Date: Wed, 17 Mar 1999 17:01:17 -0500 From: Andrew Heybey <ath@niksun.com> To: freebsd-scsi@freebsd.org Subject: NCR 895 performance Message-ID: <199903172201.RAA02671@stiegl.niksun.com>
next in thread | raw e-mail | index | archive | help
I have a machine running 3.1-RELEASE. It has three IBM DDRV 9LZX (10000RPM
Ultra2) disks and a CCD partition striped across them.
If I use an adaptec 7890 controller, bonnie reports:
-------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
AHC 512 22432 96.3 33176 42.9 11995 18.6 21094 92.3 33897 28.4 403.3 4.2
If I use a Tekram 390U2W (which uses the NCR/SYMBIOS 895 chip),
bonnie reports:
-------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
NCR 512 21438 91.4 25026 32.8 10405 16.4 18871 83.4 25415 21.2 353.1 3.7
(after tweaking ncr.c to allow 40MHz transfers; before (with default
20Mhz transfers) I was only getting 20MB/s).
Why is the ncr so much slower than the ahc? Is there something
fundamental about the controller that prevents it from going any
faster? Or is the driver not running the controller optimally? Is it
because the NCR is limited to 32 tags (so sayeth the source)? (The
ahc uses 64 tags (apparently the max that my disks can handle).)
andrew
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903172201.RAA02671>