Article View: comp.sys.ibm.pc.hardware.chips
Article #32Re: The "Case" for the AMD K6-2 and Socket 7 longevity
From: Keith R. Williams
Date: Mon, 22 Jun 1998 00:00
Date: Mon, 22 Jun 1998 00:00
50 lines
3394 bytes
3394 bytes
On Sun, 21 Jun 1998 17:32:00, "Doug Chism" <chismdm@earthlink.net> wrote: > One thing I can say is that as far as chip prices go - they are almost > always directly related to yield. when they make the chips - a large number > on the wafer cannot be used at the tolerances they designed them to. So they > are used for lower clock versions. If 95% of the chips come out of > manufacturing only meeting the 350Mhz spec - they have to charge a lot more > for the 400Mhz chip. Hogwash! Prices are determined solely by market demand. It matters not one iota what the yields are. Price and cost have no correlation in the semiconductor business. Evidence the Celeron. It *is* a Deschutes chip on a Deschutes card. It doesn't have an L2 and is clock-locked. Yet it sell for a few bits less than a PII-350 or PII-400. It is the same chip, remember. > Sure - game players dont care much about these > tolerances cause they run the hell out of them, but people who use these > things at work expect them to be reliable and not buggy. Again, Hogwash! I don't overclock business systems, or even those at home, but to say that it's technology that drives the business is just naieve. Intel got caught short with the K6 and did whatever it could to cover the low end while trying to support the margins at the high end. They packaged Deschutes into Celerons (the Covington variety) as a stop-gap. Tell me again that technology reigns supreme. Intel is so lost in themselves that they didn't see the <$1000 systems coming, though it was obvious for six months beforehand (and I'll be generous). > Its not like you > can start up a fab shop and have chips rolling out the door in two weeks > that are in tolerances. Manufacturing takes time, and money, before a steady > consistent flow of chips can be produced. Costs reflect this. Take a class > in materials engineering to see what types of tolerances have to be followed > to get clean fast silicon wafers with .25 die chips. Good grief. Intel doesn't have the market cornered on smart people. Processors are now taking their place along side memory (not to the same degree, admittadly) as commodities. Intel hasn't a clue how to deal with this. This was *obvious* at the last Intel Developer's Forum. ---- Keith R. Williams krw@ibm.net
Technical Details
Message-ID:
<pLMYl5dhX7hK-pn2-KvIl1dOJWEdo@localhost>Path:
arc1-ca.pugleaf.net!unknown!not-for-mailReferences:
<6mcgq7$q3r$1@camel18.mindspring.com> <6mhdju$mjf$1@bolivia.it.earthlink.net> <358cdc38.209015805@news.iafrica.com> <6mjg2i$f00$1@ecuador.it.earthlink.net>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10ec8c,a85535ffb77361a8 X-Google-Attributes: gid10ec8c,public X-Google-Thread: f8d87,a85535ffb77361a8 X-Google-Attributes: gidf8d87,public X-Google-Thread: 1045b9,a85535ffb77361a8 X-Google-Attributes: gid1045b9,public From: krw@ibm.net (Keith R. Williams) Subject: Re: The "Case" for the AMD K6-2 and Socket 7 longevity Date: 1998/06/22 Message-ID: <pLMYl5dhX7hK-pn2-KvIl1dOJWEdo@localhost> X-Deja-AN: 364847952 Content-Transfer-Encoding: 8bit References: <6mcgq7$q3r$1@camel18.mindspring.com> <6mhdju$mjf$1@bolivia.it.earthlink.net> <358cdc38.209015805@news.iafrica.com> <6mjg2i$f00$1@ecuador.it.earthlink.net> X-Notice: should be reported to postmaster@ibm.net MIME-Version: 1.0 Reply-To: krw@ibm.net Content-Type: text/plain; charset="us-ascii" X-Complaints-To: postmaster@ibm.net X-Trace: 22 Jun 1998 02:28:16 GMT, 32.100.110.251 Organization: IBM.NET Newsgroups: comp.sys.ibm.pc.hardware.chips,alt.games.unreal,3dfx.products.voodoo2