Higher ATA-Mode -> lower speed

[Emanuel Strobl]

Am Donnerstag, 20. Januar 2005 20:00 schrieb Emanuel Strobl:
> Am Donnerstag, 20. Januar 2005 19:48 schrieb Emanuel Strobl:
> > Dear ata interested guys,
> >
> > I observed a strange behaviour which seems to explain my often noticed
> > "16MB/s hard-limit"
> >
> > I have a UDMA133 drive (MAXTOR 6L060J3) which saturates at 16MB/s when I
> > dump anything to it, regardless of the block size. This transfer rate is
> > reached with bs=4k and doesn't increase any more even not with bs=64k.
> >
> > Now, when I set the mode to UDMA100 I get well over 40MB/s!!!!
> >
> > Can anybody confirm that for different hw? Especially people like fandino
> > who already discussed poor ata performance on -current (~16 Oct. 04).
>
> I can confirm that for my workstation here. i815 chipset and seagate
> ST380011A.
> When set to UDMA100 (the maximum) the following dump gives 16MB/s 'dd
> if=/dev/zero of=/usr/testfile bs=16k count=2000).
> When I limit the mode to UDMA66 I get 53MB/s !!!!!!!!!!!!!!!!

Sorry, that's not 100% correct.

After booting the machine I get 16MB/s. That's the hard limit.
The disk is shown to operate with UDMA100 (that's the highest mode of the 
chipset).
Now when I issue a 'atacontrol mode 0 udma100 udma100' (set the same mode as 
it's said to be) I get 57MB/s!!!!!
From this point it doesn't matter what mode I set, the 16MB/s hard-limit 
vanished :))

But the original described problem of the 16MB/s hard-limit with udma133 is 
reproducable, means if I return mode to udma133 I see the 16MB/s limit again!

-Harry

>
> It seems that's independent of UDMA133 instead it's when mode is set to the
> maximum the chipset can handle!
>
> This really hurts! But it's _the_ eplanation for all the "poor ata
> performance" reports.
>
> -Harry
>
> > Why does UDMA133 mode limit the transfer speed so badly? And why do I get
> > significantly slower transfer rates (32MB/s insted of 42MB/s) when I set
> > the mode to UDMA66 (compared to UDMA100 but twice the speed of UDMA133)?
> > There's only one device on the channel, so UDMA66 should be fine for
> > 42MB/s. The controller is a HPT372.
> >
> > Best regards,
> >
> > -Harry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20050120/5155dfd0/attachment.bin