Promise Ultra ATA 66 controller vs. FC1

[Robert D. Arendt]

Alexander Dalloz wrote:
> Am Fr, den 23.01.2004 schrieb Robert D. Arendt um 19:18:
> 
>>Steve Saady wrote:
>>
>>>I have a Promise Ultra 66 eide controller, but think it is only being 
>>>acessed at 33 Mhz.  I have looked around, but have not found how to 
>>>ascertain what rate Fedora is using it at.
>>>Is there a way to configure FC1 to use it at 66MHz?
>>>
>>
>><snip>
>>
>>>dmesg|more :
>>>ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
>>>
>>
>>I added idebus=66 to /etc/grub.conf, so my boot line now looks like:
>>   kernel /vmlinuz-2.6.1-1.53smp ro root=LABEL=/ apm=off idebus=66
>>should work with the nominal FC1 kernel as well.  Really boosted my
>>disk I/O speed.
>>
>>Hope this helps,
>>-Bob Arendt
> 
> 
> What kind of hardware are you running - no PC ware or even very old EISA
> bus hardware?
> 
> idebus=66 is nonsense and might even be dangerous on normal PC hardware,
> as simply the bus frequency is 33 MHz and nothing else!
> 
> How did you measure the speed boost? Please compare the speed at
> idebus=33 and with your setting of idebus=66 and post your values and
> how you measured the I/O speed values.
> 
> Alexander
> 
Alexander - you're right, I'm wrong.

Hardware: ASUS P4P800 motherboard (865PE+ICH5), 512 Mbyte RAM, 3.0G P4
hda: WDC WD800JB-00ETA0, ATA DISK drive

I used a bash scriptlet to write 1 Gbyte files to the test disk

rm junk*
for n in 1 2 3 4 ; do time dd if=/dev/zero of=junk$n bs=8192K count=128; done

The first time is quicker, due to the OS using available RAM as cache.
Subsequent files should get the sustained write speed.  I tried repeating
my tests in single-user boot, and got 20.0 sec/Gbyte, *regardless* of
idebus= setting.  Now I believe the idebus parameter is irrelevant,
since the drive is using the udma5 mode, not a PIO mode.

First time I tried this, I saw 30 sec vs. 21 sec;  There must have been
some background activity that skewed results;  I should have been more
careful.

-Bob Arendt