Conference hydra::amiga_v1

    For what it's worth, I've heard rumors that MicroBotics will offer
    a SCSI board 'real soon now' for the Starboard.  The advantage of
    this is obviously that two Starboards would be much more likely
    to function together than a Starboard and a Plingsnart together
    on the buss.  And, you could get 3.5 megs, a SCSI adaptor, battery
    backed clock, and 68881 all into a fairly attractive package.
    
    One option I think you've overloolked is the Insider 1 meg/clock
    board.  This memory hovers in the 'hidden' chip ram location, so
    it is not as fast as 'fast' ram.  The win here is that the 86 pin
    buss is left wide open for a combination fast ram/SCSI adaptor.
    I've heard good things about the Insider; Autoconfig, ease of
    installation, etc.. and Abel Supply will seel one for $255.
    

    re: .-1
    
    Actually, I had considered the Insider, but had the impression
    that it would have potentional problems with upward compatibility.
    I remember hearing something on the USENET about it not using true
    auto-configure.

    re: .1
    
    how much slower is it supposed to be? Are there any other advantages
    to putting the RAM there rather than on the expansion buss?

    The Insider was the topic of several USENET exchanges that I didn't
    follow that closely.  What I did glean is that Insider DOES
    auto-configure.  All this means is that 1.2 will recognize the extra
    memory on booting, without having to install 'ADDMEM' to your
    startup-sequence.  However, I think that it configures to a location
    that would have been used by the fat AGNUS chip, which can address
    two megabytes of ram.  As such, it is not considered true fast ram.
    
    What is the difference between true fast ram and the Insider?  I
    am not sure.  Maybe Randy Meyers can butt in and clear this up.
    To the user, I don't think you'd notice any difference.  I believe
    that real fast ram can be accessed at any instant, whereas non-fast
    (ie, chip, or hidden) requires that you wait for a clock cycle to
    come around.  I think this is what is meant by a wait state.
    
    What I do know is that lots of people are using the Insider without
    any problems.                    

Re: .*

I believe that the insider is recognized by the system as fast ram, runs
with out chip memory contention, and has no wait states.

Technically, the Insider board does not autoconfigure since it does not
follow the Zorro expansion autoconfigure spec.  Instead, it relies on
the Exec recognizing any memory those addresses lie within the reserved
address space of the Amiga.

In Zorro autoconfigure, the Amiga broadcasts a message to the devices
on the expansion buss asking what type of device it is and how much
address space it needs.  The device replies with a message like "I am
a memory expansion box and I need two meg of address space."  The Amiga
then sends a message to the device that says, "Fine, you now have a
base address of xxxxxxx".  The device has a latch that it uses to store
its base address at, and will then respond to any attempts to access
memory in its assigned address range.

Note that the user does not have to do anything to set the base address
of the memory expansion box.  There are no dip switches for the user to
set, etc.  I understand that one of the "innovations" of the new pc line
from IBM is an expansion architecture along the same lines as the Amiga's.

There is another way for the system to recognize memory: as the Exec boots,
it tries accessing every memory location that is on a 64K boundary in
the first 2 Meg of the address space.  This region of memory includes the
chip memory plus 1.5 Meg of memory marked as "reserved for use by Commodore."
It is fairly common knowledge that memory in this region may be used as
chip memory in a future Amiga model.  (Although, the new Amiga 500
does use this part of its memory map for the Commodore 1 Meg upgrade.)

The Exec is smart enough to correctly characterize memory found through
this boot up probe as chip memory or fast memory.  So, the Exec realizes
that the Insider memory is there, and it realizes that the memory starts
at an address not addressable by chip memory.

You find it interesting that this is the way that the Amiga 1000 realizes
if the 512K upgrade is installed.

Some of the manufactures of memory expansion hardware are unhappy that the
Insider claims to be autoconfigure when it is merely automatically recognized
because such claims cheapen autoconfigure.  Two autoconfigure boards can
be used together and the system will assign them unique address spaces.
Two autorecognized board can only be used together if the two boards are
built to live in separate address spaces.  However, autorecognized boards
have slightly less hardware than autoconfigure boards, and so may be slightly
cheaper to buy.

Memory on the Amiga can be slowed down two ways: First, the memory can
have a wait state.  Second, the memory may occasionally have "chip"
contention.

Memory with wait states built in has pretty much disappeared from the market
as far as I can tell (but it is always a good idea to make sure!).  These
memories have a design "defect" (if you will allow me to call it that) that
prevents the board from being able to deliver the memory to the 68000 as
fast as the 68000 can take it.  I believe the Insider has no wait states.

Memory which experiences chip contention is memory on the chip buss.  This
contention occurs then the Amiga's custom chips have to cycle steal, for
example, when using fancy high resolution graphics modes.

The Amiga effectively has two busses.  The first buss is the chip memory buss.
All the memory on the chip buss has its address decoding performed by Agnus.
This is the memory that can be addressed buy the custom chips, and surprise,
it is called "chip memory."

The other buss is the expansion buss (sometimes called the processor buss).
Memory on this buss has its own address decoding and refresh logic.  This
buss is not slowed down by activity on the chip buss, so the 68000 is free
to access this memory without wait states even if there is heavy chip memory
contention due to high resolution graphics.  This memory is called "fast
memory," because the custom chips never slow it down.  This memory need not
be really any faster that chip memory: if you bought a memory expansion
that had wait states, it could be slower!

The Insider is on the processor buss (literally, it is in the 68000 chip
socket!).  It should not experience chip memory contention.

(The Amiga 500 has blurred the distinction between chip memory and fast
memory.  In order to reduce the cost of the 500, the Agnus chip was
revised so that it could handle the address decoding and refresh for one
meg of memory instead of the old 512K.  However, none of the other Amiga
custom chips were enhanced to work with the additional memory.  So, on
the Amiga 500, the 512K in the 1 meg upgrade is not chip memory because
it can not be used with the custom chips.  However, it IS on the chip
buss, and so does experience chip memory contention in high res graphics
modes.  Commodore did things this way because it allows them to decrease
the number of components (the one meg upgrade doesn't need address decode
and refresh logic).)

To sum up:

The fact that the Insider is really only automatically recognized instead
of autoconfigure is only a problem if you decide to buy another product
that is only autorecognized.  For example, both the PAL and and PAL, Jr.
contain memory one meg of memory that is autorecognized, and thus would
not work with the Insider.

What you might want to do is call up the manufacturer of the Insider and
ask the following questions to confirm what I have said:

   1. Does your board have any wait states?

   2. Does your board ever experience chip memory contention?

With luck, if you call up the manufacturer, a techie type will answer
the phone.  (He might even be the guy who designed the board.)  His
answers will be authoritative.  If you are unlucky, you might get
just an answering service; they will be no help at all.  (You might
be able to get someone to call you back, though.)

    Um, yeah, that's what I really meant to say. :^)
    

Re: .5

By the way, if it wasn't clear in my previous message: In general, chip
memory has the same speed as fast memory.  In general means:

    1.  You aren't using any graphics modes not supported by an
	Atari ST.  (For example, the Workbench screen resolution
	doesn't cause the special chips to cycle steal.)

    2.  You aren't doing equally exotic things with the custom chips.
	(For example, you aren't running the blitter "in steal all the
	cycles" mode.)

    3.  You didn't buy a memory expansion with wait states.

Items 1 and 2 cause chip memory to be slower that fast memory.  Item 3
may cause fast memory to be slower than chip memory.


About the Starboard-2:

I have a Starboard-2 from Microbiotics, and am pleased with it.  When you
price a Starboard-2, there is something you should keep in mind:  The
Starboard-2 comes in several configurations.  There is the main unit
which may have only half a meg or one meg installed.  There is the upper
deck, a daughter board that fits inside the main unit and may have 0 meg
or 1 meg installed.  And then there is the multifunction board, which
is another daughter board that fits inside the main unit, and provides
a clock, parity checking, write protection, and a math chip socket.

When you get a price for a Starboard, make sure that you know what options
and how much memory is installed.  Some people sell the Starboard-2 with
.5 Meg installed and no upper deck, others sell a Starboard-2 with the
upper deck installed and a full 2 Meg of memory.

I bought my Starboard-2 as the main unit with .5 Meg installed, and an
upper deck with 0 Meg installed.  I bought the memory chips separately,
and installed them into the sockets.  Everything went very smoothly except
getting the boards back into the little metal case because it was a tight
fit.  After I bent a small tab in the metal case with a pair of needle nose
pliers, I didn't have any problem at all.  I think that anyone is capable
of performing this installation: My previous hardware projects have been
limited to changing light bulbs and I believe that soldering irons are
devices to burn your fingers on.  (I am glad the Starboard-2 doesn't
require any soldering.)

However, it is convenient to get the unit already stuffed with 2 Meg, and
I believe that some of the mail order places offer prices low enough that
there is no need for you to assemble things yourself.

    Hi...Have any of you Starboard owners bought the multifunction card
    and a math chip? If so, what is the speed increase?
    
    Thanks,
    
    PH
    

    The rumor mill suggests that the multifunction module is very close
    to shipping.  Also, Carl Sassenrath wrote the Microbotics recoverable
    ram software, so I expect it to be a qality job.  (The Starboard
    in general seems to be of the highest quality.  You could drive a
    truck over this thing and it would probably survive).
                                   
    As far as using a 68881 floating point math chip goes, I believe
    that Microbotics will supply a routine that replaces the current
    software floating point library (libs/ieeemathdoubbas) and gets the
    68881's attention.  The 68881 is still a pretty expensive chip,
    at least several hundred dollars (or more).  Expect pretty astounding
    speed increases over the current software floating point math. 
    Dave Wecker felt that Render would run around four times faster
    with a 68881.
    
    Even without the 68881, the Multi function module seems like a steal.
    Abel will proably sell it for around $79.  Add 4 256K RAM chips
    for each meg of fast ram, and you get hardware parity checking.
    A battery backed clock is also included.
                            
    All in all, the Starboard is a pretty nifty mini-card cage, all
    in a package about 1 1/4 " wide.  Abel WAS selling a Starboard stuffed
    with 2 megs for $430, but the price of 256K chips has been changing
    on a daily basis lately.
    
                                                                              
    

    Does the "sticky-disk" function of the Starboard work better than
    Recoverable RAM software?
    
    How is a parity error signaled to the user, and does parity checking
    slow things down?

Re: .10

My understanding is that the parity error is signaled by the Starboard
causing a buss error.  This would likely result is a task held requester
followed by a Guru.  (I am not sure: the Exec might consider a buss error
bad enough to go straight for a Guru without first giving a task held
requester.)

The parity checking is not there so much to allow you to recover from errors
as to provide instant warning that the Starboard-2 is having a hardware
failure.  Although, I guess, if you are an optimist, you could say that it
is there to provide you with a constant assurance that the Starboard is
working properly.

I would hope that parity checking would not slow it down, but what do I
know?

When I installed memory chips on my Starboard-2, I went ahead and filled the
eight parity sockets with the idea I would probably get the multi-function
when it came out.  However, I am being to lust after a Pal Jr, so I may
wait and see...

    
    
    Geese, Why did I have the idea that the 68881 was 50$?
    {
    
    (Was I wrong!)
    
    

    I just installed the INSIDER board and am having no problems so
    far!  Installation took about three hours, rather than the one hour
    stated in the instructions due to a number or minor matters that
    all got resolved in a straight forward manner.  The instructions
    are excellent, but I found that reading all the related notes from
    USENET helped me to both mentally prepare for the project and solve
    the minor problems.  The biggest problem and the only one not
    referenced in the instructions was that the INSIDER board did not
    fit with the internal disk drive RF shield in my (admittedly older)
    system.  As suggested in one of the USENET notes, I just swapped
    drives with my external drive.  The RF shield was smaller for it
    and that made all the difference.  It also added a hour to the project.
    When I powered up, sure enought it didn't work.  I checked all the
    connections and pushed the main pins into the socket more firmly.
    Voila!  Then final reassembly of the main box (and external drive)
    and the system has been running fine for a week.  The included software
    works fine and the user instructions for WB1.1 and WB1.2 and various
    programs are all relevant.  I love having that battery backed up
    clock.  The only thing I can't get to work according to instructions
    is having Transformer recognize a ful 640K.
    
    I ordered the INSIDER from Abel supply along with some other software.
    They were very helpful and friendly.  The software was shipped
    immediately and arrived in 2 days as requested.  They were out of
    the INSIDER and told me so on the phone.  They estimated 10 days
    for their new shipment and my board arrived within two weeks.  I
    charged it all on VISA.  They filled out separate slips and submitted
    them only when the products shipped.  Based on my one experience
    I would highly recommend them.
    

    Does it make more sense to buy the Starboard with or without RAMs
    installed?

Re: .14

It depends on what the pricing is: If you are paying list price for
the Starboard, getting empty sockets is the way to go.  However,
Abel Supply sells the Starboard with the chips installed at such
a low price, I wouldn't bother.

See the Abel price list for details.

    Re: .15 etc.
    Unless there are reasons (like financial ones) for not wanting to
    add the 2M all at one time.  I've been considering the Escort II
    unpopulated and adding .5M at a time as my pennies will allow. 
    
    Gary
    

    Anyone have any experience with the C-Ltd 1 meg memory expansion?
    It's apparently a slap on the side box rather than an internal
    expansion for the A1000. I have a chance to buy one used for $200
    and would like to hear of any experience good or bad people might
    have had with this product.
    
    Jack

    I think that it doesn't have a bus pass-thru.  You might want to
    check this out.