Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!texbell!merch!cpe!adaptex!neese
From: neese@adaptex.UUCP
Newsgroups: comp.sys.ibm.pc
Subject: Re: Seagate ST01/02 SCSI Controllers --
Message-ID: <6100018@adaptex>
Date: 19 Oct 89 17:50:00 GMT
References: <5746@portia.Stanford.EDU>
Lines: 65
Nf-ID: #R:portia.Stanford.EDU:5746:adaptex:6100018:000:3116
Nf-From: adaptex.UUCP!neese    Oct 19 12:50:00 1989


>/* ---------- "Seagate ST01/02 SCSI Controllers --" ---------- */
>Why are Seagate SCSI controllers so much cheaper than others,
>such as WD or Adaptec?

The Seagate board, which is a derivative of an older Future Domain
design is a very general purpose SCSI adapter with a minimum amount of
intelligence.  With only 3 or 4 chips on the board, it can be made very
cheaply.  The WD/Adaptec adapters are intelligent SCSI host adapters
that really take advantage of the SCSI bus.  Requiring a lot more
IC's and firmware.

>A friend told me that the Seagate
>boards take over the bus during transfers, and that while this
>makes the boards very fast, it causes problems under multitasking
>operating systems.  Is this right?

The Seagate board is a PIO/DMA device.  It does not move data for the CPU.
The WD/Adaptec boards do.  The problem under multi-tasking operating systems
with the Seagate card is due to the lack of intelligence.  The driver in the
UNIX/XENIX kernel has to do all the work for the card.  The hardware has no
provisions for multi-threading.  This makes it very difficult to multi-thread
as there are some timing requirements that must be met in order for multi-
threading to work properly.  It is very difficult for UNIX/XENIX to do this
as they are not real-time operating systems.

>While hard disk transfers are going on, can interrupts be missed?

While PIO transfers occur the CPU can't do anything else.  If a device
such as the dumb COM ports do not get serviced quickly enough, data will
be lost as they have very small hardware buffers.

>If anyone knows about problems when using these cards under, say,
>Unix, I would like to know about them.

If I have a choice between the Seagate card and a 1:1 MFM controller,
I would rather use the MFM controller.  This is again due to the difficulty
with writting a device driver for UNIX/XENIX that has severe timing
requirements.

>It sounds like there shouldn't be trouble under Dos.

Agreed.

>Alternatively, is there much of a difference in performance between various
>SCSI cards?  I had thought that the performance would be fixed by the SCSI
>device regardless of host controller, but I'm not sure.

You are partially right.  Data transfer rates are regulated by both the
adapter and the device.  If the host adapter cannot generate ACK's/REQ's fast
enough  The adapter will limit the transfer rate possible from/to the SCSI
device.  For instance, a Conner SCSI drive would perform about the same on
the Adaptec AHA-1540A and the Seagate card.  As the drive is very limiting
on the rate it can transfer data.  On the other hand, a Quantum PRO40/80S
drive will only yeild about 1.2MBytes/sec on the Seagate card and 2.6Mbytes
on the AHA-1540.  The bottom line is, how fast do you want your SCSI to be.
You can make it as slow as a good MFM controller/drive or more than twice
as fast as a fast 1:1 ESDI/drive combination.  In the case of SCSI you
definately get what you pay for, as far as performance goes.


			Roy Neese
			Adaptec Central Field Applications Engineer
			UUCP @ {texbell,attctc}!cpe!adaptex!neese
				merch!adaptex!neese