On Saturday 27 January 2001 16:43, you wrote:
> Dear Civileme, I am trying to understand your message (remember I am a
> newbie).
> Lets me comment step by step your message:
> 1) fdisk + mkfs (create partitions and file system for linux)
> 2) ide0=noautotune (I suppose is an aditional command on grub or lilo,
> isn't it?
> 2) hdparm -c1 -d1 -u1 -X68 /dev/hda; hdparm -c1 -d1 -u1 -X69 /dev/hdb
> These are comands, bu I don't understand the coments you did:
>      -use hdparm -t after each setting and start dead slow,like
> hdpar         -c1 -d1 -X66 /eachdev. Does this mean:
>       a) hdparm -c1 d1 -X64 /dev/hda; hdparm -c d1 -X65 /dev/hda..
> X66            ??? and for drive hdb continue until X68.... (I suppose
> the new          hd is hdb)??
>       b) and after reached X66 for hda and X68 for hdb: hdparm
> -t                 /dev/hda; hdparm -t /dev/hdb?
> 3) I don't understand the hparm -c1 -d1 -u1 -X64+x /dev/hdb; hdpartm -t
> /dev/hdb (you said that this produces performance of Y Mb/s!!!!); also
> the hdparm -c1 -d1 -u1 -X64+x+1 /dev/hdb (X64+x+1 mens X69 for drive
> hdb??????); hdparm -t /dev/hdb.
>
> 4) You said that when an increment X64+x+1 just produces and increment
> of +/- 5% you mst use the x setting (64+x settin??)
>
> O dear, I am absolutely lost!!!!
>
> Please, when you had time, explain me step by step.
>
> Thanks a lot for your kind attention, yours sincerely
>
>
> Francisco Alcaraz
> Murcia (Spain)
Yes they are grub/lilo appends--substitute the right channel(s) for your 
controller(s) a,b are on 0  c, d on 1 e,f on 2 g,h on 3 and if you have 4 or 
higher you have to activate after boot because 4 are all that are probed.


OK
Once you get things running, with the noautotune, you will be at less than 
optimal disk performance.

first, see if you can get 32 bit i/o enabled on each disk

hdparm -c1 /dev/hda   # I am assuming your drives are a and b though they 
could be e and g for all I know

and the same on /dev/hdb

Then once this happens, if you didn't lock up

hdparm -t /dev/hda #will time a read of 64Mb on a and report Mb/s

Then we try hdparm -c1 -u1 /dev/hda

and test the speed again

and hdparm -c1 -u1 -d1 /dev/hda

and test the speed.

Now 

hdparm -X66 /devhda
hdparm -t /dev/hda

should work and establish a baseline of data rate

so try 

hdparm -X67 /dev/hda
hdparm -t /dev/hda

is it faster? If you think you got a speed increase, then try 

hdparm -X68 /dev/hda
hdparm -t /dev/hda

Is it at least 5% faster?  If not, drop back to 67 and work on the other 
drive.

the 64+x is 64 for udma plus x for mode.  Since we can figure your disk runs 
at udma 33 or mode 2 for udma, then we can start with x=2 or hdparm -X66.  
When the speed curve flattens out, you have all the controller/disk/kernel 
combo will give you.  Of course you do have the better drives, so you should 
be able to tune them pretty well.


Seagates sometimes have trouble with VIA chipsets (a pure hardware interrupt 
routing problem).  I have also observed MASSIVE data corruption with Cyrix 
6x86 processors and SiS 5571 chipsets and linux and Seagate drives, though I 
was never able to reduce the variables and locate the exact cause.

I hope this helps

Civileme

This set and test and compare procedure should tune your drives for your 
system.  This is superior to what the autotune does and it is also reasonably 
conservative for data integrity.



Reply via email to