Path: utzoo!attcan!uunet!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!mef From: mef@aplcen.apl.jhu.edu (Marty Fraeman) Newsgroups: comp.lang.forth Subject: Re: Forth Engines / Harris Summary: No way $15 chip Message-ID: <1990Dec10.171035.27718@aplcen.apl.jhu.edu> Date: 10 Dec 90 17:10:35 GMT References: <2078.UUL1.3#5129@willett.pgh.pa.us> Reply-To: mef@aplcen (Marty Fraeman) Organization: Johns Hopkins University Lines: 66 In article <2078.UUL1.3#5129@willett.pgh.pa.us> ForthNet@willett.pgh.pa.us (ForthNet articles from GEnie) writes: >Category 6, Topic 15 >Message 215 Sun Dec 09, 1990 >F.SERGEANT [Frank] at 14:55 CST > > PK> The sad fact is that Harris can't sell an RTX for $15 > . > Phil, I don't know what you mean when you say that. Do you mean that they >can't do so and make a profit? Or, do mean no one would be willing to >purchase it at that price? > . I don't mean to speak for Phil but you might consider some simple economic facts of life here. A CMOS mask set probably costs 20K$-30K$ depending on the number of layers needed. A dedicated run of a dozen or so 4" wafers for a 2u CMOS process probably costs 25K$ although you might be able to do it for a bit less if can keep volume up. The nice 84 pin PGA ceramic packages might cost as little as $20 each if you by them in >10K lots (ie order 200K$ worth of packages at a time). In small quantities I wouldn't be surprised to hear that they cost up to $80! A 4" wafer probably can hold 40-50 die the size of the RTX. Lets say you get 50% yield (pretty good for a die this size) then a wafer run will get you 250 good chips for your 25K$ or $100 for silicon alone. Note I didn't add any costs to probe the die on your tester. Also note how sensitive the cost/die is to yield and also note I have absolutely no idea what the real yield is. For obvious reasons, the foundries guard their yield numbers more then the family jewels since you can infer the financial health of company if you know it. Contact printing is commonly used in volume production so the mask set only lasts for 20 wafer lots or so. Therefore the mask costs you about (25K$/((200 wafers) * (25 die/wafer))) or about $5/die. If a projection system is used instead then masks last much longer but the equipment used is a lot more expensive. So to summarize, the whole thing is about: $100 silicon $ 5 masks $ 20 package ---- $125 materials alone! Now what about the labor to package the parts, distribute them to customers, testing, and of course you should get at least some amount of profit for your effort? To me at least this kind of an argument shows several why the industry is headed the way it is. A 1u version of an RTX chip would be about 1/4 the area so 4 times as many die get built in each wafer lot. pushing to greater density A 6" wafer has about twice as much area so you'd get about 500 good chips per lot at not much greater lot cost although your production equipment costs a lot more. Finally plastic packages are cheaper although not as efficient getting heat out so maximum chip speed is lowered because the junctions are hotter. Marty Fraeman mef@aplcen.apl.jhu.edu 301-953-5000, x8360 Room 13-s587 Johns Hopkins University/Applied Physics Laboratory Johns Hopkins Road Laurel, Md. 20723