home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!ukma!rutgers!cbmvax!cbmehq!cbmswi!augs1!holli!holler
- From: holler@holli.UUCP (Jan Holler)
- Newsgroups: comp.sys.amiga.misc
- Subject: Re: Mike Sinz's tests? (was Re: Speeding up the A4000?)
- Message-ID: <holler.02ka@holli.UUCP>
- Date: 6 Nov 92 16:18:05 GMT
- References: <BwxvKC.686@unx.sas.com> <55535@dime.cs.umass.edu> <alien.00n1@acheron.amigans.gen.nz> <1992Nov02.170950.786@zds-oem.zds.com>
- Organization: my own mess
- Lines: 68
- X-NewsSoftware: GRn 1.16f (10.17.92) by Mike Schwartz & Michael B. Smith
-
- In article <1992Nov02.170950.786@zds-oem.zds.com> easton@zds-oem.zds.com (Jeff Easton) writes:
- > In article <alien.00n1@acheron.amigans.gen.nz> alien@acheron.amigans.gen.nz (Ross Smith) writes:
- >
- > [ Running chips faster than they are rated]
- >
- > >Suppose a chip comes in, say, 25 MHz, 33 MHz, and 40 MHz versions. This
- > >does *not* mean that the manufacturer actually makes three different chips.
- > >They make *one* chip, designed for 40 MHz plus a margin. All chips are
- > >tested at 40 MHz; those that pass are sold as 40 MHz chips, those that fail
- > >are tested at 33 MHz ... and so on down the line. You end up with four
- > >piles of chips :
-
- [stuff deleted]
-
- > >The important point being that, if you buy a chip rated at 25 MHz, that means
- > >that *that chip* (not just one of the same batch) has *already* failed at
- > >least once at 33 MHz! Trusting it because "it seems to work OK for me..."
- > >is *dumb*.
- >
- > Which is a great assumption until the chip fab has tuned the process
- > to a point that they are yeilding 40 Mhz parts all the time. Then it
- > becomes a marketing issue.
- >
- > Say this month Motorola will sell the following breakdown of chips;
- >
- > 5000 40Mhz 68040's
- > 10000 33Mhz 68040's
- > 50000 25Mhz 68040's
- >
- > The chip fab yeilds 100% good parts. The process has been pretty good
-
- [stuff deleted]
-
- I dare say that this all is not true. Take a look at the timing inside the
- chip! The most important thing is the length of the signal way inside. You
- just cannot go up with frequency for a chip designed for a certain range of
- frequency. On the other side cooling is also important!
-
- Take the Motorola 68020:
-
- RC12 RC16 RC20 RC25
- min max min max min max min max
-
- Freq 8 12.5 8 16.67 12.5 20 12.5 25
-
- cycle time 80 125 60 125 50 80 40 80
-
-
-
- What you say is, that they are all the same. Ok, so they are all 25Mhz
- parts. True? No:
-
- A RC25 fails the 25Mhz test, also fails all others, but not the 12.5MHz.
-
- Timing is halved: Freq goes down to 12.5, wich is also the lower range for
- an original RC25! Not possible to take that down to 8Mhz!
-
- Cycle time is doubled: Let's say we had a cycle time of 80ns which was
- the upper range of the RC25. So it's now 160ns. But that is way too high.
- The RC12 is just designed for 125ns cycle time.
-
- See! Don't make assumptions because of the ease of them! Sure it would be
- great to have it your way. But there is also physics to concern about!
-
- --
-
- Jan Holler - Bern, Switzerland
- uunet!cbmvax!cbmehq!cbmswi!augs1!holli!Jan_Holler
-