george@wombat.bungi.COM (George Scolaro) (11/25/90)
[In the message entitled "Re: ETH532 ???" on Nov 24, 12:47, Rick Rodman writes:] > > Here's my 2 cents: > > One place I would like to see a second CPU would be on a graphics board. > Ideally this would be a simple board with a CG16 and the RGP chipset. Les > Wilson at National can give a simple schematic for an X-terminal coprocessor > board. I would be more than happy :-D to assist with the serious (ahem) > software development. ARRRGGGHHHH!!!!! You've got to be kidding right!! The RGP and the CG16 would be the worst choice I could ever see anyone making for a graphics unit. The RGP is DEAD - no longer supported - at it's best is was a DOG and very poorly supported. Have you ever looked at the RGP performance? The cg16 is nearly a dog also, a 68k is similar for most things and even faster on data manipulation at the same clock speed. The address calculation time kills the cg16 - nice architecture but not enough silicon in the right places. It is 1980 technology, a better choice would be one of the newer riscs etc, even the new Intel 960SA. If we are playing with hardware for hobby/instruction which is what the pc532 is meant to be about, then let's try and use silicon that is at least nearly state of the art. HA, a simple X-terminal with the cg16 - tell me about it. Just take the schematic and replace the cpu with any other - it will be just as 'simple'. The cg16 has some bitblt stuff etc but this doesn't make it a graphics chip. Again a 960SA and a C&T VGA chip etc, would be a better choice in terms of simplicity in the hardware. Well, lots of things to do, but currently no time to do it. We need more hardware folk out there to take a design through to completion and fab boards. How about it? Right, got that off my chest! Now that I'm awake I can get back to some software (gad's the programmer's union will be after me) I'm doing... Regarding the complexity of the et532, I agree it has a lot of stuff on it - but then it was designed for a purpose which wasn't to be cheap and low end etc. If someone does a simple ethernet board I would suggest using something that can move data to/from the SCSI at a reasonable rate - at least 3Mbytes/sec. Anyhow best regards, -- George Scolaro george@wombat.bungi.com [37 20 51 N / 122 03 07 W]
news@daver.bungi.com (11/27/90)
George: > > One place I would like to see a second CPU would be on a graphics board. > > Ideally this would be a simple board with a CG16 and the RGP chipset. Les > > Wilson at National can give a simple schematic for an X-terminal coprocessor > > board. I would be more than happy :-D to assist with the serious (ahem) > > software development. > > ARRRGGGHHHH!!!!! You've got to be kidding right!! The RGP and the CG16 would > be the worst choice I could ever see anyone making for a graphics unit. The > RGP is DEAD - no longer supported - at it's best is was a DOG and very > poorly supported. Have you ever looked at the RGP performance? The cg16 is > nearly a dog also, a 68k is similar for most things and even faster on data > manipulation at the same clock speed. The address calculation time kills the > cg16 - nice architecture but not enough silicon in the right places. It is > 1980 technology, a better choice would be one of the newer riscs etc, even > the new Intel 960SA. If we are playing with hardware for hobby/instruction > which is what the pc532 is meant to be about, then let's try and use silicon > that is at least nearly state of the art. I disagree with using an Intel anything, and I think using a compatible processor on the graphics card would simplify things. You're a hardware person, I'm a software person. How about a TI34020, then? > HA, a simple X-terminal with the cg16 - tell me about it. Just take the > schematic and replace the cpu with any other - it will be just as 'simple'. > The cg16 has some bitblt stuff etc but this doesn't make it a graphics > chip. Again a 960SA and a C&T VGA chip etc, would be a better choice in > terms of simplicity in the hardware. VGA, eh? I doubt that making it PC-compatible would make the software any easier to write. PCs have no intelligence in their graphics. Heck, in terms of simplicity of hardware, dual-ported RAM and a 6845 would be easier. :-{ In fact I can picture a 8751 doing the SCSI side. :-/ (Feel like another healthy arrrggghhhh?) > Well, lots of things to do, but currently no time to do it. We need more > hardware folk out there to take a design through to completion and fab > boards. How about it? > > Right, got that off my chest! Now that I'm awake I can get back to some > software (gad's the programmer's union will be after me) I'm doing... > > Regarding the complexity of the et532, I agree it has a lot of stuff on it - > but then it was designed for a purpose which wasn't to be cheap and low end > etc. If someone does a simple ethernet board I would suggest using something > that can move data to/from the SCSI at a reasonable rate - at least > 3Mbytes/sec. Agree there. I just don't know about all that hardware which needs a fairly complex software program to control... or the cost. I'm an experimenter on my own budget out here.
dlr@daver.bungi.com (Dave Rand) (11/27/90)
[In the message entitled "RE: Et532" on Nov 26, 21:26, Rick Rodman writes:] > I disagree with using an Intel anything, and I think using a compatible > processor on the graphics card would simplify things. You're a hardware > person, I'm a software person. How about a TI34020, then? The Intel 960 is quite a nice part. Fast, easy to code, good tools available, and a range of processor speeds to choose from. It is also very very cheap. (I'm also a software person, or at least I think I am :-) > > chip. Again a 960SA and a C&T VGA chip etc, would be a better choice in > > terms of simplicity in the hardware. > > VGA, eh? I doubt that making it PC-compatible would make the software any > easier to write. PCs have no intelligence in their graphics. > Heck, in terms of simplicity of hardware, dual-ported RAM and a 6845 would > be easier. :-{ In fact I can picture a 8751 doing the SCSI side. :-/ > (Feel like another healthy arrrggghhhh?) Yes, VGA. George pointed out that in terms of simplicity in the hardware, a VGA chip (like the C&T 452) and the 960SA would be a nice match. PC compatibility has little to do with it. Dual ported ram and a 6845 does not make a 1024x768x256 color system... Even 'simple' frame buffer designs are non-trivial without some LSI to do some of the work for us. Also, remember that this card will reside on the SCSI bus. The main processor will not have the video memory in its adddress space, so we want something significantly more powerful than the 8751 (another Intel chip :-) doing the rendering for us. The 960 is a great cost/performance part. > > -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
news@daver.bungi.com (11/28/90)
>[In the message entitled "RE: Et532" on Nov 26, 21:26, Rick Rodman writes:] >> I disagree with using an Intel anything, and I think using a compatible >> processor on the graphics card would simplify things. It might very well do so. Read on... >Also, remember that this card will reside o>n> the SCSI bus. The main >processor will not have the video memory in its adddress space ... Not having the frame buffer accessible to the main CPU may complicate things like save unders if you do an X-windows port. Jonathan Ryshpan <...!uunet!hitachi!jon> M/S 420 (415) 244-7369 Hitachi America Ltd. 2000 Sierra Pt. Pkwy. Brisbane CA 94005-1819
dlr@daver.bungi.com (Dave Rand) (11/28/90)
[In the message entitled "Re: Et532" on Nov 27, 11:30, Jonathan Ryshpan writes:] > >Also, remember that this card will reside on the SCSI bus. The main > >processor will not have the video memory in its adddress space ... > > Not having the frame buffer accessible to the main CPU may complicate > things like save unders if you do an X-windows port. By main CPU, I am speaking of the 532 on the PC532 motherboard. Of course, the local CPU will have access to the frame buffer. -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
loeliger@mozart.convex.com (Jon Loeliger) (04/22/91)
Perhaps now that so many boards are up an running, people are more interested in newer boards like the ET532. I know of 3 that are NOT (Dave or George): Randy Hyde, Rick Rodman and myself. If all those who are interested would send me a note at loeliger@convex.com, I'll summarize and post the results to the list. Give me a confidence rate or a "I'll buy one (the board) for $300 but NFW will I get one for $350!" or some such willingness threshold. Many moons ago, I even offered to help kit them if no one else was willing to do it... (That's a backhanded offer to help kit them now if still no one else is more willing than I...) jdl
rhyde@ucrmath.ucr.edu (randy hyde) (04/23/91)
Actually, if George or Dave would summarize the features of the ETH532 it might help some people decide one way or the other. I suspect many may have forgotten what it offers or may not even be aware of it. As I seem to recall its a 32gsx32 (?) processor (a 32532 w/o MMU) 4 megs RAM, 16 serial ports and an ethernet chip. Am I missing anything? Was the clock speed 20Mhz? *** Randy Hyde
george@wombat.bungi.COM (George Scolaro) (04/25/91)
[In the message entitled "Re: ET532" on Apr 23, 6:07, randy hyde writes:] > > Actually, if George or Dave would summarize the features of the ETH532 it > might help some people decide one way or the other. I suspect many may The ET532 - Ethernet 532 is a plug in card for the PC532. It connects via the expansion SCSI bus. It also has a 50 pin scsi header and power supply connector for stand alone use (i.e. via a different SCSI based machine). 32GX32/32532-20MHz 1 or 4 Mbytes of DRAM (via fast page 256Kx4 or 1Mx4 dips) 1 Boot EPROM 1 1024bit EEPROM (for ethernet address etc) It has 16 serial ports (rs232 buffers are on a daughter card). 1 DP8490 SCSI interface 1 8 bit expansion connector (for other stuff) 1 Ethernet interface using the NSC DP8390 etc chipset. Supports thick ethernet and cheapernet (thin coax). The board is designed and routed. It has never been built. All that is required is a reasonable number of people to want it (reduce the PCB NRE etc) and it is ready to fab. The design is based upon the PC532 and an ethernet design I've already done - so even though it would be a prototype run I have very high confidence in the design. Maybe a cut and jump or two though... best regards, -- George Scolaro george@wombat.bungi.com [37 20 51 N / 122 03 07 W]
nomad@watson.ibm.com (Lee 'nomad' Damon) (04/26/91)
In article <9104241611.AA27803@wombat.bungi.COM> george@wombat.bungi.COM (George Scolaro) writes: >The [eth532] is designed and routed. It has never been built. All that is >required is a reasonable number of people to want it (reduce the PCB NRE etc) Any idea how many people this is? nomad work: nomad@watson.ibm.com - Lee "nomad" Damon - \ play: castle!nomad or nomad@castle.org \ Un*x Systems Administrator, Math Dept, IBM TJ Watson Research Center / \ "Blessed are the pessimists, for they have made backups." / \
maniac@convex.convex.com (04/26/91)
George Scolaro writes: [description of the ETH532 deleted...] > The board is designed and routed. It has never been built. All that is > required is a reasonable number of people to want it (reduce the PCB NRE etc) > and it is ready to fab. The design is based upon the PC532 and an ethernet > design I've already done - so even though it would be a prototype run I have > very high confidence in the design. Maybe a cut and jump or two though... Sounds good, did the original pc532 board come out this good? (i.e. did my board from the second round have any changes on it, or was it identical to everyone elses?) Anyway, count me in for a board. I'm not sure I'll have the money for a month or two, but I should be able to come up with two or three hundred bucks for one. -- Jon Buller jonb@vector.dallas.tx.us ..!texsun!vector!jonb FROM Fortune IMPORT Quote; FROM Lawyers IMPORT Disclaimer;
george@wombat.bungi.COM (George Scolaro) (04/26/91)
[In the message entitled "Re: ET532" on Apr 25, 21:22, texsun!vector.dallas.tx.us!maniac@convex.convex.com () writes:] > > George Scolaro writes: > > The board is designed and routed. It has never been built. All that is > > Sounds good, did the original pc532 board come out this good? (i.e. did > my board from the second round have any changes on it, or was it identical > to everyone elses?) All boards that have ever been shipped are identical and no PAL changes have even been made (only the PCB house date codes changed). In fact the original prototype only had a couple of wires - and one of those was due to misinformation from an NS person regarding a subtle timing parameter - pays to only design based on the data sheet!! Of the last half a dozen boards I've done the reworks required have been either a wire or two, pal changes or perfect - so I'm banking on a major scewup anytime now :-) best regards, -- George Scolaro george@wombat.bungi.com [37 20 51 N / 122 03 07 W]
dlr@daver.bungi.com (Dave Rand) (04/28/91)
[In the message entitled "Re: ET532" on Apr 27, 20:42, Rick Rodman writes:] > The $550 SCSI/Ethernet adapters sound very expensive for what little > hardware. Perhaps you misunderstand. The *BARE BOARD* will be in this price range, for the first few units. The way things work is that the first run of boards (which we do as few as possible, in case of major failure) cost the most. The reason for this is the cost of photoplotting (at about $30 per sheet, 9 sheets required, plus negatives); the cost of tooling from the PC board house (about $350-400); the cost of digitizing the hole positions (another $200 or so); the cost of setting up the PCB test fixture (about $300-500). These are fixed, one-time (hopefully) costs - but they add up. The boards themselves will be 'respectable' in price, perhaps as low as $230 each. But the initial costs must be borne by someone... The first run of PCB's that George and I did cost over $2000 (if you include the first aborted photoplot job - not fun). This was why the first run of 'real' boards were only $200 (that, and the fact that we had enough orders to amortize the costs better, as we had to re-plot the board). It is for this reason that we want to have maximal functionality on each board that we do - the minimum hardware required to get the job done, but the maximum functionality. It doesn't really matter what the board size is, or how many holes are on it. Simple fact is that a small run of boards will cost $2000-2500 for 5-6 boards. > > myself. If all those who are interested would send me a note > > at loeliger@convex.com, I'll summarize and post the results > > to the list. Give me a confidence rate or a "I'll buy one > > (the board) for $300 but NFW will I get one for $350!" or some > > such willingness threshold. > As I see it, there is little likelihood of a complete ET532 being > constructed for under $800. That's a lot of money for an ethernet > board. (And why does it need all those serial ports and memory?) > If the hardware cost can be kept below $500 by not populating the > serial ports, using a PPGA processor group-purchased, etc., it would > be more attractive. But I wonder what the design criteria for the > board were. It doesn't matter what people will pay for the board in the long run, it is what it costs to get it done. Those costs have to be met, and we (George and Dave) can't prototype every board out of our pockets. Yes, it is a lot of money for an ethernet board (although I would be happy to sell you my Excelan EXOS-205T board that I paid $1200 for, just to get my PC on the ethernet!). It is not a lot of money for a 10 MIP processor with heaps of memory, a SCSI port, 16 serial channels, and an Ethernet. This board can be used with the PC532, or as a standalone system attached directly to a hard disk and console. It can also be used as a parallel processor with the 532 - up to 4 of them per system. It can also be used as a SLIP-to-ethernet packet router with up to 16 serial ports. It can also be used as a vegtable peeler (It slices! It dices! It holds your door open!). Thinking about things is easy; doing them is harder. Getting multilayer high quality boards done is one of the harder things to do today. You just can't do them in your bathtub, and cheap board houses cost MUCH more in the long run than they save you in the short run. -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
wilker@gauss.math.purdue.edu (Clarence Wilkerson) (04/29/91)
What would it take to make the ET532 use 4meg chips, so that the memory capacity could be 16 megs? I was thinking along the thread of Dave Rand's last post that it could be though of as 10 mips general processor. Clarence Wilkerson
george@wombat.bungi.COM (George Scolaro) (04/29/91)
[In the message entitled "Re: ET532" on Apr 28, 19:35, Clarence Wilkerson writes:] > > What would it take to make the ET532 use 4meg chips, so that > the memory capacity could be 16 megs? I was thinking along the > thread of Dave Rand's last post that it could be though of as > 10 mips general processor. > Clarence Wilkerson > Moving the jumper that selects 256Kx4 devices to the 1Mx4 position. best regards, -- George Scolaro george@wombat.bungi.com [37 20 51 N / 122 03 07 W]
wilker@descartes.math.purdue.edu (Clarence Wilkerson) (04/30/91)
So the ET532 would be slightly slower, not AT form factor, no FPU, no hardware virtual memory support version of pc532, with the advantage of ethernet? ( Viewed as a general purpose computer instead of peripheral processor ). Clarence Wilkerson .
kls@ditka.Chicago.COM (Karl Swartz) (04/30/91)
On Mon, 22 Apr 91, Jon Loeliger <loeliger@mozart.convex.com> writes: > Perhaps now that so many boards are up an running, people are > more interested in newer boards like the ET532. One problem as I see it is ... software! The ET532 may do all sorts of great things but it won't do 'em without some software to drive it. Yes, there is the kernel that runs on the board though I'm thinking more of all the networking code. Does Minix have it? I've heard rumors of somebody working on Minix TCP/IP but I'm not sure it's out there yet, and it certainly isn't on the pc532 version. Unix can do it but at this point that only helps a few Finns. (No offense.) Unlike the CPU, most of this work can be done in advance since all but the actual device drivers are hardware-independent. You don't even have to stick with local loopback tests -- use SLIP or PPP on some of those 8 serial ports! On Sat, 27 Apr 91, Rick Rodman <rickr@virtech.uucp> writes: > The $550 SCSI/Ethernet adapters sound very expensive for what little > hardware. And Dave Rand <dlr@daver.bungi.com> replies: > The *BARE BOARD* will be in this price range, for the first few > units. The ... first run of boards ... cost the most. Dave, you're saying the first few Guinea pigs eat the fixed cost for all the follow-ons? I'd be willing to invest a larger chunk up front if I'd get a piece back on the first production run to end up paying my fair share of the thing. But paying a big chunk up front so someone else can get a free ride on the production boards? Doesn't sound like a very good deal to me. > Simple fact is that a small run of boards will cost $2000-2500 for > 5-6 boards. Or $333-500 per board. And the marginal cost for additional boards is on the order of $250 each (more than the pc532 because of the gold- plated edge connectors). So if the first production run is 20 boards, charge $270-300 and rebate the difference to the few "investors" who funded the initial board run. > It can also be used as a SLIP-to-ethernet packet router with up to > 16 serial ports. The value of a co-processor to unload much of the TCP/IP protocol processing is not to be underestimated. And for driving serial ports, running SLIP, PPP, or just as plain old terminal or modem connections. To me, at least, the design is certainly appealing for this reason in spite of what seems like quite a bit of hardware for "just an Ethernet board." One question, however. As I recall, those serial ports have minimal or no modem control. If true, their use for a router would be rather diminished since one would expect most SLIP or PPP to be over modems, especially if there's an Ethernet to be had. Right? Bottom line, for me, is that I'm interested in an et532, but not until I see some networking software that will be able to grow into it, and I'd only be willing to participate in the prototype run if I thought there was going to be an equitable sharing of fixed costs with the big production run. -- Karl Swartz |INet kls@ditka.chicago.com 1-408/223-1308 |UUCP {uunet,decwrl}!daver!ditka!kls |Snail 1738 Deer Creek Ct., San Jose CA 95148 "It's psychosomatic. You need a lobotomy. I'll get a saw." (Calvin)
dlr@daver.bungi.com (Dave Rand) (04/30/91)
[In the message entitled "Re: ET532" on Apr 29, 12:07, Karl Swartz writes:] > > One question, however. As I recall, those serial ports have minimal > or no modem control. If true, their use for a router would be rather > diminished since one would expect most SLIP or PPP to be over modems, > especially if there's an Ethernet to be had. Right? > No, all the serial ports have full modem control. It was a design requirement... -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com
david@marvin.jpl.oz.au (05/01/91)
Yet another name for the interest list. However, I would really be more interested in a simpler ETHERNET interface, which is the part that really grabs me. The machine doesn't really need another "brain", and judging by most workstations, they dont usually put on extra grunt to serve the e'net. David (Still to get our's running) .......................................................... David Magnay mail david@marvin.jpl.oz Boral Elevators was: Johns Perry Lifts 45 Wangara Road, Cheltenham 3192 phone (03) 584-3311 Victoria, Australia O/seas +61 3 584 3311
jvh@niksula.hut.fi (Johannes Helander) (05/02/91)
In article <m0jXdZE-000029C@ditka.Chicago.COM> kls@ditka.Chicago.COM (Karl Swartz) writes: > there yet, and it certainly isn't on the pc532 version. Unix can do > it but at this point that only helps a few Finns. (No offense.) We hope to have the pc532 Mach 3.0 kernel included in the CMU free distribution RSN. The ux server is unfortunately not free. If you have the Mach 3.0 ux server for the i386, we'll be able to send you diffs. To get those sources you first need a un*x source license. Then CMU may be able to help you with that. The GNU servers will probably be available in a year or so... (?). Anyways, to be able to run Mach, a MMU is required. Thus a gx32 is not sufficient. One possibility would be to put a 532 on the etherboard. But then it's going to be an "almost pc532" with ethernet. It thus comes to my mind that instead of building the ET532 a possibilty would be to make a second revision of the pc532 and put the ether chip there. This would solve the problem for future pc532 buyers and those of us who want to get an ethernet card could buy a second pc532. If the ET532 is going to be as expensive so what is the difference (except some design work)? To try to solve the cheap and simple ethernet problem, Jukka Virtanen designed a daughter board board that would be plugged in the PROM socket and one PAL socket. It is a four layer board. The National ether chip was not used because it seems to be very hard to get. We then found out how expensive it is to manufacture even four layer boards so the board has not been built. Maybe somebody (George?) could take a look at that design and evaluate it. This might anyhow be one possible solution for the ethernet issue. Johannes
bdale@col.hp.com (Bdale Garbee) (05/02/91)
> As I see it, there is little likelihood of a complete ET532 being > constructed for under $800. That's a lot of money for an ethernet > board. (And why does it need all those serial ports and memory?) The board was not originally intended, as I understand it, to be primarily an Ethernet board. Dave and George originally designed the pc532 to solve their own problem of needing a bleepingly large number of serial ports that could handle real speed without dropping bits. I've mumbled from time to time that the world needs a card that has a simple but effective processor core with SCSI to build neat widgets around, like an Ether interface, a floppy controller, maybe even an SMD interface for the 3 Fuji Eagles sitting on my basement floor (anyone want to buy one or more? :-). I have a design partially on paper for a V40 core that would work well, and I now have a design partially on paper for a 68000 core that would work well. It will be some time before I can spare enough personal cycles to get either one to the point of a complete design and/or proto. If someone else is interested in the "build a simple but capable core processor/SCSI to use for neat stuff" idea, I am more than willing to talk about my ideas, and/or work with someone else who has more time to spare... I'm swamped at my day job, and working on a small piece of contract work, and trying to get the tower up in the back yard, and... you get the picture. I've got access to excellant software tools for both processors. Some I can share, some I can't. The V40 is an Intel 80x86 type part with lots of onboard I/O, the 68000 core I envision would have fewer wizzies, but be easier to write code for and have the potential for lots more memory... both processors are pretty cheap. > But I wonder what the design criteria for the board were. Something very different than what I personally think most pc532 owners want. But, the design is done, George does totally awesome work, and if the group decides to build et532's, I'll try to find the cash to play too... Bdale
wilker@descartes.math.purdue.edu (Clarence Wilkerson) (05/03/91)
Well, I agree that another version of pc532 with ethernet might entice new 32'ers beyond the present base. My idea would be to use my present sun 3/50 as an x-terminal connected to such a board.
dlr@daver.bungi.com (Dave Rand) (05/04/91)
[In the message entitled "Re: ET532" on May 2, 13:28, Johannes Helander writes:] > Anyways, to be able to run Mach, a MMU is required. Thus a gx32 is not > sufficient. One possibility would be to put a 532 on the etherboard. > But then it's going to be an "almost pc532" with ethernet. It thus > comes to my mind that instead of building the ET532 a possibilty would > be to make a second revision of the pc532 and put the ether chip > there. This would solve the problem for future pc532 buyers and those > of us who want to get an ethernet card could buy a second pc532. If > the ET532 is going to be as expensive so what is the difference > (except some design work)? There is no difference. If you would prefer to run a 532 in the ET532 design, all you have to do is remove the GX, and install the 532. No cuts, no jumpers. The GX and 532 are pin compatible. The ET532 IS a 'pc532 with ethernet'. It doesn't have the second SCSI chip, nor the FPU, but it does have Ethernet and enough serial ports to keep even me happy! -- Dave Rand {pyramid|mips|bct|vsi1}!daver!dlr Internet: dlr@daver.bungi.com