Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!clyde!cbatt!ihnp4!qantel!ptsfa!pttesac!vanam
From: vanam@pttesac.UUCP (Marnix van Ammers)
Newsgroups: net.micro.amiga
Subject: aMEGA 1Meg RAM speedup
Message-ID: <318@pttesac.UUCP>
Date: Wed, 29-Oct-86 00:59:04 EST
Article-I.D.: pttesac.318
Posted: Wed Oct 29 00:59:04 1986
Date-Received: Wed, 29-Oct-86 22:49:57 EST
Organization: Pacific*Bell ESAC, San Francisco
Lines: 61
Keywords: CardCo C.Ltd ram speedup fix


I've had one of the CardCo (now called C Ltd) 1 Megabyte memory
boards for about 3 months now.  It's called the aMEGA RAM board.

There was some talk in this newsgroup about that board having
a problem.  It apparently would cause wait states.  I tested it
with Perry's memtest program and found that the FAST external
memory (the aMEGA board) was running 27% slower than my CHIP
internal memory.

Last week I read an article in Amazing Computing which showed me
how to fix my board.  The article said that only the older boards
have this problem.  I had to cut a trace and solder a tiny wire
between two points.  I hadn't done any soldering in years and felt
somewhat nervous about doing this operation.  But I went ahead and
did it anyway.  I felt fortunate that the board worked (and still
is) after my soldering job.  The article was on page 29 of
Amazing Computing Vol 1, Number 7.

After the fix I ran Perry's memtest program again.  To my surprise,
not only did my FAST external memory speed up, but so did my CHIP
memory!  My CHIP memory sped up by 15%, the FAST memory by 27%.
I hadn't done the memtest program with my aMEGA board disconnected
before, but apparently it had caused a slow down in my internal
memory (besides taking wait states on it's own).  I'm hoping someone
out there will explain why.

Here are the results of Perry's memtest program:

Before modifying aMEGA card:
CHIP		FAST		DIFF
35400321	45000428	9600107
35416986	45117109	9700123
35383655	44900431	9516776
35400321	44833767	9433446

After modifying aMEGA card:
CHIP		FAST		DIFF
29950288	32433664	2483376
30200281	32416999	2216718
30066951	32533662	2466711

All tests were run with nothing else running.

As you can see I ran 4 tests before and only 3 afterwards.  There
is still a difference of 8% between CHIP and RAM.  This difference
was explained in an earlier posting.  As I recall, that posting said
that under normal circumstances the difference would be around
4%.  I suppose that the poster meant that normal circumstances are
when memory is not continuously accessed.

I would like to add an external zorro box (see I now know what zorro
means!) one day.  I understand that C ltd is coming out with an
expansion box.  I just hope that I won't have to throw my aMEGA board
away in order to use their expansion.  $495 is a lot to throw away!

Marnix     <- no, that's not an operating system!   It's my name.
-- 
Marnix A.  van\ Ammers
Home: (707) 644-9781		Work: (415) 545-8334
{ihnp4|ptsfa}!pttesac!vanam	CIS: 70027,70