Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!swbatl!texbell!merch!cpe!adaptex!neese
From: neese@adaptex.UUCP
Newsgroups: comp.unix.i386
Subject: Re: Adaptec 1542 performance
Message-ID: <6700033@adaptex>
Date: 19 Oct 89 18:26:00 GMT
References: <29@van-bc.UUCP>
Lines: 28
Nf-ID: #R:van-bc.UUCP:29:adaptex:6700033:000:1520
Nf-From: adaptex.UUCP!neese    Oct 19 13:26:00 1989


Using 'dd' to measure throughput will not give you much idea of the real
overall benefits of the 1542A.
The real benefit of the 1542A is realized when you measure the overall
load on the AT bus or how much time the CPU spends messing around with
disk I/O instead of doing what it should be doing,....spending time on user
processes.  Try starting a 9600 buad communication process on a dumb COM
port and then start up some severe disk I/O.  You will find that MFM/RLL/ESDI
controller/drive combinations will cause occasional character loss, while
the 1540A in the same situation will not.
The Miniscribe drive is also limiting the speed at which things will occur.
It is an unfortunate fact that the best SCSI adapter you can use will be
directly impacted by the SCSI device you chose.  The best devices I have
found to date are the Quantum PRO40/80, CDC (any Wren after III, with read
ahead enabled), and Maxtor XT series of SCSI drives w/read ahead enabled.
You may also want to patch 'ad_buson' and 'ad_busoff'.  They are currently
set to 5 on and 7 off.  I recommend 6 on and 8 off in conjunction with
the setting to 6.7MBytes/sec data rates.  You should also install the
J1 pin pair 1 jumper to enable synchrounous transfers.
I also do not recommend the Seagate, Rodime, Conner, Priam, Miniscribe,
and Micropolis.  Not for reliability, but for performance and overall
SCSI bus usage.


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