Contents | < Browse | Browse >

/// Usenet Review:  Zeus 68040 Accelerator for Amiga 2000
    By David W. Walthour


        Zeus 68040 accelerator for the Amiga 2000


        68040 accelerator card for the Amiga 2000 with SCSI II controller.
Expandable to 64 MB of 32-bit RAM.  This review describes the 28 MHz
version of the card.


        Name:           Progressive Peripherals Inc.
        Address:        938 Quail Street
                        Lakewood, CO USA 80215-5513

        Telephone:      (303) 238-5555
        FAX:            (303) 235-0600
        BBS:            (303) 238-6326


        Due to the sudden drop in price of this card last February, I am
uncertain of its current list price; however, I believe it is about
$1495.00 (US) with no RAM installed.

        When I purchased my Zeus in February, the price was $649.50 (US) with
no RAM installed.


        The Zeus card requires that you have an Amiga 2000 computer.  The
board is designed to handle both the original German A2000 and the B2000
(rev 3.9 and greater).


        Amiga B2000 with rev 3.9 motherboard
        1 MB Chip RAM
        AmigaDOS 2.1 (Kickstart 37.175, Workbench 38.35)
        28MHz Zeus, 12 MB 32-bit RAM, with mounted Quantum 105LPS hard drive


        [MODERATOR'S NOTE:  If you are not comfortable opening up your Amiga
        or handling expansion devices, then you should have the work done by
        an authorized Amiga service center.  Opening your Amiga yourself may
        void your warranty, and careless work may even damage the machine.
        - Dan]

        The board contains the 68040 processor, which is cooled by a small
fan mounted on top of it, 16 SIMM sockets for adding up to 64 MB of RAM, an
internal and external SCSI II connector for adding either SCSI or SCSI II
devices, and a mounting bracket for attaching a low profile SCSI drive to the
back of the card.

        The board takes standard 1x8 or 4x8 SIMMs which are 80ns or faster
(28 MHz Zeus) or 60ns or faster (33 MHz Zeus).  The SIMMs are installed in
groups of 4 and must be of the same type (1x8 or 4x8).  You can mix groups
of 1x8s and 4x8s on the board to achieve many different memory
combinations.  This allows you to start cheaply with 1x8 SIMMs and add
larger amounts of RAM using 4x8s later.

        You need at least one group of 4 SIMMs on the card in order to use
it.  Without 32-bit memory on the card, it is likely that you won't notice
much speed improvement, and certain operations may actually run slower than
on a stock 2000.

        The SIMMs snap into the sockets easily; however it is recommended
that you get your dealer to install the memory for you.  If you are foolhardy
enough to install memory yourself (like me), you should wear a grounding
strap to avoid damaging the SIMMs.

        After you install memory on the board, you have to configure the
jumpers on the board.  PPI was very thoughtful in designing this board and
put these jumpers at the top of the card so that when it is installed in the
computer, they are still accessible without having to remove the card.
There are about 12 jumpers to configure which control the memory, computer
type and default settings of the board, and the manual does an adequate job
of explaining most of these.

        However, there is one set of jumpers that the manual is rather
unclear about.  These jumpers control the amount of memory that will be
autoconfigured.  Autoconfigured memory has an advantage that it is available
to the Amiga even if you boot it in "native" mode (using the 68000 instead
of the Zeus's 68040).  However, there is a disadvantage:  the autoconfigured
memory is accessed as if it were 16-bit memory rather than 32-bit memory
when in 68040 mode, resulting in a 60% slowdown when the '040 uses this
memory.  So you should consider the specifics of your setup before choosing
the amount of RAM to autoconfigure (0, 2, 4, or 8 MB).

        A special note for Amigas with only Chip RAM:  if you choose 0 MB
autoconfigured, then Kickstart will get loaded into Chip RAM, slowing down
your performance.  So you need to set aside some Autoconfigure memory
(typically 2 MB) so that Kickstart will be in Fast 16-bit RAM; however, it
is not possible to get them opened in the much faster 32-bit RAM.

        Mounting a SCSI drive is easy using the mounting bracket on the card
and the additional hardware provided.

        Now you can install the Zeus into the 2000.  You should also have a
dealer install the board for you, but here are a few tips if you decide to
go it alone.  Installation of the board is somewhat hampered by the fact
that the fan atop the '040 sticks rather far out of the board.  As a result,
to insert the board, the manual directs you to remove all other cards from
your 2000 and swing the card in from the side (the manual has pictures to
help describe this).  This was rather inconvenient for me because I have
quite a few other cards in my system, so I removed the screws from the drive
bay/power supply cage in my 2000 to move it slightly out of the way so I
could insert the card straight down.  Either method gets the job done, but
neither is particularly easy.  It would have been nice for PPI to design the
board for easier installation.

        The only thing remaining is to install the software.  This consists
of placing a few libraries in your LIBS: directory and placing a program
called INIT040 at the beginning of your startup-sequence.  This program adds
the 32-bit memory to the system, places a Kickstart image in 32-bit RAM and
moves the system vectors to 32-bit RAM.  A program is provided to move these
things from the installation disk to your hard drive, but unfortunately my
installation disk was slightly damaged and so I had to install these
programs and libraries myself.


        When I got the Zeus configured and installed, and I booted my
machine, I was amazed by how much faster it was.  To quantify this speed
increase, SysInfo 3.18 reports about 21 MIPS (millions of instructions per
second) and about 5.33 MFLOPS (millions of floating point instructions per
second).  This is about 29 times faster than a stock 2000 in MIPS and about
350 times faster in MFLOPS.  Using the benchmark program AIBB 6.1, I found
that, overall, the Zeus card was 1.74 times faster than an Amiga 4000 at
integer tests, 1.15 times faster at floating point tests, but only 0.62
times as fast at graphics tests (due to the 4000's AGA graphics chips).

        Quantifying the speed of the SCSI II controller is not possible for
me because I don't have a SCSI II drive.  However, with my Quantum 105 SCSI
drive, I find that I can potentially achieve a transfer rate of about 1.1
MB/sec from Diskspeed 4.2 which seems to be the limit of the Quantum drive.
This performance unfortunately drops off quickly as the drive becomes

        I have used my Zeus for about 4 months now and have had very little
software trouble with it.  Some games and some PD demos programs have not
worked with the 68040 processor, but in such cases it is easy to reboot the
machine in 68000 mode to run the program.  The vast majority of software I
have used is compatible with Zeus.


        The documentation is well written; and except for the lack of
information regarding the technicalities of the autoconfig memory (see
above), it is very complete.


        Overall, I am extremely satisfied with the Zeus accelerator.  From a
hardware point of view, I think it is extremely well designed. However, I do
have two dislikes about the software:  one relatively small and one large:

        My smaller dislike is with the INIT040 program in that it
initializes the system in a nonstandard way.  For instance, when it remaps
the Kickstart into 32-bit memory, it does so by not adding 1 MB of the
32-bit RAM to the system and using this to hold the 512 KByte Kickstart
image.  The result is that the remaining 512 KBytes that aren't configured
are wasted and can't be used by the system.  Furthermore, INIT040
initializes the Memory Management Unit (MMU) tables of the 68040 processor
in a non-standard way which can conflict with other programs (such as
EMPLANT) which also try to use the MMU.  For these reasons, I suggest that
anyone using Zeus NOT use the INIT040 program.  Nic Wilson's Set040 (on Fred
Fish #642) can be used to map the Kickstart ROM.  There are also a variety of
programs to add memory to the system and to move the system vectors to
32-bit RAM, and the latest version of Commodore's SetPatch and 68040.library
will set up the MMU tables in the standard way.

        My large dislike is with the device driver for the SCSI II
controller.  For some unknown reason, PPI wrote the driver in such a way
that it steals nearly ALL of the CPU cycles when it is reading or writing to
any devices attached to it.  <FLAME ON>  This is an Amiga, not a MAC!  Device
drivers should be properly written to utilize the Amiga's multitasking
architecture and should not use busy loops to poll for data from the
device!  The stupid programmers at PPI better get their act together and
rewrite the device driver properly! <FLAME OFF>

        Neither of these problems is big enough, however, to recommend
not purchasing the Zeus card.  It is a phenomenal value.


        I have never used another 68040 accelerator for the 2000, so I can't
compare it to other 68040 products.  However, I have used a GVP '030 Combo
accelerator and can say that Zeus compares favorably with it.  I feel that
the Zeus has a superior hardware design in that it is built around using
inexpensive 1x8 and 4x8 RAM SIMM modules, whereas the GVP accelerators use
proprietary SIMMs which cost quite a bit more ($50+ per MB for the GVP vs.
$33 per MB for the Zeus).  The GVP accelerators, though, have more
thoughtfully designed software which operates with AmigaDOS in a more
standard way (like not using busy loops in their device drivers).


        I have found no bugs other than those already listed above.  Some
other owners on Usenet have complained that they have examined their 68040
chip and found it to be marked as a "68EC040", the version which does not
contain an integral MMU, and some users who have checked have complained
that they could not get the MMU to operate.  PPI claims that they received a
shipment of '040 chips from Motorola which were incorrectly marked.  I have
not examined my '040 chip and my MMU does operate correctly, so I can't
really comment on this issue.


        Vendor support from PPI has generally been poor.  When you attempt
to contact them, you almost always reach an answering system, and they have
taken over a week on average to respond to any questions I have left on
their system.  If you have questions regarding Zeus, I would recommend
turning to USENET or your local dealer before trying to contact PPI.


        PPI offers a limited 1 year warranty on Zeus which extends only to
the original purchaser.


        Overall, I feel that the Zeus accelerator is a very good product and
would give it a rating of:

o       4 out of 5 overall

o       5 out of 5 on hardware design

o       2.5 out of 5 on software for the poorly designed, but bug-free,
        device driver

o       4 out of 5 for the documentation

I would recommend it to anyone buying an accelerator for their 2000.


An update from Alan Quirt...

I just read David W. Walthour's review of the PPI Zeus board. I followed the
discussion in c.s.a.hardware when the price reduction was big news, and kept
notes on problems and suggested solutions.  Two things should perhaps be
added to the review.

1. The Init040 program supplied by PPI should NOT be placed at the beginning
of your startup-sequence if you are running any OS version earlier than 2.1.
It will crash your Amiga.  Instead it should follow LoadWB.  (Of course you
may not want to run it at all, for reasons given in the review.)

2. Some Zeus boards came from the factory with the autoconfig jumper set for
4 MB of autoconfig memory. If you have only 4 MB on the board and leave that
jumper unchanged, nothing works.  Several people said to avoid using over 2
MB autoconfig.  I think it would be best for anyone with other fast RAM to
leave all Zeus memory non-autoconfig.

Both of these hints come from USENET, not from personal experience.
However, several people reported the problems, and I saw no contradictions
in the posted solutions.

Regards, Al Quirt