Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!bbn!apple!versatc!mips!mash From: mash@mips.COM (John Mashey) Newsgroups: comp.arch Subject: Re: 68040 and 80486 Message-ID: <17178@winchester.mips.COM> Date: 13 Apr 89 18:36:00 GMT References: <1032@myrias.UUCP> <12289@reed.UUCP> <1049@myrias.UUCP> <1928@trantor.harris-atd.com> Reply-To: mash@mips.COM (John Mashey) Organization: MIPS Computer Systems, Sunnyvale, CA Lines: 46 In article <1928@trantor.harris-atd.com> thaker@trantor.harris-atd.com (Gautam Thaker) writes: >Why is comp.arch so quiet these days? What happened to the >good old days of "processor wars"? 68040 and 486 have both >been announced and there is almost nothing about them in >the newsgroup. We have debated the question of which is >better, the 386 or the 68030? Now can someone tell me which >is better in the next round; the 68040 or the 486? Please, let us not fill THIS newsgroup with: "I think the xxx will be better than yyy because xxx's vendor says it will be N mips, and xxx's only says it will be M mips". "The xxx's architecture is awful, hence yyy must be better." "xxx has always lied, so yyy must be better". "I heard a fourth-hand rumor that yyy will be really fas "yyy runs at Z MHz more than xxx, so it must be better" "xxx and yyy are both scum, zzz is clearly better anyway." "xxx has more transistors than yyy, so it must be better" Many of us would welcome DATA, especially as both these chips are important; they're also interesting from an architectural view. a) It will be fascinating to compare 486 and i860 performance, costs, and delivery, given that this is probably the best RISC-vs-CISC comparisons around [same company; same technology; same-sized chip; designed about the same time.] b) For both 486 and 68040, it will be interesting to see the techniques used to improve processor efficiency, and what can be learned from them. In particular, did the use of VLSI lead to different improvement methods than those used over the years by mainframe/supermini vendors? However, other than making gross estimates, it's pretty hard to say very much about performance until one has: c) Well-documented simulations of buildable systems. (zero-wait-state main memories don't count :-) OR, REALLY: d) Lots of benchmarks run on real machines. How long will it be until one can get d)? (a while) Hence, I suggest that this issue may take a while to settle, or even reach reasonable conclusions. It would certainly be useful if knowledgable people from Intel & Moto would post FACTUAL information, so that we're not bombarded by second/third/ fourth - hand factoids.... -- -john mashey DISCLAIMER: UUCP: {ames,decwrl,prls,pyramid}!mips!mash OR mash@mips.com DDD: 408-991-0253 or 408-720-1700, x253 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086