Conference hydra::amiga_v1

    Microbotics seems to run ad campaigns only when they are introducing
    new products.  There hasn't been much news from them lately.
    
    There really haven't been a lot of new controllers released since the
    last round of magazine review/tests about 6 months ago.  My guess is 
    that the HardFrame is still Top-of-the-Heap.  It will be interesting 
    to see how Commodore's new A2091 stacks up.  I suspect it will do
    fairly well.
    
    The only 'DMA problems' are due to improperly designed plug-into-the-
    68000-socket units, like CMI's Processor Accelerator.  (And WordSync
    is from Supra, not GVP.)
    
    I can't comment on the ST296N drive, I went for a 105 meg Quantum on 
    my HardFrame.  Love it!

Re: .0

>Is there any credit to the claim by GVP that DMA controllers
>can cause all kinds of woes? 

Re: .1

>    The only 'DMA problems' are due to improperly designed plug-into-the-
>    68000-socket units, like CMI's Processor Accelerator.

The "DMA is bad" claims are mostly a result of a problem in the A2090
and A2090a disk controllers from Commodore.  These controllers had a
problem that caused them to experience enormous slow-downs when the
custom chips in the Amiga were cycle-stealing.  The new A2091 doesn't
have this problem at all.

The A2090(a) depended on being able to store the data coming off the disk
into memory with little delay.  If the buffer the A2090(a) was trying to
store into was in chip ram, and if the custom chips were monopolizing
that memory (cycle-stealing) the A2090(a) would fail to be able to
load its data into memory, and would abort the disk operation.  The
disk driver would then retry the operation, and would take about
10 retries (or more) before succeeding.  Disk I/O would enormously
slow down.

DMA disk controllers do stress the system more: for DMA to work, the
hardware (not just the disk controller, but memory cards, processor
accelerators, etc.) must be designed correctly.  Marginal hardware
that breaks a few rules might fail to work under intense DMA
activity.  The rules for building Amiga hardware are written so that
DMA will work.  There have been a few instances of hardware developers
breaking the rules.

DMA has the advantage that it's faster, and allows the CPU to do other
work while disk I/O is going on.  In effect, it is like another "custom
chip" that off-loads work from the 68000.  Non-DMA controllers make the
CPU copy the data from a private buffer on the disk controller into
the Amiga's memory.  If there is disk I/O being done, the 68000 isn't
doing anything else.

Re: .0

I'm a bit down on Seagate drives: one died on me a month after the
warranty ran out.  The Quantum drives are faster, and have a warranty
that lasts a year longer than the Seagate warranty.

    re: A2091 Disk Controller
    
    Has anyone used the A2091 controller with an Adaptec or an Omti
    adapter to an ST506 style drive?  I have a varied mixture of drives
    on my system and was wondering how well the A2091 dealt with
    non-imbeded SCSI drives with adapters.
    
    Anyone get an RZ23 up on the A2091 yet?
    
    John
    

John,

>   Has anyone used the A2091 controller with an Adaptec or an Omti
>   adapter to an ST506 style drive?  I have a varied mixture of drives

My understanding is the 2091 is VERY similar to a HardFrame. Both host
adapter drivers were written by Bob (Kodiak) Burns. The "should" be 
very compatible. I have not tried the 2091 yet, but I have seen it.
It takes advantage of custom LSI to reduce the part count dramatically.

	-SES

    Re: .3
    
    As far as my experiences with the A590 and A2091 (basically the
    same thing) show, both should have no problems dealing with your
    setup. They're far more versatile than the 2090(a), that is, the
    software is...
    
    Regards,
    
    <CB>

    Hi,
    
    While on the subject of good controllers has anyone used the GVP scsi
    controller that can have 8Mb (SIPPS) ram onboard (with the disk off-card).
    
    I am probably going to change to this rather than the scsi+2mb I have
    been using - but if there are any horror stories.....
    
    regards
    kevin

Kevin,

>    I am probably going to change to this rather than the scsi+2mb I have
>    been using - but if there are any horror stories.....

If *I* were you, I would consider another alternative... Move to an A2620
or A2630 and be happy with 5 (or adding chips) 7 meg of memory. The A2620's 
are readily available used and are most addictive. It will also make an
improvement in performance of the non-dma GVP controller.

Might eventually consider moving to an A2091 or HardFrame rather than a
GVP for the DMA. It makes a noticable difference in multitasking performance
on my Amiga.

	-SES

    re .6
    
    	If you are already using the GVP SCSI + 2MB then the SCSI + 8MB
    will work for you. Both controllers are exactly the same in the SCSI
    area ( same driver, same boot ROMS ).
    
    
    		Steve