From owner-freebsd-hardware  Thu Aug 22 05:51:21 1996
Return-Path: owner-hardware
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id FAA11435
          for hardware-outgoing; Thu, 22 Aug 1996 05:51:21 -0700 (PDT)
Received: from tbd.gfoster.com (dyna237.intr.net [204.157.123.237])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id FAA11429
          for <freebsd-hardware@FreeBSD.ORG>; Thu, 22 Aug 1996 05:51:16 -0700 (PDT)
Received: (from gfoster@localhost) by tbd.gfoster.com (8.7.5/8.6.12) id IAA02195; Thu, 22 Aug 1996 08:50:50 -0400 (EDT)
Date: Thu, 22 Aug 1996 08:50:50 -0400 (EDT)
From: Glen Foster <gfoster@gfoster.com>
Message-Id: <199608221250.IAA02195@tbd.gfoster.com>
To: freebsd-hardware@FreeBSD.ORG
In-reply-to: <199608220336.AA169064976@hpnmhjw.sr.hp.com> (message from Darryl Okahata on Wed, 21 Aug 1996 20:36:15 -0700)
Subject: Re: SCSI again: Asus SC200 vs. Adaptec 2940 ... 
Sender: owner-hardware@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

A couple of data points with Adaptec 2940s:

***** Bonnie:
              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
P166      100  1414 23.8  1356  5.1   700  4.3  1736 27.5  2081  7.7  42.7  1.9

***** IOZONE:
Writing the 128 Megabyte file, 'iozone.tmp'...96.320312 seconds
Reading the file...66.140625 seconds

IOZONE performance measurements:
        1393451 bytes/second for writing the file
        2029278 bytes/second for reading the file

Machine is a P166, with 64MB 60ns parity DRAM, ASUS P55T2P4 w/256K PB
cache, a 1GB Fujitsu M2266S-512 002C, and an Adaptec 2940 SCSI
controller running 2.1.5R.  Disk was about 3/4 full at start of test.
This drive is an older 5 1/4" 3600 RPM drive that is not doing
syncronous transfers, if I get a wild hair I'll flop some jumpers and
see if sync xfers can be enabled.

***** Bonnie:
              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
P100      100  2056 51.4  4226 19.3  1609 12.9  2076 50.7  4291 19.7  72.1  3.9

***** IOZONE:
Writing the 256 Megabyte file, 'iozone.tmp'...96.984375 seconds
Reading the file...63.914062 seconds

IOZONE performance measurements:
        2767821 bytes/second for writing the file
        4199943 bytes/second for reading the file

Machine is a P100, with 32MB 70ns non-parity DRAM, unknown MB w/256K
PB cache, a 2GB Seagate ST32430N 0300, and an Adaptec 2940 running
2.1.5R.  Disk was about 1/2 full at start of test.

Without trying to get too deep into the numbers, I think one could say
that one's money is more effectively spent on disk, RAM, and CPU than
on a "better" SCSI controller.

>Date: Wed, 21 Aug 1996 20:36:15 -0700
>From: Darryl Okahata <darrylo@hpnmhjw.sr.hp.com>
>
>> How do these two boards actually compare?  Disregarding drivers
>> are they fairly similar in capability and speed?
>
>     The NCR driver is quite nice.  It's certainly hard to go wrong for
>the money.  Here are some benchmarks, using an NCR815-based controller:
>
>-------------------------------------------------------------------------------
>***** Bonnie:
>              -------Sequential Output-------- ---Sequential Input-- --Random--
>              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
>Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
>PCI/Q4.0  256  3350 66.8  3387 14.8  2027 15.9  4671 90.0  6065 24.6  87.5  4.1
>
>***** IOZONE:
>$ time iozone 256
>
>Writing the 256 Megabyte file, 'iozone.tmp'...77.492188 seconds
>Reading the file...43.460938 seconds
>
>IOZONE performance measurements:
>        3464032 bytes/second for writing the file
>        6176476 bytes/second for reading the file
>-------------------------------------------------------------------------------
>
>This is on a P133, with 64MB 60ns parity FPM DRAM, ASUS P55T2P4 w/512K
>PB cache, a 4GB Quantum Atlas (34300S), and a DTC 3130B SCSI controller.
>Note that these measurements were *NOT* done on the outer tracks of the
>drive; the drive was half-full when these benchmarks were done, and so
>I'd guess that they were done around the tracks 1/3 of the way in.