jacobson@uiucuxe.CSO.UIUC.EDU (08/06/86)
se interested here is some of the information I received
with the memory card I received today in the mail at my dealers.
This is the aMEGA card by C. Ltd. (formerly Cardco). As you can
see it has the ground planes and they chose the Intel 8206 Ram
controller because of reliability and the fact that it is
available in low power CMOS version. Thus they have addressed the
most important items given by Perry in his warnings about RAM
cards. They do have wait states during RAM refresh that in Worst
Case give you 4.5% slow down. But as they explain below that
should only average 3% in memory intensive operations. Sounds to
me like such small loss is entirely acceptable. Anyway I post
this for your info and as a response to the very important
questions raise by Perry. All in all the board is well
constructed, and I am very happy with it. I hope this information
will help those making choices in which card to go with. C. Ltd's
card at this time is still the best buy and seems to be well-done
overall in my opinion.
aMEGA Board Instructions - by C. Ltd. - (316) 267-6525
The aMEGA Board - Design Perameters
The aMEGA Board was designed to fully comply with
the memory expansion methods designed by Commodore-Amiga for the
Amiga Computer. Additionally, the board was designed to operate
with the methods of memory expansion that are currently
available,prior the Com##odore-Amiga's release of the full memory
expansion software boot-up routines in Kickstart/WorkBench 1.2.
All components on the board are socketed to facilitate repairs.
The circuit board is a four layer board with full ground plane,
and it is housed in a metal case to eliminate RF noise and
interference. In genera# the board was designed using readily
available components and proven, time tested, dynamic RAM design
techniques.
Automatic Memorv Confi#u#ation with Rev.1.1 Kickstart/Workbench
To make a Rev.1.1 Workbench disk that allows the use of the
ConfigMem (see AutoConfig) routine with the aMEGA Board you must
boot up under Rev.1.1 Kickstart, then Rev.1.1 Workbench. Now
copy the INSTALL:c/ConfigMem file from the aMEGA INSTALL disk to
the c: directory of your Rev.l.l Workbench disk. Then you must
copy the INSTALL:libs/ExpansionMemory file from the libs:
directory of the aMEGA INSTALL disk to the libs: directory of
your Rev.l.l Workbench disk.
With these files on your Rev.1. 1 Workbench disk you
can now call the ConfigMem routine from CLI ( or in your
s/StartupSequence flie) just as you would any other
command.
Automatic Memorv Confiouration with Rev.1.2
Kickstart/Workbench
Memory Configuration with Rev.1.2 is totally automatic
and the routines are imbedded in the Rev. 1.2 Kickstart.
These
routines are not available to users.
The aMEGA Board - Soeed of Ooeration and Wait-States
The speed of operation of an Amiga external memory
board is determined primarialy by the number of "wait
states'' that are induced by the Dynamic RAM Controller IC
chip. A ''wait state" occurs when the microprocessor is
trying to read information from a memory location while it
is being "refreshed''. Dynamic RAM chips will only store
data for a short time, so they must be
"refreshed'' periodically, ''refreshing" the RAM chips is
one of the many functions performed by the RAM controller
chip. The Dynamic RAM Controller is the one central
component around which the entire RAM board is designed.
While it is possible to design a custom RAM controller that
induces no "wait states", we chose to use the very reliable,
well proven and easily available single chip Intel 8206 RAM
controller.
Page 10
aMEGA Board Instructions - by C. Ltd. - (316) 267-6525
This RAM controller will induce some "wait states" but
our decision to use it was based on our general design
concept that
some sacrafice in performance was acceptable in
exchange for
proven reliability and serviceability. The Intel RAM controller
is a several year old product, with proven performance. The
controller is easily obtainable to the general public in case a
replacement is needed. And, this controller is available in a
Low-Power CMOS version, which requires less power than other
controllers (putting less demand on your computer's power supply)
and produces far less heat than other controllers (which greatly
increases the life expectancy of other components on the board).
(Feel the case of the aMEGA Board during operation and you will
notice very little heat build up.)
As we said before a "wait state'' occurs when Your Amiga's
central processor is trying to access the external RAM while the
external RAM is being "refreshed". There are two factors that
determine how often these collisions happen. The first is how
often the RAM controller refreshes the RAM chips. Obviously, the
more often these refreshes occur, the more often a collision is
possible. With our Intel RAM controller, the time between
refresh cycles can be determined by the designer of the board to
fit the needs of the system. The Dynamic RAM chips available
today are specified by thier manufacturers to be refreshed at
least once every 4 milliseconds. P#freshing at this rate would
cause a worst case slow down of about 1.##. We, following our
general design philosophy of choosing reliability over raw speed,
chose to
refresh the RAM chips about 20# more often
than the RAM chip
makers require. This means that our board will have a worst
case slow down of about 4.5%. This also means that as your
RAM chips deterioriate with age and from heat build-up, our
board will be refreshing them more often and this will
considerably improve thier reliability and extend their
usefull life.
As we said this worst case slow down would be about
4.5%.
The second factor that determines !:ow often "wait
states" happen
lowers this worst case number considera#ly. "Wait states"
can onlv occur when the Amiga's microprocessor is tyring to
read information located in external RAM. In normal
operation the microprocessor ''fetches" a byte of data on
during one clock cycle, then takes several clock cycles to
process that information. Different instructions (ie.
adding vs. subtracting - subtracting takes less) require a
different number of clock cycles for the Amiga's
microprocessor to do its internal calculations, so what
the program is actually doing will change the number of
"wait states'' that occur, because "wait states" onlv happen
when the microprocessor is actually getting information from
the external RAM. So, our worst case slow down is based on
a program that is resident in the external RAM, uses short
instructions (with few internal microprocessor operations)
with all of its data in external RAM and all opertaions
using the external RAM. In the real world this happens in
only very, very few situations.
A good question at this point is "O.K., in reality, exactly
how much does this board slow down my Amiga?" Well, during
graphics operations, screen updates and any operations that take
place in the internal RAM of the Amiga, our board will not slow
down your conputer at all! During very memory intensive
operations like ''in memory sorting" and large spread sheet
recalculations you may notice the worst case slow down (on the
average of about 3%) during these operations. Based on our
observations of typical use we would estimate the average user,
would loose less than 15 seconds for each hour of computer use.
Well thats about all I had, sorry about the mispellings, I missed
them in the document which I used an optical scanner on. Does not always get
the words right.