--------- START OF FORWARDED MESSAGE --------- From System ....: Internet From Conference : EMail MessageID ......: <9460.7031T826T2039@ci.educ.lu> Originally from : Manou BILLA Originally to ..: ahoff Subject ........: FWD: Re[2]: Re[2]: Re[4]: Gateway --------- START OF FORWARDED MESSAGE --------- From System ....: RESTENA (TCP/IP) From Conference : PIOS one Mailing List MessageID ......: <199704020635.BAA18623@zeus.jersey.net> Originally from : Dave Haynie Originally to ..: Mailinglist 'TeamONE' Subject ........: Re[2]: Re[2]: Re[4]: Gateway On Tue, 01 Apr 1997 13:15:24 -0800, Chuck Davis yammed on about: > On 01-Apr-97, Dave Haynie wrote: > >On Mon, 31 Mar 1997 13:24:44 -0800, Chuck Davis > >yammed on about: > > >> On 30-Mar-97, Dave Haynie wrote: > >Funny....he MSRP. > > Alright! OK, my apologies to you, Dave Haynie. Your design points are > well taken. I think we agree that, at the advent of the BeBox, we were > _all_ looking for something new to keep us from caving to Micro$loth :) > I browsed your website,for the first time, with this new VoyagerNG, > last evening and was able to see the pics and diagrams. Looks clean. Thanks. I'm not here claiming to be doing an A\Box. I don't know how many people Phase V have, but they need about 10 on chips, 2 on systems, to even start talking about that kind of project. AAA had more on chips, only me on systems, but we would have added 2-3 more to deliver a final system based on the AAA chips and my "post A4000" system architecture. > CHRP is great, but I missed reasoning for TransAm & Maxtreme? I'm > new to this list, so perhaps you've already done that. Hey, I missed it too. It's just plain marketing. Down inside you find the same basic hardware. There might be a few connector differences, the TransAM now has a Siamese System bundled, the Maxtreme, of course, has the MacOS (which won't be ready quite as soon as a shippable BeOS). Primarily it's just a marketing thing. > > I'm not that enamored with the MacOS, for all the known reasons. > And when I was requested to port my system to the 68K Mac, I when > searched thru all their docs for an equivalent RBF_Interrupt and didn't > find one, it just deepened my disdain for the machine. I did wish for > a 12-16 bit serial port on the Amy, though ;) The Mac 68K architecture was really, really primitive. We used to laugh at them, I took great pleasure in being able to come out with a faster machine using the same CPU, even if they gradually kicked our butts on things like RTG, and thus, graphics chip flexibility. The PowerMacs were much better, though they're kind of behind the curve these days compared to what you'd expect on a PClone, or what you'll get on most CHRP systems I suspect (though again, that's less certain since multiple companies are involved in CHRP solutions, whereas all PowerMacs and clones come from the same Apple architectures). > >It's also fairly ironic. After all, when the BeBox came out, I was one > >of the first Amiga people to support it, to get one, and to support the > >design decisions Joe Palmer made in the machine. Most of the community > >was still hung up on "custom chips". > > Yes, I appreciated that you were amongst the first to do that, and they > really attracted a mass of knowledgeable follower-developers, too. > > Perhaps, in these days of much faster CPUs, custom chips on the > motherboard might be passe'. We'll have to see how well the A\U\Box > faires in the coming paradigm. In the sense of high-integration, you're getting custom chips on the PIOS One. I mean, aside from buffers and simple TTL chips (you know, under $1.00 stuff), there are only four main chips, and a clock synthesizer. You'll find more stuff on an A500 or A1200 motherboard ( sure, they do have more of the pieces there too). It's just that PIOS didn't make these custom chips, other folks did. Intel made one, the same PCI to ISA bus bridge chip you'll find on many PClones these days. SMC made one, a Super I/O that does most of the PC ports. The Apple designed Hydra (made by several chip companies) does basic system glue, OpenPIC interrupt control (a CHRP requirement), SCSI, and the fast RS- 422 serial ports. Finally, Analog Devices made the sound chip. > Perhaps you were handcuffed by Motoringrolas not keeping their processors > up to the speeds where Insmell was headed and I wondered why was there > any `design room' left for the existence of Warp, et al, on the 3-4000? > I guess I didn't know of your bosses, at the time :( Those guys in charge really sabotages the development efforts in the latter days. We wanted something at least up to the AA specs for the A3000 graphics. 25MHz 68030 wasn't bad in 1990, that's what Apple had in their IIci, which sold for nearly twice the price of the A3000. But it also had 8-bit graphics. Management wouldn't pay the price. Next came the A4000, which was not anything like what we would have wanted. In late 1990 I started working on the "A3000+", the first system that used the AA chips. In addition, I had an AT&T DSP3210 chip in there, 16-bit audio I/O, an audio management chip, other cool stuff. The software for it was cool, the AT&T chip has its own 32-bit multitasking kernel, so one task didn't necessarily take it over, like in other DSP systems. And it did floating point about 10x faster than the '040. That's when Ali and Sydnes stepped in, kicked out the management (the folks responsible for the A500, A2000, and A3000), and installed Sydnes' folks. These guys had the primary job, it would seem, of making sure Ali knew the previous management was bad, so they did nothing for six months. They built the A600, cancelled the A500. Bonehead moves. The A3000+ was cancelled, my partner on the A3000, Greg Berlin, was told to build something we called the "A1000jr" (after Sydnes' involvement with the PCjr), a $1000 ECS machine, supposedly to go out in April of '92. No C= sales offices would order a single unit, so it was cancelled. Then in May, the A1000jr was given a partial brain transplant from the A3000+ ( just the AA stuff and Zorro III, no SCSI or DSP), the "cheapest '040 card possible without custom chips", originally made for A1000jr upgrades, became the A4000's CPU module. Started in May, it shipped in September, a record for a "new" machine at Commodore. Greg and a other guys had worked up a replacement CPU module, more like what I orginally envisioned when I did the A3000's CPU slot. One chip would hook the '040 bus to the '030 local bus, and since you had control and datapath all in one chip, a custom chip at that, whole burst cycles between the CPUs could be converted at once. So it could run nearly as fast to A3000 memory as the '030 could, rather than about 1/2 speed like the A4000 did. Plus, the 64-bit memory bus on-card would be much faster yet. Also, by virtue of the integration, the CPU didn't have to run at motherboard bus speed, so 33MHz and 40MHz '040s, or 50/66MHz '060s would work in the same module. The design was 95% completed and simulated on the CAD systems, Greg went to management, and they shot it down. This was something that would have payed for itself in TWO MONTHS of production. Plus opened up higher performance machines. But they didn't want to spend the money on a gate array. > >If so, why not just go playsomewhere else -- > > Now, isn't that just like a liberal? I didn't realize that you owned > this list :( I'm sure that we all are honored and appreciate that > PeterK and you are here. Look man, I just want to talk about computers here. I've been around the nets for nearly 19 years, I know what kind of crap can break loose when you get off topic too much, and that'll drive me away, simply because I don't have the time. I'm not afraid of a fight, but I'm only going to fight the ones I'm really interested in. Right now, that's building a new computer system (which, I have always believed, also involves a community, which is why I'm here). I have little time for other stuff -- the wife and kids, a good 3-5 mile run, a few good beers, and at least two hours of sleep a night, whether I need it or not... > I'll wait for the SMP G3 kinda thingy for my personal machine :) The first G3 is very cool. Well, literally, too, at about 5W, doing the same work a Pentium Pro or Pentium 2 does per clock tick. We do have a multiprocessor project starting after the PIOS One is in production, and we're looking to use cheaper parts than the PPC604 family. Nothing wrong with them, but we don't want to be another Daystar right now. I totally subscribe to Be's idea that regular people can benefit from multiprocessing. On the surface anyway, the first G3 chips look very good for SMP (well, in the MEI way that the PPC603 family supports, not as efficient as the PPC604's MESI, but it works). Dave Haynie | V.P. Engineering | haynie@pios.de PIOS Computer | ...no RISC, no Fun! | http://www.pios.de Be Dev #2024 | System Engineer Amiga 2000, 3000, 4000 "Taken, all I'm taking, from the old men with the power" ***************************************************************** * The TeamONE List. To unsubscribe send Email to * * listserv@nostromo.gate.net and in the body of message type * * "delete yourname@yoursite TeamONE" (minus the "") All other * * questions should be sent to damocles@nostromo.gate.net. * ***************************************************************** ---------- END OF FORWARDED MESSAGE ---------- ---------- END OF FORWARDED MESSAGE ----------