6ns self clocking design coupled with 160 Byte of 3ns RAM
- Memory Controller (MCU) connects to CPU, QCache and external world.
The ROM can be used to store emulation code for CISC instructions ( I assume
this is similar to what Clipper has). Price abotu 600 Deutsche Mark ( 400 $)
External RAM interface:
Separate Data and Address, 160 MByte/sec. bandwidth. Qcache is 32 instructions
deep. Integrated support for dual-ported SRAMS in MCU, claim that this enables
multiprocessor systems. At last a (buggy) table showing planned routes
into the future:
Name When Price Technology MIPS
PgC7600 1/91 400 $ Bipolar 200
PgC7610 2/92 40 $ CMOS 80
PgC7620 1/93 100 $ Bipolar 250
PgC7700 2/93 400 $ CMOS 1000
PgC7710 4/93 200 $ Bipolar 2000
If you ask me: Too little, too late, no serious performance, no software.
Forget it.
Burkhard Neidecker-Lutz, CEC Karlsruhe
------------------------------[end of article]---------------------------------
If the PgC series has the capability of adaptable microcode and CISC emulation,
then it would merit consideration by Atari as an alternative to the Motorola
68000 line. At the very least, it would offer Atari a little leverage against
Motorola's monopoly of the '030 and '040 market. At best, Atari would have a
computer that could run 80x86/68000/6502 etc. binary code, by loading the proper
microcode.
Although the German author perceives the PgC chips as unviable, there are other
firms that are ready to market similar chips (NexGen, San Jose, CA) that have
CISC emulation and multiprocessing capabilities. The pricing and performance
seem to be exceptional to me and are some of the qualities associated with
Clive Sinclair (and perhaps Atari), so that a Timex 200000K could soon be in
our hands?
The British press probably has reported on the PgC chips more extensively and
I wonder if anyone there could make corrections, additions and their perspect-
ive on the preceding information?
Thanks,
Jackelmar@lut.ac.uk (Mohammad A. Rahin) (04/01/91)
Keywords: Motorola 68040 In article <12229@uhccux.uhcc.Hawaii.Edu> kiki@uhunix1.uhcc.Hawaii.Edu writes: >[someone speculates about evolution of Atari TT models with Motorola 68040 cpu] > >I guess it would be safe to assume that Atari is considering the '040 cpu for >their next generation along the TT product line. But the microprocessor spec- >trum has shifted from the dominance of the 80x86 and 68000 families from Intel [stuff deleted] > >-------------------------[start of article]------------------------------------ > >>From: neideck@kaputt.enet.dec.com (Burkhard Neidecker-Lutz) >Subject: PgC 7600 (was Re: Second-generation RISC) >Organization: CEC Karlsruhe > >I'm typing what I've read in a local computer magazine here in Germany. >The magazine isn't exactly known for overwhelming accuracy, so beware. > >Summary: I'm underwhelmed. > >Done by small British company called PgC, Ltd. (Yawn). Funded by Clive >Sinclair (Oha !). RISC machine with a couple of onboard systems to make cheap >machine implementation possible. The components: > [stuff deleted] > >Name When Price Technology MIPS >PgC7600 1/91 400 $ Bipolar 200 >PgC7610 2/92 40 $ CMOS 80 >PgC7620 1/93 100 $ Bipolar 250 >PgC7700 2/93 400 $ CMOS 1000 >PgC7710 4/93 200 $ Bipolar 2000 > >If you ask me: Too little, too late, no serious performance, no software. >Forget it. > > Burkhard Neidecker-Lutz, CEC Karlsruhe > >------------------------------[end of article]--------------------------------- > >If the PgC series has the capability of adaptable microcode and CISC emulation, >then it would merit consideration by Atari as an alternative to the Motorola >68000 line. At the very least, it would offer Atari a little leverage against [stuff deleted] >The British press probably has reported on the PgC chips more extensively and >I wonder if anyone there could make corrections, additions and their perspect- >ive on the preceding information? > >Thanks, >Jack See Byte latest (March 1991) issue. Dick Pountain wrote a very informative article on this chip (PgC). Cheers. Rahin ------------------------------------------------------------------------------ _ _ _ __ _ __ ' ) ) ) / ) ' ) ) / / / / /--/ /--' __. /_ o ____ / ' (_/ (_ / \_(_/|_/ /_<_/ / <_ M.A.Rahin@uk.ac.lut -> from within UK M.A.Rahin@lut.ac.uk -> from outside UK
keithb@hemel.bull.co.uk (Keith Bedford) (04/03/91)
Re the Pgc cpu - see Byte March 89 (maybe only the International edition?) for more details. ============================================================================== Keith Bedford keith.bedford@hemel.bull.co.uk Bull HN Information Systems Ltd Maxted Road - UK03-HM14 Hemel Hempstead Herts HP2 7DZ England +44-442-232222 ext 4520 ==============================================================================
plinio@turing.seas.ucla.edu (Plinio Barbeito) (04/04/91)
In article <12229@uhccux.uhcc.Hawaii.Edu> kiki@uhunix1.uhcc.Hawaii.Edu writes: >[someone speculates about evolution of Atari TT models with Motorola 68040 cpu] A reasonable speculation, IMHO. And as long as we're talking about a 1993 time frame, it would also be fair to bring even the 68050 into the discussion. >I guess it would be safe to assume that Atari is considering the '040 cpu for >their next generation along the TT product line. But the microprocessor spec- >trum has shifted from the dominance of the 80x86 and 68000 families from Intel >and Motorola. One alternative is mentioned by the author of the following art- Not in terms of volume sold, you mean the ever-elusive claims of performance by chip manufacturers. >icle. And although he seems to disdain the specifications of the chip, I think >it might be of significance and worthy of serious consideration. > >-------------------------[start of article]------------------------------------ [...large portion deleted] > >Name When Price Technology MIPS >PgC7600 1/91 400 $ Bipolar 200 >PgC7610 2/92 40 $ CMOS 80 >PgC7620 1/93 100 $ Bipolar 250 >PgC7700 2/93 400 $ CMOS 1000 >PgC7710 4/93 200 $ Bipolar 2000 > >If you ask me: Too little, too late, no serious performance, no software. >Forget it. >------------------------------[end of article]--------------------------------- > >If the PgC series has the capability of adaptable microcode and CISC emulation, >then it would merit consideration by Atari as an alternative to the Motorola >68000 line. At the very least, it would offer Atari a little leverage against >Motorola's monopoly of the '030 and '040 market. At best, Atari would have a It would be more fair to call it a monopoly if Motorola's prices were unreasonable (like intel's). >computer that could run 80x86/68000/6502 etc. binary code, by loading the >proper microcode. Promising 80 MIPS by Feb. 1992 for $40 in (large volumes, I assume)? Sounds great, but it is just that -- a promise. Has intel ever made good on the promise that its i860 would run at 50MHz so that the 150 max MIPS figure hyped by the press could come to light? How fast does the chip run under normal conditions (i.e. with interrupt processing ON, using usual compiler output, using affordable (read slower) chips for memory)? Anyone want to take bets on whether intel's benchmarks had interrupts enabled, or ran the dhrystone benchmark from 80nsec memory? At least with a 68040, with its familiar instruction set, we know what kind of performance to expect by simply looking at the average number of instructions per cycle and the capacity of the cache. Other questions to be resolved: how efficiently would the processor emulate a 68000; is the microcode well geared to emulate its rich set of addressing modes and (small) irregularities? Would it have enough micromemory to do it? Can it address enough registers, or would it have to draft part of the cache for this role, inevitably slowing things down? How fast can it do a multiply or divide (it would have to do this very quickly in general purpose microcode to beat out an 040 with a built-in FPU with dedicated microcode)? We wouldn't know all of this until the machine and emulation software were implemented. That's a long way off. Also, a lot of fast cpus will be introduced in this year of 1991, so that 80 MIPS (whatever that means in real terms of performance) may not seem as fast by comparison in one year. And 1993? Well, in computer years, that's ages hence. By then, the 68050 may be a more familiar sight. It is difficult to speculate on numbers relating its performance with the information we have now, but...there was a recent announcement that Moto just built the largest, most dense chip yet (for the US Navy, too advanced for mortals like us to buy, with a good amount of space devoted just to nuclear-proofing redundancy), and one must begin to wonder what they're going to be able to do with their commercial products (that don't need to waste any space for redundancy) by 1993. >CISC emulation and multiprocessing capabilities. The pricing and performance >seem to be exceptional to me and are some of the qualities associated with >Clive Sinclair (and perhaps Atari), so that a Timex 200000K could soon be in >our hands? I'm all for power without price as much as the next guy, but my general feeling is that it's too soon to tell. As long as one can speculate all the way out to 1993, one can speculate about a lot of other things, many of them silly...like the Common Market banning American chips because the high frequencies might disrupt brain processes that control the ability of consumers to choose for themselves between labels stating "with hormones" or "without", or the Japanese opening their market to American rice due to a widespread boycott of Japanese submarine propellers :-P. When I see the TT040, or the PgC-ST at my dealer...or even at Toys-R-Us...I'll be convinced. Until then we can only make suggestions to Atari and hope that they'll listen. Right now we are having enough trouble getting them to put a 1.44 Meg drive into the TT...(by the way, TRH, your portable design is fine the way it is, IMHO). plin -- ----- ---- --- -- ------ ---- --- -- - - - plinio@seas.ucla.edu This page intentionally left blank so that it could contradict itself.
vsnyder@jato.jpl.nasa.gov (Van Snyder) (04/05/91)
In article <2318@lee.SEAS.UCLA.EDU> plinio@turing.seas.ucla.edu (Plinio Barbeito) writes: >... >Sounds great, but it is just that -- a promise. Has intel ever made >good on the promise that its i860 would run at 50MHz so that the 150 >max MIPS figure hyped by the press could come to light?... The 150 MIPS figure is the speed it's guaranteed not to exceed :-) -- vsnyder@jato.Jpl.Nasa.Gov ames!elroy!jato!vsnyder vsnyder@jato.uucp
johns@maccs.dcss.mcmaster.ca (Conan the Barbarian) (04/05/91)
Are people seriously replying to this message? Wouldn't it be wise to be at least a little skeptical? >Article 32880 of comp.sys.atari.st: >>From: kiki@uhunix1.uhcc.Hawaii.Edu >Newsgroups: comp.sys.atari.st >Subject: Atari cpu evolution >Keywords: Motorola 68040 >Message-ID: <12229@uhccux.uhcc.Hawaii.Edu> >Date: 1 Apr 91 07:31:56 GMT ^^^^^ Does this date mean anything to anyone? >Sender: news@uhccux.uhcc.Hawaii.Edu >Organization: University of Hawaii Or maybe this is one of them there prolonged jokes, and the joke is on me for not recognizing it as such. <insert fav. smiley> -- John Schmitt johns@maccs.dcss.mcmaster.ca ...!unet!utai!utgpu!maccs!johns
Bob_BobR_Retelle@cup.portal.com (04/05/91)
Lotsa speculation about Atari using '040 and beyond.. Golly.. Lenonard Tramiel has already stated that Atari will never use the 68040 (I think it was at the Glendale AtariFest last year..?), so why worry about it..? Atari wouldn't lie to us, would they..? BobR
buggs@cup.portal.com (William Edward JuneJr) (04/07/91)
What about all treal CHEAP & POWEFULLY SPARC clones? Seems that'd be a better choice, at least to me... You uld have either software emulation or hardware coprosessors to run existing software.
kiki@uhunix1.uhcc.Hawaii.Edu (Jack W. Wine) (04/08/91)
[previous statements about British PgC cpu series and influence on Mot./Intel] >From: plinio@turing.seas.ucla.edu (Plinio Barbeito) >It would be more fair to call it a monopoly if Motorola's prices were >unreasonable (like intel's). I would contend that their decision to not license their 030/040 and 386/486 architectures to other companies was very short-sighted. If Motorola had allowed their chips to be second sourced, especially with Japanese companies like Toshiba and Hitachi, they would have attained a much larger share of the market. Distributing the development of the cpus would have allowed Motorola and their licensees to debug and produce their designs much faster. Instead, the 040 arrived more than a year overdue and does not meet its performance expectations. As for Intel, from what I read, they seem to be simply unscru- pulous with their licensees and I guess it would have been out of character for them to have an open architecture. [PgC 7610 specs and availability...] >Promising 80 MIPS by Feb. 1992 for $40 in (large volumes, I assume)? >Sounds great, but it is just that -- a promise. Has intel ever made >good on the promise that its i860 would run at 50MHz so that the 150 >max MIPS figure hyped by the press could come to light? Someone posted that there was an article on the PgC chips in the international edition of Byte. It is supposed to be in the March '91 (not '89) issue and perhaps someone from Britain might post more information. >Other questions to be resolved: how efficiently would the processor >emulate a 68000... We wouldn't know all of this until the machine and >emulation software were inplemented. The following information about a similar processor might provide more proof of the viability of the PgC design: Teraplex (Champaign, IL) developed a 32 bit processor with a 128 bit instruction word (PgC 7600 uses 96 bit instr.). The Teraplex cpu does not have microcode or decoding circuits so it uses the the first 64 bits of the instr. word for hardware control, while the second 64 bits are the instruction operands. The instructions are very primitive so that it is capable of mimicking other processors. Actual systems built with this MISC (minimal instruction set computer) execute MS-DOS programs at about 4.5 times the speed of a 33Mhz '386. Although it does not have a FPU, the Teraplex 32 bit CMOS design is about as fast as a MIPS R3000 with an FPU. Other systems based on the MISC are running MIPS and SPARC programs and the company is also investigating running Motorola 68000 code. Teraplex hopes to have commercial workstations and desktop computers based on the chip, by the fourth quarter of 1991. Refer Byte, Nov. '90, pgs. 19-20. >Also, a lot of fast cpus will be introduced in this year of 1991, so >that 80 MIPS (whatever that means in real terms of performance) may >not seem as fast by comparison in one year. And 1993? Well, in >computer years, that's ages hence. By then, the 68050 may be a more >familiar sight... It's hard to make predictions, but it seems certain that one of the dominant OS running on the new processors will be UNIX and the SPARC RISC was designed to run it optimally. Also, Sun's decision to have an open architecture is resulting in a flood of systems based on the SPARC chip set. A partial list of companies introducing SPARC systems are Northgate, Chicony Electronics, CompuAdd, Hyundai, Tatung, DTK Computer, Opus, RDI/TriGem, DCM Data, Mars, Sampo, Solbourne, Meiko World, ICL, Solarix, Toshiba, Twinhead, etc... It would appear that the SPARC-UNIX combo will mirror what happened with the MS DOS-INTEL systems; you will be able to get a very powerful system, very cheap. Atari will have a hard delivering an '040 system that will be competitive with the SPARC-based systems. If the specs for the PgC and Teraplex chips hold true, then it would seem to me that Atari could have an amazing machine that would be capable of running programs faster than the target system! >From: Bob_BobR_Retelle@cup.portal.com >Golly.. Lenonard Tramiel has already stated that Atari will never use >the 68040 (I think it was at the Glendale AtariFest last year..?), so >why worry about it..? I was reading some of the old Znet and STreports on Freenet (thanks to the people who upload them) and Alwin Stumpf of Atari Germany supposedly said in an interview that they would use the 68040 if its price were reasonable. Is Leonard Tramiel head of R&D? I remember reading an article about a guy name Shiraz Shivji. Is he still around at Atari? Jack
Shervin.Shahrebani.Of.250/744@f744.n250.z1.FidoNet.Org (Shervin Shahrebani Of 250/744) (04/09/91)
Now why the hell did Lenny say no 040's? What kind of a stupid, premature decision is that? I hope he is wrong. S.S.
mathew@mwowm.mantis.co.uk (mathew) (04/10/91)
In <1991Apr3.115106.4220@hemel.bull.co.uk>, Keith Bedford writes: >Re the Pgc cpu - see Byte March 89 (maybe only the International edition?) >for more details. Indeed, it's only in the "International" edition, since it's a well-known fact that Americans aren't the slightest bit interested in anything which is happening in Europe. The so-called "International" section of Byte is one of the many reasons why the magazine sucks. mathew -- If you're a fan of John Foxx, please mail me.
vsnyder@jato.jpl.nasa.gov (Van Snyder) (04/11/91)
In article <12371@uhccux.uhcc.Hawaii.Edu> kiki@uhunix1.uhcc.Hawaii.Edu (Jack W. Wine) writes: > >It's hard to make predictions, but it seems certain that one of the dominant >OS running on the new processors will be UNIX and the SPARC RISC was designed >to run it optimally. Also, Sun's decision to have an open architecture is >resulting in a flood of systems based on the SPARC chip set.... From an article in April Datamation, I see that Sparc International is only half-open: Their "shrink-wrap" API requires Sun-OS. The article was about the 88000. 88Open doesn't require a certain OS to be compliant. As a result, even though only 6 US vendors have picked the 88000, there are over 2000 applications available, while for the Sparc, with some 30 or so vendors using it, there are only about 2500 applications. The article praised the 88000 and 88Open on technical grounds, but mentioned that Motorola's marketing was "methodical at best." It also mentioned that Motorola has cut the 88000 price by 2/3, and that the 88010 will be out "soon". Top-of-the-line 88000 is about as fast for non-floating-point as the fastest Sparc or MIPS, but has better support for multiprocessor architectures (according to Harris, who build a multiprocessor server from it). When 88010 is available it should be substantially faster, even for floating point, than current Sparc or MIPS. That is, almost as fast as the new HP PA chips %). (57 and 76 Mips are the speeds the newest HP machines are guaranteed not to exceed. $k12 and $k20 in greyscale, diskless.) -- vsnyder@jato.Jpl.Nasa.Gov ames!elroy!jato!vsnyder vsnyder@jato.uucp
Shervin.Shahrebani.Of.250/744@f744.n250.z1.FidoNet.Org (Shervin Shahrebani Of 250/744) (04/14/91)
Shiraz Shivji left a while ago. Leonard Tramiel (I don't know if he is head of R & D or not) has less authority than Alwin Stumph. If Alwin thinks that the TT should have a 68040, then he will convince Jack and the R & D team. I very much doubt what Leonard said. I think that eventually, Atari will be forced to use the 040. They cannot wait for the 050. They should not go to Intel because of architecture inferiority and incompatibility with their Motorla 68k line. The could and are going RISC. Atari U.K is developing a RISC computer. But I still think there will be a need for an 040 to bridge the gap. S.S.