Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!seismo!amdahl!kim From: kim@amdahl.UUCP (Kim DeVaughn) Newsgroups: comp.sys.amiga Subject: Re: The Next Amiga Message-ID: <6369@amdahl.UUCP> Date: Sun, 26-Apr-87 07:21:52 EDT Article-I.D.: amdahl.6369 Posted: Sun Apr 26 07:21:52 1987 Date-Received: Sun, 26-Apr-87 23:17:10 EDT References: <3367@udenva.UUCP> <6248@ukmj.ukma.ms.uky.csnet> <1734@cbmvax.cbmvax.cbm.UUCP> Organization: Amdahl Corporation, Sunnyvale, CA 94086 Lines: 96 Yes, I know this probably doesn't "belong" in this news.group, so you may want to hit your "n" key now ... In article <1734@cbmvax.cbmvax.cbm.UUCP>, grr@cbmvax.cbm.UUCP (George Robbins) writes: > >In article <6151@amdahl.UUCP> kim@amdahl.UUCP (Kim DeVaughn) writes: > > > >Or how about using a MIPS R2000 chip (12.5 MHz flavor), and is available > >running 4.x BSD ... > > In all fairness, it seems that the MIPS chip requires more than a simple > "chip" to obtain it's nominal performace level. Well George, my posting was a somewhat "tongue-in-cheek" response to someone at AMD who suggested using their Am29000 chip in a future Amiga ... guess I should have sprinkled in a few :-) :-). While the Am29000 looks like a very nice chip according to the Preliminary specs that have been published, it has yet to be realized in silicon. I will reserve judgement on it until I have seen a real part, and have some idea as to availability, pricing, etc. You are quite correct in saying that the R2000 requires more than a simple chip to obtain it's nominal performance. While the R2000 can be made to work with a few PALs, some reset and clocking circuitry, a few bus drivers and receivers, and a little glue, this will not provide "nominal" performance (though it will still give "reasonable" performance). Nominal performance requires the addition of I-fetch and Op caches, and a faster interface to memory than most standard buses (VME, MultiBus II, etc.) can support. (The MIPS board and system level products use the VME bus, but MIPS can supply a proprietary dual-ported memory board with a fast, private interface to the CPU, in addition to the VME interface.) Another performance addition are gate-array chips that can buffer "writes" to memory (in favor of "fetches"). And for any serious floating-point work, you'd want the FPU coprocessor. In any case, this is all somewhat moot, as the R2000 (and the Am29000, and the Clipper) is really an engine for a high-performance product running UNIX(R), with lots of DASD, and several users. For a high-end Amiga class product, an 020 or 030 makes far more sense! > The Clipper seems to > provide a much higher level of integration, including cache and MMU in a > basic 3 chip set. Not so! The single R2000 chip has all of the above on it except for the cache RAM itself ... 64-entry fully associative TLB, cache-control, etc. Why waste valuable *custom* silicon on RAM-chip equivalents, when RAM chips are widely available, and cheap? Plus, you get to design the size of the caches you want/need. You should also be aware that the Clipper, in it's *initial* incarnation, is already pushing existing technology at the high end, which makes higher performance versions of the chip (set) less likely, and definitely harder to do. For example, the current Clipper achieves it's rated performance at 33 MHz; the MIPS R2000 does it's thing at a measly 12.5 MHz. Do not be surprised if you see a 25 MHz R2000 out considerably sooner than you'll see a 66 MHz Clipper. Also, board design tends to get just a little "hairy" at 50+ MHz :-)! I do not wish to start another round of "chip wars", the Clipper is a good product; I just don't agree with some of the design tradeoffs that were made, and I see less "room to grow" for it. > Fast 68020's and vaporous 68030's are a good bit slower, but at least have > the benefit of executing the same flavor of object code... Agreed (for the Amiga). > Is there any good overview of the current R2000 chip available as opposed > to research papers on it's predecessors? Yes. There is a considerable amount of information and documentation on the R2000 available from MIPS. Email me if you want their business address. There have also been several articles in various trade rags within the last year or so (early 1986 issues of VLSI Design, and Micro-Mini Systems [or is it Mini-Micro Systems] come to mind). Several of the folks at MIPS also frequent comp.arch, especially John Mashey who heads up the OS group there. His email address is mash@mips.UUCP. /kim -- UUCP: kim@amdahl.amdahl.com or: {sun,decwrl,hplabs,pyramid,ihnp4,seismo,oliveb,cbosgd}!amdahl!kim DDD: 408-746-8462 USPS: Amdahl Corp. M/S 249, 1250 E. Arques Av, Sunnyvale, CA 94086 CIS: 76535,25 [ Any thoughts or opinions which may or may not have been expressed ] [ herein are my own. They are not necessarily those of my employer. ]