disk drive hell (aka/etc/hdinit)

[Jon Tara]

In article <496 at megatest.UUCP>, palowoda at megatest.UUCP (Bob Palowoda) writes:
> 
>    Interesting my DTC controller seems to be slow also. I started using
>    a benchmark program Coretest. It on a Vertex V150 it reports 280K/Bytes/sec,
>    23ms random track access, and 6ms track to track access, under MSDOS.
>    I just ran the same benchmark on a friends PC with and OMTI RLL controller
>    with a CDC and it reports 720K/Bytes/sec 23ms and 4ms. Opps forgot to mention   my DTC is also RLL. The OMTI could format 1 to 1. That's about 5x increase
>    in data throughput at approximately the same price. Does anyone have any
>    benchmark tests to measure disk IO on SCO Xenix? Also has anyone used 
>    RLL with SCO Xenix? If so what controllers and/or drives. Any help would
>    be appreciated.

I have an OMTI 8627, which does ST506 RLL, as well as ESDI.  With
a Vertex V170 in an AT at 9.5 mHz, Coretest gives it 658k/sec and
28 msec.  With the "high speed I/O" jumper off, it's much lower -
maybe 250k.

I upgraded to a 386, and a CDC Wren III ESDI drive.  Much to my
horror, throughput went DOWN!  The best I can do is about 500K/sec,
with 17msec access.  Initially, I could only get the thing to
run on the 386 by turning the high speed I/O off (that's done by
INSTALLING a jumper...) or by slowing down the AT bus to 1/3
the 20 mHz processor speed, instead of 1/2.  Eventually, I discovered
that it would work reliably with a 10mHz bus by turning on the
Chips & Technologies 82C206 'EMR Bit' option.
  
Loading the 8627 BIOS extension into 32-bit RAM improves it to
about 550K/sec, but works only sporadically, when wait states
are added to main memory - apparently there's some time-critical
code.
  
The theoretical maximum would be about 1.1MB/sec (36 sectors x
512 x 60 rev/sec.).  The theretical maximum with RLL is 798k/sec,
so the controller's getting just about all there is to get for
RLL, but doing a pretty poor job with ESDI.
   
So, what does this have to do with Xenix?  Well, SCO recently came
out with a seperate version, called '386ESDI' (or 286ESDI) which
supports this board.  It supposedly has additional instructions to
take advantage of ESDI drives (the OMTI, that is.)  But OMTI
has discountined the 8627, and is now coming out with seperate
models supporting 1:1 for MFM, RLL, and ESDI, but no more combined
models.  And guess what?  They're all Western Digital compatible,
so they'll need the NORMAL version of Xenix...  They're claiming
much better performance from the yet-to-be-released ESDI controller.
  
Does anyone know of an ESDI controller, available now and that will
work with Xenix that can get me close to that magic 1.1MB/sec?


-- 
  jtara%m-net at umix.cc.umich.edu          ihnp4!dwon!m-net!jtara

 "If this is all a dream, I can't wait to see what it's like when I
  wake up."  - Explorers