sata slower than ata ?!

[SzczechoO]

Hi!

I was playing a little with hdparm and I've found out that my sata drive with 8mb cache is actually slower than ata drive with 2 mb cache.
Here is the specification of my hardware:

abit nf7-s v 2.0
seagate barracuda vii 7200 120gb 8mb cache sata connected to sil3112
seagate barracuda vii 7200 80gb 2 mb cache ata connected to promise 20268

hdparm from sata

[Little Nemo]

You should at least connect both disks to the same controller before making comparisons. An inadequate driver might slow it down. hdparm produces the following result for my disk:

[SzczechoO]

[Little Nemo]

Okay. Of course the SATA driver for your controller might be slowing down the disk, too. If you know someone else who has a SATA controller and is willing to let you test your disk under Linux, then you could make sure it's really the disk. I still cannot believe it, though.

[srs5694]

I don't know enough about the Seagate model lines to know if you should be getting equivalent or better performance from the SATA disk; it's possible that the SATA disk is just plain not as fast as the PATA disk. I'd look into this issue before getting too worried about this. Get some spec sheets from Seagate and compare the internal data transfer rates for the two drives. (The external data transfer rates are meaningless; that's just the interface speed, which is almost certain to be much higher than the true data transfer rates.) If those rates are identical, or if they're higher for the SATA drive, then you should start looking for causes.

As you're using a Silicon Image controller, you can try switching from the ATA to the libata driver. If the Silicon Image ATA driver is built into the kernel, you'll need to rebuild your kernel, disabling it or at least building it as a module instead. You'll also need to activate the libata driver, which is in the SCSI section of the kernel configuration. Until recently, the Silicon Image libata driver was considered experimental, but I think that may have changed with 2.6.5 (I haven't actually checked with a 2.6.5 kernel, though, so I'm not positive of that).

I also recall hearing that Seagate's SATA implementation is a bit wonky. I didn't look into that in great detail, though; I just read a couple of comments and so bypassed Seagate when I bought my SATA drive. You might want to look into it further, though. Try a search for "Seagate SATA" on this forum, for instance.

[ravenqr]

My Seagate Barracuda 7200.1 80 gig hdparm's at 56 mb/s


got half of that sata drive in raid0 with a 40 gig ide drive and i get 111 mb/s with the raid0


seagate sata's are good drives...only true sata drive out there

i would check into the chipset performance for via...that could be it

[Kiroku6]

Sorry to take this off subject a little but, why do you say Seagate is the only true SATA?

[den_RDC]

I think he means that most (if not all) Seagate's SATA drives are "really" native sata (they aren't PATA drives with a convertor onboard). You can spot "converted" drives easily by the fact that the usually have a small square chip on them with a big "M" printed on (these are pata-sata bridges made by Marvell). The difference between a "converted drive" and a "native" drive is almost non-existant, but the seagate drives support the new and upcoming SATA queuing scheme that looks & works a lot like SCSI (and we all like scsi very much ofcourse). Afaik this queing scheme needs a compatible controller and driver, so it might be linux can't do this yet. But it's nice to have i guess.

PS. I am pretty shure all this is correct, but don't shoot me if it isn't
_________________
Fan of the "Survivor Warriors of the Evil Empire of Bloody Destruction and Bloody Darkness"

[eNTi]

[GentooBox]

My SATA hardisks is faster than my ATA harddisks.

SATA:

[irondog]

[Nate_S]

[srs5694]

[marcalj]

Hi, I'm now proving my new SATA HD and I get the same problem. My PATA drive is faster than SATA drive (in the same Sil3112A chipset).

[dugb]

Someone suggested to me the other day that older mother boards have the SATA controller attached to the south bridge chip via the PCI bus. Its only the newer boards that have the controller directly attached to south bridge chip. Of course nearly all boards these days have PATA as part of the south bridge or directly attached to it.

The PCI bus is clocked at 66Mhz, a SATA or PATA controller connecting directly to south bridge can be clocked faster than 66Mhz.

I admit I have not sat down and done the math yet to see if this would really make any difference.

[hw-tph]

Here's my single Seagate 160GB SATA disk on my NF7-Sr2 using libata (hence the SCSI device name):

[PrakashP]

Samsung 160GB 8MB 7200rpm PATA drvie, with PATA->SATA converter attached to SiI3112A, using libata, Abit NF7-S Rev2.0, using 2x512MB in DC mode

[marcalj]

How it works "libata" ? is configuration in the kernel? Can you especify where?

How can I know if I use libata?

Thanks.
_________________
AMD 64bits X2
porlaminima.com

[hw-tph]

libata can be found under SCSI Low Lever drivers. If you're using libata your SATA disk will be called /dev/sda since it's a SCSI device to the kernel when you use libata.

[janiskr]

mm then result of
hdparm -tT /dev/hda

[johntramp]

I have the same problem. Is there nothing that can be done about it?
I guess I will just have to get another sata drive and set up a raid array.

[hw-tph]

Do you have performance problems when using the libata driver or when using the IDE/ATA driver?

By the way - there is nothing that makes a SATA drive faster than an IDE drive. The disk, cache management and so on are the same. The only difference is the interface which the controller uses to speak to the drive. Some higher rotation speed disks are appearing as SATA only though, like there have been 10000rpm SCSI drives for years.

Using SATA-150 won't get you any better speeds than good old ATA-100 if you're using a single disk. One single disk will not (and this will probably be true for a couple of years) saturate the 100MB/s connection that ATA-100 provides. The best disk speeds I get except from SCSI is still from my Promise 20267 ATA-100 controller on my old Asus A7V using a new 8MB cache Hitachi ATA disk. It averages about 58MB/s in hdparm.


Håkan

[johntramp]

ok, so the only real performance seen out of sata is via raid then?

[rasmuslp]

[hardcore]