Parity error using AHA1542B

[Bill Norcott]

In article <133 at chansw.UUCP>, chan at chansw.UUCP (Jerry H. Chan) writes...
>Anybody out there ever see a problem with parity errors while running
>the Adaptec 1542B SCSI controller? We had some problems a while back
>on some 386 motherboards -- systems would take a parity error during
>high disk I/O activity; the fix was to use a different HD controller
>(i.e., Future Domain). Now my 486 box is exhibiting the same behavior...???
> 
The Adaptec 1542B has much better performance than the Future Domain 860 & 885
which are not bus mastering.  I have run CORETEST on DOS with a very slow
Seagate ST296N  84 MB drive, and increased transfer rate from 438 KB/sec
to 663 KB/sec by dumping the Future Domain in favor of the Adaptec.  Using
the faster Seagate (CDC) Wren drives CORETEST cranks out 1.3 MB/sec
asynchronous, and 2.2 MB sec synchronous mode.

By the way I am running a 20 MHz 386 with AMI BIOS and Chips & Technologies
NEAT chipset, so I can program bus & DMA speeds, etc during setup.

Make SURE the jumper settings for the Adaptect parity enable/disable match
those on your SCSI device.  Try using parity disabled on both.

>Might this somehow be related to BUSON and BUSOFF parameters? I am now
>using the default parameters a la ISC, and have dropped the DMA rate down
>to 5mb/sec. Might the Adaptec be violating the 15 usec refresh rule on memory
>during bus mastering cycles?
> 
I can (running DOS) crank the bus up to 10 MHz transfer rate and also on the
Adaptec, with no problems.  The bus on/off times are programmable on the 
Adaptec (with their driver in DOS, and I think also via jumpers).  The
DMA rate is settable by both jumpers and their ASPI4DOS driver.

You can test the various DMA rates from DOS DEBUG by entering the following
command at the debug prompt:  g=dc00:9 
(assuming that your BIOS address for the board is DCOO, the default)

This runs a diagnostic to test the onboard DMA at the currently selected 
speed.  If I can run full blast on my 386 your 486 should work 

>BTW -- No amount of memory testing seems to be able to detect bad memory.

The controllers BIOS diagnostic will detect snags in the DMA test, 
usually by hanging if the board is cranked up too fast.

> 
>??? Help ???
> 
>-- 
>Jerry Chan 508-853-0747, Fax 508-853-2262  |"My views necessarily reflect the
>Chan Smart!Ware Computer Services & Prods  | views of the Company because
>Worcester, MA 01606                        | I *am* the Company." :-)
>{m2c,jjmhome}!chansw!chan                  \---------------------------------


Regards,

Bill Norcott
Digital Equipment Corporation