Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!think.com!mintaka!bloom-beacon!eru!hagbard!sunic!mcsun!hp4nl!telmail!neabbs!ajbrouw
From: ajbrouw@neabbs.UUCP (ALBERT-JAN BROUWER)
Newsgroups: comp.sys.amiga.hardware
Subject: Re: A3000UX, is the hardware the same?
Message-ID: <625269@neabbs.UUCP>
Date: 14 Apr 91 17:02:34 GMT
Organization: NEABBS multi-line BBS +31-20-5733533 (20x), Amsterdam, Holland.
Lines: 45


John Veldthuis wrote:
>>I have an A2620 in my Amiga and I have lockup troubles when DMA'ing to 32
>>bit memory with my HardFrame Harddisk and want to set the MASK so that the
>>HardFrame only uses the 4 meg of Fast memory starting at $00600000 as
>>buffers

I've exactly the same problem John has. Config: Hardframe, 2630, Quantum.
Lockups are only on _writes_ from 32bits RAM. Shuffling the boards to less
optimal slot positions makes reads to 32bit RAM cause trouble as well.

Dave Haynie wrote:
> There are actually two potential problems.  An unknown number of A2000 
> motherboards prior to Revision 4.5 have a particular brand of buffer chip in
> one location between local and expansion buses that causes a marginal condition
> (...)
> This is cured by adding a 1K resistor between pins 11 and 20 of U605 in the
> A2000

I've a rev 6.0 2000 upgraded to 6.2. and tried the U605 patch. It didn't make
a difference ofcourse (6.0 being > 4.5)

> Early Hardframes also had a problem, essentially a marginal bit of timing on
> their DMA handoff logic, nothing off enough to affect the slower 68000, but
> trouble for any coprocessor board with fast onboard memory.  Microbotics has
> a fix for this, if you have one of the older boards.

I've the latest HardFrame chips. Have sent it to MicroBotics to have this
verified as well as getting the latest ROM. So that aint the problem either.


So... Does anyone know about other conditions, in addition to those mentioned
above, that cause DMA lockup notably FROM 32bits memory?


As a temporary solution I've hacked up the FastFilingSystem to allocate
buffers in 16 bit RAM, and write through the buffers instead of direct
DMA, when writing from 32bit memory. However I feel little for repeating
this procedure for every new FFS revision, and 32bit originating writes
are horridly slow this way.


-Albert  (hp4nl!neabbs!ajbrouw)

"After 5 days of debugging efforts, I decided to
 let it win a core-wars competition instead."