Hello!
I've got here an i845PE-based motherboard with the usual
Marvell 88i8030 PATA->SATA bridge chip. When I connect a
SATA drive to this one, the BIOS recognizes it as an UDMA100
device, but the kernel (2.4.22) only uses UDMA33, and hdparm
also only reports support up to udma2, which results in a
performance of about 15MB/sec (hdparm -t) vs. the >50MB/sec
these drives can normally push.
I have tested this with two different drives, a fully native
SATA drive (ST3160023AS) and a pseudo-SATA drive (MX6Y120M0)
which uses the same Marvell chip internally to convert SATA
back to PATA because thats what the drive really is. The
result is that both drives show up as UDMA100 in the BIOS
and as UDMA33 in the kernel.
What is the reason for this?
Any comments/fixes out there already?
Thanks!
Regards,
Szabolcs Rumi
In a futile gesture against entropy, Rumi Szabolcs wrote:
> I have tested this with two different drives, a fully native
> SATA drive (ST3160023AS) and a pseudo-SATA drive (MX6Y120M0)
> which uses the same Marvell chip internally to convert SATA
> back to PATA because thats what the drive really is. The
> result is that both drives show up as UDMA100 in the BIOS
> and as UDMA33 in the kernel.
IIRC this is due to a problem with the cable-detection code in 2.4.22 -
it looks for an 80-conductor cable (which SATA certainly doesn't have!)
and sensibly refuses to enable modes above UDMA2 if one is not present.
I believe this is fixed in 2.4.23-pre, so the best answer is probably
either to install the latest of those or to wait for .23-final to
appear. (I have the same issue; I've opted for the latter.)
--
Adam Jones ([email protected])(http://www.yggdrasl.demon.co.uk/)
.oO("sudden vision of a klein pizza" )
PGP public key: http://www.yggdrasl.demon.co.uk/pubkey.asc