pcg@aber-cs.UUCP (Piercarlo Grandi) (11/01/90)
On the subject of a SPARC based PC/AT clone: somebody> It is fairly cheap (starting prices ~$4999). In article <P0R6UO2@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: peter> Sorry. That's not "fairly cheap". That's not even "moderate". AT peter> bus machines with 386es in them start under $2000. On 30 Oct 90 15:45:04 GMT, davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) said: davidsen> Not with 4MB, FPU, ethernet, and 1152x900 displays, they davidsen> don't. These are aimed for a totally separate market. I davidsen> haven't tested one yet, but on paper they will produce better davidsen> cost/performance than the Sun SS+, at least at list price. Not to mention an open architecture, with the ability use hundreds of low cost accessories. But the $4999 really ought to include an hard disk (say, 80MB ESDI 20msec.) to be competitive with a Sun SLC -- also the OS may not be included. davidsen> 25MHz SPARC, FPU, MMU, 64 bit memory path, 64k cache. Color, davidsen> harddisk, 386, and thinnet are options. Incidentally, my MIPS PC/AT idea was based on seeing an advert in EDN for an R3000/R3010 evaluation module that with 64K cache and 128K SRAM would cost something like $895. This looks expensive, but still it is significantly less than an i486, or an i386+i387+cache controller, and gives about the same performance (it is actually better than a 486 on floating point, as John Mashey keeps telling us). I have no doubt that a MIPS or SPARC or AMD 29k or M88k based CPU module should cost significantly less than an i486. Could anyone with an idea of list prices for quantities like 100 pieces give us an idea of the cost of a CPU module (CPU, FPU, MMU, minimum effective cache) for each of the above architectures? Can anybody tell us how difficult it would be to design a plug in board for an i486 socket with any of the above chip designs? (I expect this to be hard). Note: i486 motherboards, without the 486, are not much more expensive than 386SX ones. Get my drift? Can anybody tell us how difficult it would be to redesign an existing i[34]86 based motherboard around each of the aboved chip designs? (I expect this to be not that difficult, but not easy either). And so on... -- 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
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (11/01/90)
In article <2081@aber-cs.UUCP> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: | Note: i486 motherboards, without the 486, are not much more | expensive than 386SX ones. Get my drift? I don't know about in thousands, but I have been looking hard at boards for my project cheap-ix, and the SX boards with CPU are about $350 in Q5, the 486 $900-1500 Q5, without CPU. Allowing $150 for the SX chip (I can't find anyone selling them separately in the pile I have handy), that would make the board about 4x more expensive. I think the use of 32 bit paths, more expensive bus control chips, and more layers in the board to keep it reliable at higher speeds account for this. Note that this is not to agree or disagree, because I think you could argue that the price diference is what counts rather than the ratio, and the ratio in a complete system price would be only 10-20% increase. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Halloween costume: stark naked in a body bag
sysmgr@KING.ENG.UMD.EDU (Doug Mohney) (11/01/90)
In article <2081@aber-cs.UUCP>, pcg@aber-cs.UUCP (Piercarlo Grandi) writes: >On the subject of a SPARC based PC/AT clone: Well, heck with that! How much trouble would it be to build a PC-clone which could double as an X-term? I'm talking about a one-piece board with built-in Ethernet and 2MB of RAM, and a '386, add a couple sockets for ROMs. You crank many of them out in Taiwan, sell them as dirt-cheap clones, or add the ROMs and a high-res monitor and you have Mr. X-term.... I guess the best bet would be a modified PC-clone, with the built-in Ethernet, some support for a local disk, and an option to replace/toggle-between the ROMs for bootup.
jonah@dgp.toronto.edu (Jeff Lee) (11/02/90)
sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >Well, heck with that! How much trouble would it be to build a PC-clone which >could double as an X-term? I'm talking about a one-piece board with built-in >Ethernet and 2MB of RAM, and a '386, add a couple sockets for ROMs. [...] >I guess the best bet would be a modified PC-clone, with the built-in Ethernet, >some support for a local disk, and an option to replace/toggle-between the >ROMs for bootup. A group here is tossing around a similar idea, but minus the disk controller and (possibly) plus a SCSI port, with SIMMs for memory (1/4/16MB). However, we're considering a RISC processor in place of the 386 (running at memory speeds to eliminate the cache and reduce the chip count). Does anyone working on current small systems design have any suggestions on how easily [i.e using little glue] current processors (29K, 88K, MIPS, SPARC, or even [34]86, 680[34]0, or ???) would or would not fit into such a low-chip count, low-end, low-quantity(?) product? Also, how easy is it to add an FPU along with those chips that don't include it? Alternatively, does anyone know of an existing board-level product with at least >=1M pixel memory mapped graphics, CPU, keyboard/mouse/serial ports, EPROM socket(s), and >=2MB RAM? Ethernet, SCSI, FPU, and TOD-clock are optional, but full hardware documentation is essential. Jeff Lee -- jonah@cs.toronto.edu || utai!jonah
torbenm@freke.diku.dk (Torben [gidius Mogensen) (11/02/90)
jonah@dgp.toronto.edu (Jeff Lee) writes: >sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >A group here is tossing around a similar idea, but minus the disk >controller and (possibly) plus a SCSI port, with SIMMs for memory >(1/4/16MB). However, we're considering a RISC processor in place of >the 386 (running at memory speeds to eliminate the cache and reduce the >chip count). Does anyone working on current small systems design have >any suggestions on how easily [i.e using little glue] current >processors (29K, 88K, MIPS, SPARC, or even [34]86, 680[34]0, or ???) >would or would not fit into such a low-chip count, low-end, >low-quantity(?) product? Also, how easy is it to add an FPU along >with those chips that don't include it? Try looking at the ARM (Acorns Risc Machine) chip set, produced by VLSI technology. The ARM is a 32bit RISC cpu, currently found in two versions, ARM2 and ARM3. It sounds like the ARM2 is best suited for your purpose, as it is without cache and designed to work fast even so. There are currently 3 support chips: MEMC, VIDC and IOC. MEMC is a memory management unit, VIDC is a video and sound controller and IOC is an I/O controller. VIDC is able to support memory mapped dispalys with virtually any resolution in up to 8 bits per pixel. It is programmable regarding frequency, and can be used on both normal frequency monitors, VGA monitors, multiscan monitors and high frequency monitors. Acorn computers have produced computers and workstations that support all of these types of monitors on the same machine. Prentice Hall publish the book: "VL86C010 32bit RISC MPU and Peripherals Users Manual" ISBN: 0-13-944968-X The book has the datasheets for: VL86C010 - ARM2 VL86C110 - MEMC VL86C310 - VIDC VL86C410 - IOC The present production version of MEMC is MEMC1a, which is an improved version of the MEMC described in the book. The cost of the complete chip set (ARM2, MEMC1a, VIDC and IOC) is lower than the price of a 386. The ARM3 is an ARM2 expanded with cache plus a few syncronisation commands. It is currently available in versions running up to 40MHz. Acorn has announced an FPU, supposed to be shipping at the end of the year. >Alternatively, does anyone know of an existing board-level product >with at least >=1M pixel memory mapped graphics, CPU, >keyboard/mouse/serial ports, EPROM socket(s), and >=2MB RAM? >Ethernet, SCSI, FPU, and TOD-clock are optional, but full hardware >documentation is essential. Acorn sells full machines with the above specifications, but I do not think they sell boards only. The top level products are UNIX workstations supporting screen modes of 1152x900 B/W uning a high frequency monitors or 800x600x256 colours using super VGA monitors. >Jeff Lee -- jonah@cs.toronto.edu || utai!jonah Torben Mogensen (torbenm@diku.dk)
sysmgr@KING.ENG.UMD.EDU (Doug Mohney) (11/02/90)
In article <1990Nov2.000650.18866@jarvis.csri.toronto.edu>, jonah@dgp.toronto.edu (Jeff Lee) writes: >>I guess the best bet would be a modified PC-clone, with the built-in Ethernet, >>some support for a local disk, and an option to replace/toggle-between the >>ROMs for bootup. > >A group here is tossing around a similar idea, but minus the disk >controller and (possibly) plus a SCSI port, with SIMMs for memory >(1/4/16MB). However, we're considering a RISC processor in place of >the 386 (running at memory speeds to eliminate the cache and reduce the >chip count). Doesn't count :-) PC compatability (for better or worse) and decent X-term capability would be more better. I suppose you could be really perverted, use a RISC processor and stick in Soft-PC in ROMs. Flip a switch and that $2000+ X-term becomes a $1000 PC-clone. Aw nuts, you'd still need to throw in a disk controller for a 3 1/2" floppy, and you'd probably have to hack the Soft-PC up a bit to support SCSI disk storage. >Does anyone working on current small systems design have >any suggestions on how easily [i.e using little glue] current >processors (29K, 88K, MIPS, SPARC, or even [34]86, 680[34]0, or ???) >would or would not fit into such a low-chip count, low-end, >low-quantity(?) product? Also, how easy is it to add an FPU along >with those chips that don't include it? You might wanna look at how the SparcStation SLC was put together; I understand there are a couple of companies offering up glue chip-sets for Sparc.
henry@zoo.toronto.edu (Henry Spencer) (11/03/90)
In article <0093F1A8.A28E4920@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >Doesn't count :-) PC compatability (for better or worse) and decent X-term >capability would be more better... Some people would consider neither a virtue. :-) Both are de-facto standards that are hideously ugly and do terrible things to your system design if you enshrine them as fundamental goals. (Retaining the potential for them, on the other hand, is easy: any fast CPU with a large address space can emulate the early Intel processors at higher speed than Intel chips, and X is not difficult to port to a sane machine with a clean frame buffer.) >...use RISC processor and stick in Soft-PC in ROMs... A small practical difficulty with this approach is that Soft PC is licensed software. It also does a less than wonderful job of emulation, as our department is finding out on a bunch of DEC RISC workstations for undergrad teaching... >You might wanna look at how the SparcStation SLC was put together; I understand >there are a couple of companies offering up glue chip-sets for Sparc. The two parts of this sentence are unrelated. :-) I haven't seen details on the SLC, but Sun normally uses proprietary MMU designs that bear no relation to (e.g.) the "Sparc Reference MMU". Worse, not only are they proprietary but they are Top Secret, although apparently Sun has entirely forgotten why, since they can't offer any rational reason for it when asked. -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
rick@ameristar (Rick Spanbauer) (11/03/90)
In article <1990Nov3.052952.1786@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >The two parts of this sentence are unrelated. :-) I haven't seen details >on the SLC, but Sun normally uses proprietary MMU designs that bear no >relation to (e.g.) the "Sparc Reference MMU". Worse, not only are they >proprietary but they are Top Secret, although apparently Sun has entirely >forgotten why, since they can't offer any rational reason for it when >asked. It may be a simple reason, like not wanting users to be able to buy only SunOS tapes from Sun and then run the binaries on cheap sparc clones, or to make the job of clone companies harder (ie buy a SunOS sources + unix guru to rewrite the MMU code). Suns choice of what they consider trade secret has never been very user friendly at all - for years, one couldn't get schematics, buy keyboard and power supplies directly from their suppliers, get compiler, rasterop lib, ND, or MMU source, etc. We tend to pay more for our Suns, yet know less about them than our Amigas, PCs, or Macs :-) >"I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology Rick Spanbauer Ameristar
sysmgr@KING.ENG.UMD.EDU (Doug Mohney) (11/04/90)
In article <1990Nov3.052952.1786@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes: >In article <0093F1A8.A28E4920@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >>Doesn't count :-) PC compatability (for better or worse) and decent X-term >>capability would be more better... > >Some people would consider neither a virtue. :-) Well, neither is the United States Congress, but we have to work with the systems which are installed, eh? ;-) > Both are de-facto standards >that are hideously ugly and do terrible things to your system design if you >enshrine them as fundamental goals. (Retaining the potential for them, on >the other hand, is easy: any fast CPU with a large address space can >emulate the early Intel processors at higher speed than Intel chips, and >X is not difficult to port to a sane machine with a clean frame buffer.) People will want X-terms and people will want PC-compatability. It is probably easier to build a PC with 1024 x 768 monitor, and be able to drop in a couple of bootstrap ROM which access your (built-in) Ethernet to load up the X-software. Of course, you could insist upon Token Ring as well, but I'm hoping it will go away. I wouldn't do this with anything less than a '386; you can keep the "earlier" Intel processors. Since AMD and others are coming out with '386 clones, the chip count would be pretty small. >>...use RISC processor and stick in Soft-PC in ROMs... > >A small practical difficulty with this approach is that Soft PC is licensed >software. It also does a less than wonderful job of emulation, as our >department is finding out on a bunch of DEC RISC workstations for undergrad >teaching... I said this was perverse. I didn't say it was optimized. I noted it would require some hacking to get it running properly. >>You might wanna look at how the SparcStation SLC was put together; I understand >>there are a couple of companies offering up glue chip-sets for Sparc. > >The two parts of this sentence are unrelated. :-) I haven't seen details >on the SLC, but Sun normally uses proprietary MMU designs that bear no >relation to (e.g.) the "Sparc Reference MMU". So? I don't think Sun has patended the way they use board real estate. See how they did things (Board in the back of the monitor <hm. What will all those zappo evil EM rays coming out the back to do the chips in the long run?>, small configuration). Go get the commercial Glue chips and go to town. Doug
henry@zoo.toronto.edu (Henry Spencer) (11/04/90)
In article <0093F295.10626840@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >People will want X-terms and people will want PC-compatability... By this reasoning, nobody should be building anything except PC-compatible X terminals. Strangely, many manufacturers manage to make money building machines that are neither. -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
henry@zoo.toronto.edu (Henry Spencer) (11/04/90)
In article <1990Nov3.150202.27353@ameristar> rick@ameristar (Rick Spanbauer) writes: >>... Sun normally uses proprietary MMU designs that bear no >>relation to (e.g.) the "Sparc Reference MMU". Worse, not only are they >>proprietary but they are Top Secret, although apparently Sun has entirely >>forgotten why, since they can't offer any rational reason for it when >>asked. > > It may be a simple reason, like not wanting users to be able to buy > only SunOS tapes from Sun and then run the binaries on cheap sparc > clones, or to make the job of clone companies harder... I said "rational reason". The problem with the idea of keeping it secret from the competition is that it's not that hard to reverse-engineer the stuff if you try. The basic design concepts were published long ago; all that is being kept secret is the details, exactly the sort of thing that a competent engineering team could figure out in a month or two, given a couple of machines to play with and modern tools. Any would-be clone company has ample resources to figure this stuff out without Sun's help. This policy makes life harder only for the legitimate customers. I can see keeping things like this secret *briefly*, when the machine is new and hot and it's worth throwing up even small obstacles to the cloners, but why are the MMU details of the Sun 2 still secret today? -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
paul@taniwha.UUCP (Paul Campbell) (11/05/90)
In article <1990Nov3.235958.21976@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >I can see keeping things like this secret *briefly*, when the machine is >new and hot and it's worth throwing up even small obstacles to the cloners, >but why are the MMU details of the Sun 2 still secret today? Probably they are still embarassed by them ...... Paul -- Paul Campbell UUCP: ..!mtxinu!taniwha!paul AppleLink: CAMPBELL.P What most people don't realize is that those plastic cover slips that your 3 inch floppies come in are actually condoms for protecting your computer from harmfull computer viruses - practice safe computing ..... :-)
graeme@labtam.labtam.oz (Graeme Gill) (11/05/90)
In article <1990Nov3.235115.21250@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes: > In article <0093F295.10626840@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: > >People will want X-terms and people will want PC-compatability... > > By this reasoning, nobody should be building anything except PC-compatible > X terminals. Strangely, many manufacturers manage to make money building > machines that are neither. From our experience I can say that Xterminal performance is limited by by three things: 1) For small operations, network bandwidth is the ultimate limit. The x11perf poly point performance multiplied by 4 is usually a good estimate of the available network bandwidth in bytes per second. 2) For large areas, memory write speed is the ultimate limit. To get good speed you need to take advantage of 32 bit access, burst writes, interleaved memory and any special write modes Vrams support. 3) Between 1) and 2), performance depends on cpu speed. Small operations area also fairly sensitive to instruction cache size, function call overhead and the working register set size. The more registers the processor has the better, since graphics code is notorious for having large numbers of function parameters, and a large number of working variables. Given the above considerations, the current PC/AT clones do not measure up well. Their network interfaces are generally based on 16 bit controller cards which shuffle data via small static ram buffers, limiting network performance to (typically) 50 - 200 Kbytes/sec. This compares to 600 - 800 Kbytes/sec for a purpose built design. The common video cards are accessible only 8 or 16 bits at a time, over the slow PC bus, and may involve bank switching and plane wise access to pixels. Neither 286 nor 386/486 processors support burst write to memory. All these problems are reflected in the performance quoted for PC X-terminal emulation - generally 10 to 100 times slower than purpose built Xterminals and workstations. Hence the market for Xterminals - workstation performance graphics on your desktop at a lower price. Graeme Gill Labtam I.S.D. Pty Ltd graeme@labtam.oz.au
davecb@yunexus.YorkU.CA (David Collier-Brown) (11/05/90)
davecb@yunexus.YorkU.CA (David Collier-Brown) writes: > Interestingly enough, there is at least one 68k-cpu pc around: the >Institute of Space and Terrestrial Studies here at Ork helped develop >a machine called the "lab 100", which is a small 68xxx-based system >which uses at-style peripheral controllers and acessory boards, and fits >in one of the small-footprint pc cases. OOPS: that should have been Lab30! Tyler Ivanco wrote back in response to my posting and gave the specs, which include OS9 instead of Unix, in part for its rtealtiem capabilities, in part for the ease of writing device drivers... Tyler writes: | Well here I am out in Calgary reading about this box. Actually, | it is called the Lab30. Avy Moise at york (x55359) can show you the unit. | Briefly: | | 16-20Mhz 030/FPU, | High Speed SCSI, | 8 32/16/8 bit DMA channels, | Any of the 8 channels can be "connected" to any request | source. | 23+7 Vectored Interrupt channels, | Series of high resolution timers, | 4MB on board DRAM, | 32KB SRAM, | 128KB EPROM, | 2 1Mbit serial channels, | Real Time Clock, | 8/16 Bit AT bus (doesn't support bus master mode), | LABus (just the 030 bus signals) to permit memory expansion, | etc. | Split board sandwich. The PCB can be altered so that it can | be cut in half and sandwiched thus reducing the footprint | but about 50%. | PC keyboard interface, | OS/9 Real Time O.S. with support for the following: | NI GPIB, | SCSI H.D., Tape Drive, Optical Drive, | Serial channels both on the motherboard and AT card, | Real Time clock, | Parallel port (on an AT card), | General Purpose event card (custom design for one of our | experiments). | Device drivers in development: | Keyboard | floppy disk | ATI VGA graphics | OS/9 Windows | OS/9 Internet (via ethernet) | AT Ethernet card | | | Most of the work is real time based hence the reason for OS/9. This | O.S. is field proven and mature. It is also easy to create, debug, | and install device drivers. The system can run without mass storage. | Unix could be run on the system, but we have no plans to port as there | are many other UNIX platforms out there that can do this job quite | nicely. | | Anyway, hope that this help you. | | Tyler | | Talk to avy. He can give you a more complete description. Alas, this tends to support both my and Henry's points (:-)) --dave -- David Collier-Brown, | davecb@Nexus.YorkU.CA, ...!yunexus!davecb or 72 Abitibi Ave., | {toronto area...}lethe!dave or just Willowdale, Ontario, | postmaster@{nexus.}yorku.ca CANADA. 416-223-8968 | work phone (416) 736-5257 x 22075
rpeglar@csinc.UUCP (Rob Peglar) (11/05/90)
In article <1990Nov3.235115.21250@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes: > In article <0093F295.10626840@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: > >People will want X-terms and people will want PC-compatability... > > By this reasoning, nobody should be building anything except PC-compatible > X terminals. Strangely, many manufacturers manage to make money building > machines that are neither. Truth is stranger than fiction. Henry's post reflects the fact (well, OK, generalization) that a) the mass market (probably ~90 %) doesn't really know what they want - they buy what they're told, or what's "hot", or whatever they see in a magazine, or what's on sale, or ... you get the point. b) the above-mentioned technology "PC-compatible X terminals" may be the "right thing" and all, but (most) companies exist to make money for the shareholders - not to spread "rightness" to the user base. c) a) + b) means that marketing, not technology or architecture, dominates the "computer" business. No one should be surprised. Sad, perhaps, but not surprised. Rob -- Rob Peglar Comtrol Corp. 2675 Patton Rd., St. Paul MN 55113 A Control Systems Company (800) 926-6876 ...uunet!csinc!rpeglar
richard@aiai.ed.ac.uk (Richard Tobin) (11/06/90)
In article <1990Nov3.052952.1786@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >Both [PC compatibility and X] do terrible things to your system design if you >enshrine them as fundamental goals. (Retaining the potential for them, on >the other hand, is easy: any fast CPU with a large address space can >emulate the early Intel processors at higher speed than Intel chips, and >X is not difficult to port to a sane machine with a clean frame buffer.) I would be interested to hear you elaborate on this - in particular, what does "enshrining X as a fundamental goal" involve other than having "a sane machine with a clean frame buffer". [Which reminds me: I was recently reading the "documentation" that came with a super VGA board, and was disgusted to see that it can only be addressed 128k at a time - you have to change banks by outputting something to a port. It occurred to me that this could all be hidden behind the 386's paging; have a 512k area with only 128k mapped and switch banks when you get a page fault. Has anyone tried this? Would it be fast enough? Followups to somewhere sane, please.] -- 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
phys169@canterbury.ac.nz (11/06/90)
In article <5530@labtam.labtam.oz>, graeme@labtam.labtam.oz (Graeme Gill) writes: >> In article <0093F295.10626840@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >> >People will want X-terms and people will want PC-compatability... >> > [good reasons why PC's running X give very limited performance] > > All these problems are reflected in the performance quoted for PC > X-terminal emulation - generally 10 to 100 times slower than purpose > built Xterminals and workstations. Hence the market for Xterminals - > workstation performance graphics on your desktop at a lower price. > Still, the demand for PC-type equipment remains - e.g .where an organisation has a big commitment to PC's already, or where stand-alone equipment is more appropriate (ever try managing a bunch of Unix systesm, one-per-site, where the users can barely manage DOS?). So, if you can't make a good Xterminal out of a PC by adding software, how about a replacement graphics card with Ethernet & 2nd CPU on-board? That would solve bandwidth problems and compatibility problems (given the right hardware design). My opinion is that now the "leading edge" of technology applies to a relatively small number of users - before, when great improvements in speed came, every computer user (from the home user up) could actually do with the extra performance; now, there are becoming less and less buyers that actually need that extra technology (although everyone will say it's nice). The "bread and butter" for computer manufacturers will remanin with the majority of users in offices and the like, where something like a PC is good enough. Power-hungary programs may come along to raise the performance requirements a bit, but not enough to justify the huge cost of R&D that an increasingly small section of the market wants. Not that I, for one, think I have enough computer power at the moment, but as more people start to realise they would be paying more for power they don't need, they will be content to stick with (say) Asian manufacturers turning out acceptable-performance machines using yesterday's technology. This seems to be the real limit of growth, not bus bandwidth or the speed of light. The one thing, other than performance, that grabs the "average" buyer (and especially corporate buyers) is the fear of equipment becoming obsolete. Curiously enough, this is working in favour of PC's and against Unix workstations (when many would argue it ought to be the other way around). The PC's have been here for such a long time, and the Unix-based workstation market seems so volatile. Manufacturers of the latter (especially m88k-based ones, according to the previous thread) should stop squabbling, and adopt some of the PC-world stability and respectability, even if the (relatively small number of) people who appreciate what's going on under the hood see the respectibility in modern designs. As I indicated above, you don't have to be saddled with all the limitations of the PC standard, to accomodate PC users. All this, of course, has a motive behind it; I've been wanting to buy some workstations which will run X-windows, with good speed, and yet not loose anything we had with 286-based PC's. Hopefully, some intelligent reader will take to heart what I have said, and produce the workstation of my dreams, by Christmas (which year?), please! :-) It would have a 16" monochrome screen with at least 32 shades (and accept any VGA mono/colour screen), an AT/VT220 compatible keyboard, rodent, optional diskette drive, 386 processor, RS232 and RS422 and parallel I-O on the motherboard, thin-wire Ethernet and multi plane video card with 2nd (purpose-built) CPU on a single card, small box (expansion can be via the network), and cost about the same as a conventional big-box 386-based PC. Okay, this isn't the place for wish-lists. The real point is, though, people have different ideas of what an ideal workstation should be an do. It becomes difficult to cater for all tastes, and still be competative. PC's have many different screens, keyboards, and so on; manufacturers need not produce all these themselves. It is tempting to say that X-windows compatibility is all that is required, that what goes on inside the box isn't the user's concern. Even now, X-terminals connected to a fast Unix box fail to give each user a good PC emulator, for example, since only one person can use the computer comfortably when that is running. Other people can probably think of good reasons why the local workstation has to perform in its own right, and what it does (and how compatible the hardware behind it is) remains important. I wish it wasn't, but people designing without that realisation will be doing their sales pitch to a smaller audience. I apologise for taking such a long time to say this, and for inserting apologies every other sentence for defending antiquated PC technology, but there are some important plusses there, and the temptation to reject it completely is rather fool-hardy, in my humble opinion. Bye for now, Mark Aitchison, Physics, University of CAnterbury, New Zealand.
peter@ficc.ferranti.com (Peter da Silva) (11/06/90)
In article <1990Nov6.101902.9683@canterbury.ac.nz> phys169@canterbury.ac.nz writes: > now, there are becoming less and less buyers that actually need > that extra technology (although everyone will say it's nice). The "bread and > butter" for computer manufacturers will remanin with the majority of users in > offices and the like, where something like a PC is good enough. I've got news for you... this has been the case since 1980 at least. Most users of IBM-PCs would be perfectly well served by a 64K Z-80-based machine running CP/M. The "gee whiz" technology really does sell machines. > All this, of course, has a motive behind it; I've been wanting to buy some > workstations which will run X-windows, with good speed, and yet not loose > anything we had with 286-based PC's. I personally can't think of anything off the top of my head that a 286 based PC offers that I want. What are you talking about here? For running MS-DOS software, an 8088 based PC is fine. For anything else, a 68000 based PC is much better. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
guy@auspex.auspex.com (Guy Harris) (11/07/90)
>The two parts of this sentence are unrelated. :-) I haven't seen details >on the SLC, but Sun normally uses proprietary MMU designs that bear no >relation to (e.g.) the "Sparc Reference MMU". I've read stuff that leads me to believe that LSI Logic has, in addition to their SPARC+reference MMU chip set, a chip set that includes a Sun MMU (perhaps the one they sell to Sun for the various Desktop SPARC machines).
graeme@labtam.labtam.oz (Graeme Gill) (11/07/90)
In article <1990Nov6.101902.9683@canterbury.ac.nz>, phys169@canterbury.ac.nz writes: > > Not that I, for one, think I have enough computer power at > the moment, but as more people start to realise they would be paying more for > power they don't need, they will be content to stick with (say) Asian > manufacturers turning out acceptable-performance machines using yesterday's > technology. This seems to be the real limit of growth, not bus bandwidth or > the speed of light. > There seems to be a definite 'knee' in the performance/usability curve of an X server. Measured by the Xstone benchmark (and the sort of applications currently available) that knee occurs somewhere around 20K to 25K Xstones. PC based X terminal emulators have a performance around the 2 - 5K xstones, well short of the mark. What I am saying is that there is still a lot of room for performance improvements in PC products. I also have information that suggests that a number of asian manufacturers are seriously looking at mass manufacture of cheap fast X terminals. > > So, if you can't make a good Xterminal out of a PC by adding software, how > about a replacement graphics card with Ethernet & 2nd CPU on-board? That > would solve bandwidth problems and compatibility problems (given the right > hardware design). > > Mark Aitchison, Physics, University of CAnterbury, New Zealand. An excellent idea that would provide a standard graphics interface for all software running on a PC type platform, even MS-DOS programs. Unfortunately for it to work there would have to be wide acceptance of the interface standard within the PC community, and I suspect that the window of opportunity for this product is nearly closed. It would need: 1) A massed produced X interface card ie: 80960/29000/34020 + 82596 + 2/4Meg RAM + Vram + ASIC + BT459/ecl shift register on a card for < $1000 2) Software support from some notable operating systems and applications developers - ie Microsoft, Autocad, Lotus etc. 3) Inclusion as standard from some notable manufacturer, ie IBM, Compaq etc. To my mind a graphics card with X as the interface makes much more sense than the current mish mash of low performance products and standards - e.g. VGA, TMS340XX standard, 8515A etc. Since the X server would most likely be uploaded, there would also be scope for keeping up with changes in graphics standards, and the splitting of the workload would lead to better performance. - and another thing, the PC applications would then become X11 compatible; at last you could run all those nifty PC programs from your Xterminal/Workstation etc. Graeme Gill Electronic Design Engineer Labtam I.S.D. Pty Ltd
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (11/07/90)
In article <2PX6KJ@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: | I personally can't think of anything off the top of my head that a 286 | based PC offers that I want. What are you talking about here? For running | MS-DOS software, an 8088 based PC is fine. For anything else, a 68000 | based PC is much better. If you want UNIX+DOS capability the 80486 seems most cost effective. If you want just UNIX, or must have BSD, you are probably better off with a RISC machine. Until the 68040 is available the 68k family seems to be slightly down on performance relative to the 486 and RISC machines. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) The Twin Peaks Halloween costume: stark naked in a body bag
henry@zoo.toronto.edu (Henry Spencer) (11/08/90)
In article <3698@skye.ed.ac.uk> richard@aiai.UUCP (Richard Tobin) writes: >>Both [PC compatibility and X] do terrible things to your system design if you >>enshrine them as fundamental goals... > >I would be interested to hear you elaborate on this - in particular, >what does "enshrining X as a fundamental goal" involve other than >having "a sane machine with a clean frame buffer". The problem with fundamental PC compatibility is obvious: it dooms you to using a grossly obsolete CPU architecture. The problem with X is more subtle: it's so huge and slow that you end up struggling desperately to provide more and more computational resources just so X will run tolerably, at the expense of rather greater cost and complexity than would be needed in a just world. I may perhaps have been a little harsh on X in that comment. A little. -- "I don't *want* to be normal!" | Henry Spencer at U of Toronto Zoology "Not to worry." | henry@zoo.toronto.edu utzoo!henry
keith@mips.COM (Keith Garrett) (11/08/90)
In article <2PX6KJ@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: [...] >I personally can't think of anything off the top of my head that a 286 >based PC offers that I want. What are you talking about here? For running >MS-DOS software, an 8088 based PC is fine. For anything else, a 68000 >based PC is much better. if you check the requirements for the latest versions of many DOS applications, they now require 286 compatibility. -- Keith Garrett "This is *MY* opinion, OBVIOUSLY" Mips Computer Systems, 930 Arques Ave, Sunnyvale, Ca. 94086 (408) 524-8110 keith@mips.com or {ames,decwrl,prls}!mips!keith
sysmgr@KING.ENG.UMD.EDU (Doug Mohney) (11/09/90)
In article <42906@mips.mips.COM>, keith@mips.COM (Keith Garrett) writes: >if you check the requirements for the latest versions of many DOS applications, >they now require 286 compatibility. Homie don't think so! Anything which runs on a 80286 can (97.5% of the time) run on a 8088, abet slower. My roomie in particular grumbles about the slowness of MechWarrior on his paltry IBM Portable (yes, when IBM tried to clone the Compaq sewing machine..), and often covets my machine ('286) for an increase. Of course, the '286 gives you extended/expanded (pick one, I don't keep track of the difference), but if you need something THAT high powered, you should (in my humble opinion) proceed directly to Mr. '386 DX @ 20Mhz. (Yes, I s'pose you could use the '386 SX, but somehow my knee jerks at a chip which was basically designed to kick the '286 clones into the dustbin, thereby maintaining Intel's monopoly upon that market).
staff@cadlab.sublink.ORG (Alex Martelli) (11/09/90)
graeme@labtam.labtam.oz (Graeme Gill) writes: ... >would lead to better performance. - and another thing, the PC applications >would then become X11 compatible; at last you could run all those nifty >PC programs from your Xterminal/Workstation etc. According to Quarterdeck advertising (a neat brochure in a recent issue of BYTE mag), this compatibility can be yours next January with X/Desqview. A SW-only solution, so performance is in question, but... -- Alex Martelli - CAD.LAB s.p.a., v. Stalingrado 45, Bologna, Italia Email: (work:) staff@cadlab.sublink.org, (home:) alex@am.sublink.org Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; Fax: ++39 (51) 366964 (work only), Fidonet: 332/401.3 (home only).
kls30@duts.ccc.amdahl.com (Kent L Shephard) (11/09/90)
-STUFF DELETED- >> All this, of course, has a motive behind it; I've been wanting to buy some >> workstations which will run X-windows, with good speed, and yet not loose >> anything we had with 286-based PC's. > >I personally can't think of anything off the top of my head that a 286 >based PC offers that I want. What are you talking about here? For running >MS-DOS software, an 8088 based PC is fine. For anything else, a 68000 >based PC is much better. That's why more Intel based (286 & 386) machines run UNIX than any other. 68000 may be comprable but not better INMO. Not wanting a flame war but just wanted the facts straight. >-- >Peter da Silva. `-_-' >+1 713 274 5180. 'U` >peter@ferranti.com Kent. -- /* -The opinions expressed are my own, not my employers. */ /* For I can only express my own opinions. */ /* */ /* Kent L. Shephard : email - kls30@DUTS.ccc.amdahl.com */
phys169@canterbury.ac.nz (11/12/90)
In article <0093F67E.F8E473A0@KING.ENG.UMD.EDU>, sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: > Anything which runs on a 80286 can (97.5% of the time) run on a 8088, albeit > slower. Some new software (like big speedsheets and DTP-type stuff) needs a 286 or better. But the main thing is having 16-bit slots for some cards. Any software that runs well enough on an 8088 is old and/or trivial! ;-) Mark Aitchison.
brett@cayman.amd.com (Brett Stewart) (11/13/90)
In article <0093F295.10626840@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >In article <1990Nov3.052952.1786@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes: >>In article <0093F1A8.A28E4920@KING.ENG.UMD.EDU> sysmgr@KING.ENG.UMD.EDU (Doug Mohney) writes: >>>...use RISC processor and stick in Soft-PC in ROMs... Some makers of 29K-based X terminals are considering doing something similar to this. The idea is Xterm/PC/3270 based on a software switch. Only one neat, quiet box on the desktop. >> >>A small practical difficulty with this approach is that Soft PC is licensed >>software. It also does a less than wonderful job of emulation, as our >>department is finding out on a bunch of DEC RISC workstations for undergrad >>teaching... I had the good fortune to host a COMPCON session a few years ago ('89) that was quite well attended. The subject was execution of a guest architecture on a host. There were to be three papers: Luther Johnson, then of Phoenix Technology, on Phoenix's MIMIC-derived version od DOS on 'other' John Banning, then (and possibly now - are you still there John?) of Hunter Systems, on Hunter's compile-the-binaries strategy of DOS on Unix, and Insignia Solutions on SoftPC strategy. Unfortunately, Insignia was a no-show to the session, so I did not get to learn more than is in the published papers. My understanding is that the Insignia guys are switching over to the Phoenix method, and licensing some of Phoenix' technology. As for pricing, I dont know about SoftPC at retail, but the OEM pricing I have heard shouldn't make anyone blink. Interested persons can see about this technology by referring to the paper, by Cathy May, 'MIMIC: A fast System 370 simulator' or something like that, in the proceedings of an '87 SIGPLAN symposium on interpretation. This paper is more a starting point than a comprehensive how-to paper. Many customers of our AMD 29K are interested in preserving old binaries of CISC machines by using a pure-software, or hardware-assisted technique to allow them to execute on the 29K. I believe the preference for the 29K architecture for this (and hence the relevance of this posting) is due to the ability of the 29K to store a model of the 'volatile state' of the guest architecture in its capacious register file and still have enough room left over to run the MIMIC-type part of the interpretative system. On the subject of architecture neutrality, did anybody visit the OSF's demonstratiomn of 4 different Architecture-Neutral-Distribution- Format (ANDF) technology last week? 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