Article 125919 of comp.sys.amiga.hardware: Newsgroups: comp.sys.amiga.advocacy,comp.sys.amiga.hardware Path: news.PEAK.ORG!engr.orst.edu!osshe.edu!news.uoregon.edu!news.sol.net!newspump.sol.net!uwm.edu!news.cse.psu.edu!news.cc.swarthmore.edu!netnews.upenn.edu!dsinc!scala!news From: dave.haynie@scala.com (Dave Haynie) Subject: Re: Why the Amiga is truly dead. Sender: news@scala.scala.com (Usenet administrator) Message-ID: <1996May30.200031.8526@scala.scala.com> Date: Thu, 30 May 1996 20:00:31 GMT Lines: 135 Reply-To: dave.haynie@scala.com (Dave Haynie) References: <4nhmqn$250k@news-s01.ny.us.ibm.net> <1996May20.202434.7207@scala.scala.com> <4ns530$ba3@news-central.tiac.net> Nntp-Posting-Host: gator Organization: Scala Computer Television, US Research Center Xref: news.PEAK.ORG comp.sys.amiga.advocacy:157620 comp.sys.amiga.hardware:125919 In <4ns530$ba3@news-central.tiac.net>, dmeyer@tiac.net (David Meyer) writes: >Dave Haynie (dave.haynie@scala.com) wrote: >: In <4nhmqn$250k@news-s01.ny.us.ibm.net>, fheitka@ibm.net (Frederick Heitkamp) writes: >: >Response from Dave Haynie indicates that it is not practical for a new AGA >: >generation of chips >: Yup. The real problem is that the AA architecture, which is still >: heavily based on the original Amiga architecture, is just too old to >: be extended much further. > Sort of like the 68K architecture and the x86 architecture are >too old to be extended much further? Only worse. The x86 architecture is a good basis for comparison. The original, like Amiga chips, were based on an exposed 16-bit architecture (the 68000, by contrast, was 16-bit architecturally, but it offered a 32-bit progamming interface). When they went to the '286, the x86 architecture got a few improvements, but stayed much within the framework of the original x86 series. Much like going from ECS to AA. For the '386, they built a whole 32-bit architecture on top of the existing 16-bit architecture, so the old stuff could be supported, but new code would use the new features (or so they thought). This is very much like what the AAA architecture set out to do. The problem is that it's just more difficult to extend an Amiga graphics chip architecture this way than a CPU. Most CPUs were available from the start in several speed grades, different systems gave you different real-world performance, etc. So nothing is all that dependent on a CPU behaving just so -- in fact, the whole key to the personal computer business has been that, in most cases, going faster is a very good thing. And when you can't (the '386, for example, ran many '286 instructions a tad slower than a '286 at the same clock speed), no biggy. The Amiga chips, on the other hand, are very exactly timed, and any great change to this, any real change to blitter or copper performance, etc. and all kinds of stuff will break. Even AAA built itself around the standard 280ns bus cycle of the original Amiga chips, although the architecture itself was very different. AA worked pretty well for two reasons: it's based exactly on the ECS chip behavior. Alice is just a Fat Agnus that's learned a few new tricks. Lisa is all new, but being the display chip, she's slaved to whatever Alice feeds her. So like the x86's real mode, you boot up in compatibility mode, and unless you thrown the magic switch (in the FMODE register), you're just like an ECS chip -- blitter, display, bus bandwidth, everything (well, everything with the power of the chip designers). Going to AAA would have been expensive. And lots of stuff would have broken anyway. For example, in a VRAM system, there was absolutely no way to make the system run as slow as an ECS or AA system. That makes games and other hardware bangers execute very differently. Some don't care, most do to some extent. When you do to the 64-bit system (six chips, friends and neighbors), it's impossible to support copper lists compatibly anymore. The OS doesn't care, and neither do apps that run under the OS, but most of them don't care about S3 or Tseng SVGA chips, either. Anyone making their own screen will have big problems. Any major upgrade of the Amiga architecture will have these problems; one reason AAA was so late was the extent to which these problems were researched (well, that and Commodore's typically weak budgets). >: That's not to say that VIScorp might not benefit by merging all >: three chips and adding some features. > VIScorp will be in the business of making extremely low-end >computers. That's what set-top boxes are. My point exactly. But at least at Commodore, the AA chipset wasn't that expensive, maybe $15-$20. So you get it down to under $5.00, as a single chip, and save a few more bucks on PCB real estate and power supply. Line up a good custom process and you could suck in the DACs as well (another $3.00), maybe some additional system glue (one problem with AA being so heavily based on ECS is that Alice is basically Agnus with some of the 16-bit specifics cut out, but nothing much to help on 32-bit implementations). Get Motorola involved and you could suck in the 68020 in as well. You're still well under a 1/2 million transistors. >: Even a low-end Power Amiga needs to have fast 3D graphics, >: MPEG, 16-bit audio with sythesis, etc. or it's not going to find its >: proper niche between the game machines and the PC Clones. > It does? When I see people buying personal systems in the store, >I don't hear mention of these features by the customers. Most consumers are morons. That's why PClones and Budweiser sell so well. And yet, consumers in the gaming market are one of two kinds -- folks considerably more savvy about systems than PC buyers (of course, game machines are easier to understand), or they're kids who've been convinced by advertising or "what my friend has". A US$500 computer can be competition to the game consoles. It has a good chance of picking off the more demanding game folks who might also be lured by the idea of the net, but aren't yet really interested in computers (that's 65% of US households, 90% of European households). 3D graphics and CD-quality audio with synthesis were important enough to show up on all of the current game consoles. The Sega Saturn uses a Yamaha DSP for sample-based audio synthesis, with a 68EC000 dedicated to handling the actual MIDI procotols. Sega hasn't done as well as expected, but Sony's managed to sell around 3-4 million PSXs, and that's before they dropped to US$199. If the PSX plays a better game than your Power Amiga, you're not going to attract a ton of game players. >The customer can tell if the machine is good enough by watching and >listening. Sure. And it's going to be pretty clear if the 3D game on your computer is anemic compared to the one over there on the PSX. And just as obvious if it's blowing the PSX out of the water. Sure, software has to be there too. All I can do is answer for the hardware. >Would having these specs be better than not having them? Of course! >Is it a make-or-break requirement at the -low- end? I doubt it. Of course it's a wish list. I do believe we have a better shot with a US$500 system without PSX-beating 3D graphics than a US$1000 system with them. >You go over the $1,000 MSRP and these things start to become more >important. Not because the user is more likely to -need- them, but >because the user is more likely to -want- them. Today's necessities were always yesterday's luxeries. That 1 MIPS A1000 sure seemed cool back in 1985. Now folks are moaning when that can't have 100 MIPS. Dave Haynie | ex-Commodore Engineering | for DiskSalv 3 & Sr. Systems Engineer | Hardwired Media Company | "The Deathbed Vigil" Scala Inc., US R&D | Ki No Kawa Aikido | info@iam.com "Feeling ... Pretty ... Psyched" -R.E.M.