Xref: utzoo comp.sys.apple:23784 comp.sys.mac:50995 comp.sys.apple2:303 Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uunet!cbmvax!daveh From: daveh@cbmvax.commodore.com (Dave Haynie) Newsgroups: comp.sys.apple,comp.sys.mac,comp.sys.apple2 Subject: Re: Accepting the Mac (was Re: More Macweek Rumors) Message-ID: <10278@cbmvax.commodore.com> Date: 20 Mar 90 22:01:13 GMT References: <1848@crash.cts.com> <18491@boulder.Colorado.EDU> <12667@thorin.cs.unc.edu> <1990Mar17.105403.17776@spectre.ccsf.caltech.edu> <10254@cbmvax.commodore.com> <1990Mar20.153543.5841@spectre.ccsf.caltech.edu> Reply-To: daveh@cbmvax (Dave Haynie) Organization: Commodore, West Chester, PA Lines: 212 In article <1990Mar20.153543.5841@spectre.ccsf.caltech.edu> toddpw@tybalt.caltech.edu (Todd P. Whitesel) writes: >daveh@cbmvax.commodore.com (Dave Haynie) writes: >[ in response to stuff I wrote ] >>Yeah, right. You're going to grow very old waiting for 65816s that can compete >>with the latest 680x0s. >Why should I wait for a CPU that does more than I need when I will be able to >buy a 20 mhz 65816 in a month or two? I don't do anything that requires the >raw power of a 68030/040, and I'm not going to pay for one because I'd be >wasting my money on features I won't use. My point being that, at any level of performance, the 65xxx solution these days costs more than then 680x0 solution, for general purpose computing. We still use 65xx-family parts where they make sense. Commodore's 7 port serial card is managed by a slow (3.5MHz) 4502. But 4502s cost pennies, all serial ports deal with are interrupts and byte pumping. Real computers with GUIs need more performance. Which is exactly the reason you're concerned with 65816s. >Most of what I do are jobs best done by coprocessors Specific purpose jobs are done well by coprocessors, as long as the coprocessor can do the job better, or at the same time the CPU is doing something else. But if the CPU can't run in parallel and could do the job nearly as well as the special purpose unit, than that extra hardware is a waste of money. Just to keep things in prespective. >>It really doesn't matter if WDC eventually comes out with a 20MHz part >They might or might not. But ASIC Technologies (two college students) reverse >engineered a 65816 state machine. On rented supercomputer time they silicon >compiled it onto a gate array and the prototypes are due in silicon Real Soon >Now. We'll see. Most chip design isn't generally done on supercomputers, and especially something as simple as a 65816 core. The CSG 4502 I mentioned was a redesigned CMOS version of the 6502 that brings in all the Rockwell instructions and a mess of new ones, and goes up to 10MHz or so. The redesign was pretty simple at the gate level, as a gate array. Also, you don't use silicon compilers on gate arrays -- gate arrays and standard cell chips are automatically routed from a gate description. Pretty standard stuff, anyone can bop on over to LSI Logic or Motorola, plunk down about $10K NRE, and get their own gate array made. The first Sparc chip was done this way, that's just a tad more complex than any 65xxx you might dream up. But if you have any hope for the future of the 65xx family, getting away from WDC is a good idea. Far as I know, WDC never made production quantities on their own -- like I mentioned, GTE was the first to make any. But everyone did count on WDC and Mr. Bill for new designs. Throwing a few new ideas at the thing, cache, a wider data bus, etc. would keep it competitive for another few years at the low end. But more than likely give software compatibility the big one-two. Most 65xx code is extremely bad and intolerant, in modern terms. >They shafted him from them start, anyway; Apple requires two >second sources and "won't buy from design companies" so after WDC licensed >VLSI technologies and California Micro Devices, Apple quit ordering from WDC. When the GS came out, the only company making '816s was GTE. The main reason I couldn't get any 4MHz '816s in quantity was that Apple bought them all. They could make a real deal on 2MHz parts, since the yield on 4MHz was so low, they had more of those than they knew what to do with. That was back in '85, I haven't kept up-to-date on who's cranking these babies out now. Back then, at least, WDC didn't have a production volume foundary anyway. And "buying from design companies" certainly didn't scare Apple away from the 68030, only available from Motorola. Though I do believe that no one would have trusted any production item on WDC as a single source had they been the one and only '816 vendor. >Like I said, if they CPU isn't what's being used for everything, then who >cares, especially if price is a major concern? Most people are going to be >using the graphics performance which should be taken care of by a cheap >blitter and NOT the CPU. Even a rather complex blitter like the Amiga's requires a good CPU behind it. The blitter does a good bit of work for you, but there are many graphics issues that are handled much better with the CPU. In general, they work together. >Hold it, you're talking like a power user, and not like your typical Apple II >customer. Does it really show? True, the Apple II is plenty fast enough for lots of people. But here you are proposing a more extensive implementation of that architecture, to go much faster. Yet even the current GS isn't on a par with something like the A500, speed-wise. It doesn't seem very likely that you can do all that much to build a faster, more powerful GS for any less money than the current one (while, of course, keeping up against any desire on Commodore's part to build a faster, more powerful A500 for less money than the current one). The GS is in what I would call a techological bottleneck. Same place the C64 is. My suggestion for the C64's future was to put the whole thing on a single chip or so and sell it in cardboard/shrink wrap packages in the checkout lines at K-Mart for $25.00. Or something like that. Same principle applies to the Apple II, if Apple even wants it to stay around. >>Same reason you aren't seeing serious competition from fast 8088s anymore. >Serious competition? Do you care about low end price/performance or just wicked >fast machines? Personally? I think speed is the ONLY thing that matters. That's what they pay me for. But seriously, the 8088 machines are a good example -- they're getting cheaper, but not faster. You brought up the speed issue, talking about killing A500s with a faster Apple GS. >>Thing is, it's going to cost more to build your 20MHz Apple //f, significantly >>more, that the el-cheapo color Mac. >I really doubt that, after seeing what Apple can do when they acutally pour >a little money into the Apple //. Right now the Low Cost Mac sounds like it >will be about as powerful as a Transwarped GS but will run 'professional' >Mac programs if you force it to, and for about the same price. Apple must, of course, decide to build a low-cost color Mac. But they're already well on the way to it -- the video controller in the Mac IIci, or something like it, is the first step. This uses cheap memories. Apple is capable of doing a cheap Mac. They might have to bend the rules. The GS doesn't have the traditional Apple 5x cost to retail markup like the Macs do; the cheap Mac wouldn't either. >> Way back in '85, a 4MHz '816 cost noticably more than an 8MHz 68000. >> Things are going to be even more skewed now. >Not really; Bill Mensch won't be doing the production anymore. It doesn't matter. There are all kinds of companies making the basic 68000, and this has driven the price way, way down, at least in volume. You can't get a small-to-medium gate array from LSI for less than a 68000. >That's because they've been waiting for a faster 65816 and Bill Mensch >hasn't been able to give them one. If Apple really wanted a faster Apple II, they could have banged out a clone in about a year, at the rate they're cranking out new gate arrays. Sure, it's easier for them if Mensch does the work, but it also could wind up costing much more if they looked at this as a real product with a chance at making them some money. >Marketing has gotten it into their heads that the Mac is the only computer >Apple sells and so they've been trying to strangle the Apple // without >realizing what it's costing them in customers or customer support. Nobody feels good if their computer is apparently abandoned by the parent company. But you can't expect altruism from Apple or anyone else either (nice if you get it, but don't ever expect it). If the IIGS isn't making them enough to justify the II's continued existence, at least Apple's bean counters will want to dump it. >The IIGS system software is also a LOT cleaner than the Mac's because no new >CPUs have forced it to be patched all over yet. The 680x0 series was designed pretty much from the start to allow painless updates. At least to a certain extent, Apple didn't follow the guidelines necessary to follow through with that. I really don't know how badly that affects Mac stuff today. Updates SHOULD have been completely transparent. >Besides, most of us buy for reasons OTHER THAN THE BLOODY CPU, so please don't >assume that everyone wants a 68K based machine just because it is a 'better' >CPU. It is also much more expensive and I for one am happy paying less for less >because the 65816 suits me just fine. As I've pointed out several times, the 68000 costs less than the 65816. >> The current //gs already costs more than an equivalently set up Amiga 500. >The 'current //gs' uses five year old gate arrays and a ten year old >architectural structure but with major beef-ups. The current A500 is using four custom chips designed in '85 or earlier, and two finished in 1986. Heck, even the 68000 has been around much longer than the 65816. New architecture can sometimes lower a system's cost, but you have to pay for that new architecture. By the time you've been making a part, any part, for 5 or 10 years, if you're paying for more than plastic, metal, and sand, you're paying too much. >By doing to the //gs what Apple did to the Mac IIfx (and don't try to tell me >that designing the system intelligently costs that much money) they will save >gobs over the current design even after adding the bare essential new features. It costs quite a bit of money for Apple to build the new stuff in the IIfx, no doubt. In some respects, the very successful Mac II line may have already funded that R&D. Apple makes lots of money on Mac IIs, they have a very high markup. Only Apple knows for sure, but I doubt they could fund a major Apple II revamp project with the current profits from the Apple IIs they still sell. >Do not assume that the IIGS is the best that the Apple II can do. I've never assumed that about any Apple II machine -- they're always had too many parts. But Apple only started getting serious about gate arrays near the end of the Apple II's current history, and applied the best of that new (to them) ASIC capability to the Mac II. You can always do better, like the "C64 on a chip" I joked about up above (though it has moved that way over time -- today's C64 board is less than 1/2 the size of the original). Somebody has to justify that "doing better". Back before the A500 was done, the C128 group I was working with wanted a similar cost reduction and feature improvement, to build a new C256 or some-such. No matter what we came up with, the PC board was only a small part of the cost for either machine. Casework, power supplies, keyboard, floppy disk drives, etc. are essentially the same for both. It doesn't take a marketing wizard to figure out that "16/32 bit, 7MHz, Blitter, Copper, 880K, 68000" are easier to sell than "8 bit, 2 or 4MHz, 360K, 65/85/45-something-or-other" at the same price. >Todd Whitesel >toddpw @ tybalt.caltech.edu -- Dave Haynie Commodore-Amiga (Systems Engineering) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy Too much of everything is just enough