> Any suggestions for testing programs? If not, I'll run some real-world
> tests with both drivers...
Thanks. Just do it.
As you can see at the results below, a ncr driver giving 15% better
performance with my configuration cannot exist.
I am interesting in your performance problem with the 53c8xx driver and I
will be happy to track it and to fix it (if necessary).
Do all benckmarks you like to do and send me accurate informations about
conditions of tests as I asked you in my previous mail.
Thanks in advance.
Gerard.
bonnie -s 200
-------------
-------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
XPW-4kfs 200 5991 96.5 7045 27.3 3233 27.1 5650 91.9 6791 26.7 97.7 3.7
XPW-1kfs 200 5022 93.6 6823 43.1 2976 34.2 5007 87.2 6404 31.2 53.2 3.1
S12-4kfs 200 3914 60.8 4036 16.0 1922 15.0 4045 65.4 4377 15.6 66.9 2.6
S12-1kfs 200 3715 68.3 3927 22.3 1761 19.2 4030 69.5 4405 20.3 43.4 2.9
hdparm -t /dev/sdb (Atlas wide)
------------------
/dev/sdb:
Timing buffer-cache reads: 32 MB in 0.91 seconds =35.16 MB/sec
Timing buffered disk reads: 16 MB in 2.35 seconds = 6.81 MB/sec
Estimating raw driver speed: 16 MB in 1.90 seconds = 8.44 MB/sec
hdparm -t /dev/sda (IBM S12)
------------------
/dev/sda:
Timing buffer-cache reads: 32 MB in 0.90 seconds =35.56 MB/sec
Timing buffered disk reads: 16 MB in 3.69 seconds = 4.34 MB/sec
Estimating raw driver speed: 16 MB in 3.24 seconds = 4.94 MB/sec
Configuration:
--------------
P133 / Tyan Tomcat II / 512K plb / 32 MB EDO / NCR53C825 rev. 2
Quantum Atlas XP32150W / IBM S12 OEM
Linux-2.1.7 / NCR53C8XX 1.14b (patch available at linux-scsi list)
(NCR53C8XX 1.12c and 1.14a should give same results)
- tagged commands enabled, up to 12
- 25 ns transfer period factor (i.e: 10 MHz synchronous data transfer)
- disconnections allowed