kurt@fluke.UUCP (Kurt Guntheroth) (12/12/85)
On bug lists: We (Fluke) never had any trouble getting bug lists for the 16000 under non-disclosure. To the contrary, when we were investigating processors it was Motorola who denied the existence of any bugs, even in the presence of publicly demonstrated bugs and brain damage in Motorola parts. I have personally seen a bug list for a very old 16000 rev and the bugs were so minor they would be virtually undetectable under normal circumstances. Bugs are a fact of chip design. Motorola has bugs, some were serious. Intel had very serious bugs in early '186's (remember those?). Yes, National has some too. It may strike some of you as funny that DEC still comes by regularly and installs ECO's on our 10 year old PDP-11's, so the chance of getting it all right on a single chip doesn't seem to high. On speed: Does the 68020 beat the 32032? Should you be surprised? The 32032 is the same die as the 32016 with more address lines led out. This part has a single 32 bit ALU and simple internal architecture. The 68020? Three, count 'em three ALU's. 256 byte cache. Incredible gyrations for speed. And about 6 times as many gates. That means bigger dice and bigger bug lists potentially. That means higher cost too. Basically all this junk was necessary to repair mistakes and inelegencies in the design of the 68000. The 32000 has a cleaner instruction set and is potentially a higher speed architecture. If you want to check relative speeds of the architectures, look at the 32016 against the 68010, which it is directly comparable with. The 32016 is faster, and generates smaller code modules for the same (C) program. If NSC and TI don't stand still with their architectures, you can expect to see a competitor with equal or greater speed. If they do stand still, you can still expect the simpler chips in the 32000 line to have a cost (or profit) advantage over 68000 parts. On flamage: What is it about these processors that generates so much flamage? There is the most ridiculous lot of misinformation and stupidity going around. No bugs is a claim that is impossible to support, yet Mot fans tried for awhile. Comparing the 32032 to the 68020 is foolish when they are so different. So why attempt to build a case for the superiority of the 68000 architecture over the 32000 architecture on such a comparison? Neither Mot nor NSC will ever beat Intel's market share, yet it is the conventional wisdom that both parts are easier to program with than the x86. It seems to me that each part has its niche. NSC for elegant and original software at a low price, Mot for high power at a high price, and Intel for compatibility at any price. Let us relax, secure in the knowledge that probably all three parts will suddenly be destroyed in the market by some strange parallel machine or RISC that is lurking on some design table, simulation system, or fevered brain.