Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga.hardware
Subject: Re: Overdriving 68030 CPUs
Message-ID: <10725@cbmvax.commodore.com>
Date: 9 Apr 90 21:01:08 GMT
References: <1990Mar31.223641.13301@aucs.uucp> <10602@cbmvax.commodore.com> <3163@rorschach.oakhill.UUCP> <10598@netcom.UUCP> <101213@convex.convex.com>
Reply-To: daveh@cbmvax (Dave Haynie)
Organization: Commodore, West Chester, PA
Lines: 43

In article <101213@convex.convex.com> swarren@convex.com (Steve Warren) writes:
>In article <10598@netcom.UUCP> mcmahan@netcom.UUCP (Dave Mc Mahan) writes:
>>In a short summary:
>> If the chip was able to be driven faster safely, don't you think Motorola
>> would have sold it as able to go that fast?  

>Well, I don't know how Motorola does it but a lot of companies don't have
>a seperate process for higher speed devices.  All the devices come off the
>same die, and they test for proper function at higher speeds until they
>fill their quota for that speed range.  Every device on the die may be
>capable of running at the higher speed.  But only the first N-devices
>that pass the test will be marked as higher-speed.

Last I heard, Motorola has two 68030 versions, the lower speed process can
yield only 16MHz, 20MHz, 25MHz, and 33MHz parts, the faster process can
yield only 40MHz and 50MHz parts.  The dropouts from the 50MHz process
can't serve as real 33MHz parts, because their dynamic characteristics
won't operate at the minimum speed required for 33MHz operation.  RAM
systems like cache aren't the only dynamic elements in fast CPUs; they use
some really weird ideas like dynamic latching.  No refresh is ever involved,
because such signals are guaranteed to be relatched if the CPU is run at
the appropriate speed.  These designs not only make the part smaller,
but faster.  Not something people thinking about standard TTL based system
level design like to think about, but when you consider than some signals
get in and out of the CMOS 50MHz 68030 much faster than the best F series
TTL parts, you have to figure that something's up.

You really never know how a process yields, and that's why overdriving any
part is a really bad idea.  You're OK when things work 100%, and you're
OK when things fail (eg, you fix them, slow down the clock, and they
work again).  But when they start acting flakey, you may waste a great
deal of time on a couple of bits dropping here and there.  You have
to depend on a reliable computer system.  You can't know if the 25MHz
chip you put in last January at 33MHz is going to be 100% this summer
in your 90F computer room, no matter how nicely it acted this up to now.


>--Steve      DISCLAIMER: All opinions are my own (I don't speak for Convex)

-- 
Dave Haynie Commodore-Amiga (Systems Engineering) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
                    Too much of everything is just enough