[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: hdparm



<tko@westgac3.dragon.com> writes:
| Martin Oldfield writes:
| > 
| > 
| > I'd like to improve the IDE performance of my system. The IDE
| > controllers are on a newish Intel motherboard; /proc/pci says:
| > 
| > IDE interface: Intel 82371AB 430TX PIIX4 (rev 1).
| > 
| > The drives are older:
| > 
| >  Model=QUANTUM FIREBALL ST6.4A, FwRev=A0F.0800, SerialNo=15672304
| >  Model=QUANTUM FIREBALL_TM3840A, FwRev=A6B.1T00, SerialNo=39662361
| >  Model=ST32140A, FwRev=08.08.01, SerialNo=JBF24417
| >  Model=ST32140A, FwRev=08.08.01, SerialNo=JB770285
| > 
| > Can anyone suggest more aggressive (yet safe!) options for hdparm to
| > make things run more quickly; alternatively is there a repository of
| > known good settings.
| 
| Here's a script I added to /etc/init.d (with link in /etc/rc2.d) for better
| performance. Use the '-i' option alone to find out the number for the '-m'
| option. (man page explains all) This script is the last thing executed during
| bootup. Enjoy! Oh BTW, my transfers jump from 5 Mb/sec to 35Mb/sec.

You're saying you get 35 mega Bytes per second? That seems highly
unlikely!

I'm not disputing the fact that your script might improve performance,
haven't tried it, but there's not a hard drive in existence, excluding
specialty solid state drives and RAIDs, that can sustain 35MB/s (I'm
assuming by Mb you meant Mega Bytes and not Mega bits, which is what
Mb is generally used for?). Shoot, I don't even think the UDMA bus can
acheive that? I believe it's theoretical maximum is 33MB/s. Whatever
you're using to get this performance number isn't measuring your disk
throughput but your cache performance.

Of course, if you meant what you wrote and get 35 mega bits/sec I
could believe that, although if the drive was getting 5Mb/s to start
with it's time for a new drive!

Try using

hdparam -t -T /dev/<whatever>

for a little better estimate.

Gary


--  
Unsubscribe?  mail -s unsubscribe debian-user-request@lists.debian.org < /dev/null


Reply to: