NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/sys/apple2/20269

Text File | 1992-09-12 | 3.9 KB | 74 lines

Newsgroups: comp.sys.apple2
Path: sparky!uunet!elroy.jpl.nasa.gov!nntp-server.caltech.edu!toddpw
From: [e-mail] (Todd P. Whitesel)
Subject: Re: Accelerator performance
Message-ID: <[e-mail]>
Sender: [e-mail]
Nntp-Posting-Host: punisher
Organization: California Institute of Technology, Pasadena
References: <[e-mail]> <[e-mail]>
Date: Sun, 13 Sep 1992 05:06:08 GMT
Lines: 61

[e-mail] (Harold Byron Bouma) writes:

>    Well, up until the 14Mhz 65816 came out, all Zips above the speed of
>8 Mhz were 8 Mhz chips pushed to go faster.

Not true. Zip has been using high-speed WDC prototypes also. When I went in to
get my 8-running-at-10 chip replaced, they gave me one of those and I was told
that the chips could go up to 11.19 mhz and that 45ns cache (what I have) was
fast enough to keep up at that speed. The prototype chips had no speed stamp
on them; WDC was not selling _those_ as 8 mhz parts, they were charging an arm
and a leg for them (in comparison).

>Todd Whitesel (sic) had written
>to Zip a while ago upset that his 10 Mhz chip had become damaged from the
>high speeds, and Zip wrote back saying that out of the 10,000+ 10 Mhz zips
>sold, none had become damaged from being ran at 10 Mhz. And later the real
>reason why Todd's processor becoming damaged had been discovered, which was

... STATIC DAMAGE THAT OCCURRED MANY MONTHS BEFORE I NOTICED ANY PROBLEMS.
That is why I never suspected static damage until I called (not wrote) Zip and
talked to tech support voice.

>    Now, before the 14 Mhz chip, Zip would buy large quantities of 8 Mhz
>chips and then run them at higher speeds to see if they could run at that

This is what they originally did. In the beginning, the best WDC could do was
6 mhz -- 7's were only available just before the ZipGS shipped, and since 7 is
the first speed at which you notice AppleTalk problems, it explains why Zip had
to scramble to deal with the AppleTalk problem weeks before the ship deadline.

>speed. WDC sells chips rated at 8 Mhz, but many are capable of running faster.
>But there isn't anything wrong with running a chip faster than its rating, or
>at least nothing proven yet.

The whole point of rating chip speeds is so you don't HAVE to prove anything --
you know they will run at that speed because the manufacturer tested them for
you. Usually the manufacturer has a more restrictive list of timing specs that
each chip has to meet, since they take into account normal degradation of the
chips and expect them to meet spec for many, many years. Depending on how you
design your board you can often get away with running the chip faster ... this
technique is used all too often in the industry and when these "aggressive"
designs get older, they tend to stop working because the chips start to degrade
to the point where they no longer run as fast as they used to, but still within
the original timing spec (which is being violated by the "aggressive" design).

In other words, I do not feel comfortable running a chip above its rated speed
unless I know I can downgrade TO the rated speed when normal wear-n-tear bring
the chip back down to less optimistic timing specs. That is why I now have
crystals to run my Zip at 10.5 (current), 10, and 8 mhz. When I had the problem
with my first 10 mhz CPU, I dropped down to 8 mhz and ran at 8 for two months
while I tried to arrange a car ride to Zip's office on a weekday (and since I
was still at college this was a lot harder than it sounds).

The reason nothing is proven yet is that nobody's accelerator is old enough to
exhibit problems due solely to aging. I do not have a very good feel for the
statistics here, so the best guess I can make is in the 5-10 year range. That
is long enough to explain how the people I learned all this from found out for
themselves, the hard way...

Todd Whitesel
toddpw @ cco.caltech.edu