Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!newstop!sun!amdcad!brahms!phil
From: phil@brahms.amd.com (Phil Ngai)
Newsgroups: comp.sys.ibm.pc.hardware
Subject: Re: AMD Math Coprocessor again.
Message-ID: <1990Dec20.000822.15651@amd.com>
Date: 20 Dec 90 00:08:22 GMT
References: <1990Dec19.121014.3338@hellgate.utah.edu>
Sender: usenet@amd.com (NNTP Posting)
Organization: Advanced Micro Devices, Inc; Sunnyvale, CA
Lines: 30

Making semiconductors is a little like trying to figure out
how much sleep people need. Some people need more than
others. Everyone will need more or less depending on how
much they did that day (hard day chopping wood or easy
day relaxing?). Semiconductors generally can run faster
when they are cooler (there's a company, Velux, that sells
chip coolers to let you push your processor). Also the
factory includes at least a little extra margin to
guarantee the chip will work at spec. If the "yields"
are good, there may be a lot of margin.

As the product matures, the average speed will probably go
up from process and design improvements and tweaks.

Can you use a 10 MHz 287 at 10.67 MHz? Maybe, maybe not.
It is less than 7% over so I wouldn't be surprised if
it did work.

All you KNOW is that if you try it and it doesn't work,
the factory will not consider the device to be defective.
It is very unlikely that any physical damage would be caused
by such an experiment.

This is my opinion exclusively and I do not speak for the
company or work anywhere near that group.

Someone ought to put this kind of discussion in the list
of frequently asked questions, if it isn't already.

--