after last night's meeting, i decided to play with hdparm. here's my
findings:
i did 3 read buffer tests. did another 3 tests. the statistics for the 2nd
set of of 3 are:
mean: 4.507 MB/sec standard deviation: .021 MB/sec
next, i enabled some of the things we talked about last night:
/sbin/hdparm -c 1 -u 1 -d 1 -m 16 /dev/hda
it didn't let me enable the "use dma" flag. i had to recompile the kernel
and set the following options:
generic pci ide chipset support
sharing pci ide interrupt support
generic pci bus master dma support
use pci dma by default when available
intel PIIXn chipset support
PIIXn tuning support
i was still curious about speed gains, so i ran the read buffer test with
everything enabled EXCEPT the "use dma" flag.
mean: 8.32 MB/sec standard deviation: .000 MB/sec
nearly double the throughput for my hard drive! rebooted again and ran:
/sbin/hdparm -c 1 -u 1 -d 1 -m 16 /dev/hda
this time, everything was enabled and i ran the read buffer test again;
3 times, then 3 times again. the stats for the 2nd set of three are:
mean: 17.270 MB/sec standard deviation: .276 MB/sec
nearly 4 times the original throughput.
these numbers are well and good, but i gave my drive the ultimate test --
i loaded star-ing-at-the-wall office 5.2 and net-sleepwhileitload-scape 6.
can't really tell if they load 4 times faster or not, but i can say with
confidence that their load time DEFINITELY improved a great deal. we're
talking perceptively; not a shaving of a few milliseconds...
hope this helps.
peter
--
Just upgraded to Woody? Don't have permission to run X? linux
In Xwrapper.config, change allowed_users from root to console. -
------------------------------------------------------------------ ._.
Did you vote for Nader? How does it feel [EMAIL PROTECTED] /v\
to have elected a Republican president? http://www.dirac.org/p // \\
------------------------------------------------------------------ ^^ ^^
GPG Fingerprint: B9F1 6CF3 47C4 7CD8 D33E 70A9 A3B9 1945 67EA 951D rules
PGP signature
