how can this be possible? sata bad perfomance

||kafre|| · Tux's lil' helper Joined: 15 Sep 2003 Posts: 114

ummm any idea of why this sata is going soooooooooooooo slow?

Linux kafre 2.6.2 #13 Mon Apr 12 22:43:46 UTC 2004 i686 AMD Athlon(tm) XP 2600+ AuthenticAMD GNU/Linux

kafre kafre # hdparm -tT /dev/hde1

/dev/hde1:
Timing buffer-cache reads: 1736 MB in 2.00 seconds = 866.40 MB/sec
Timing buffered disk reads: 50 MB in 3.02 seconds = 16.53 MB/sec

kafre kafre # hdparm -d1X69 /dev/hde

/dev/hde:
setting using_dma to 1 (on)
setting xfermode to 69 (UltraDMA mode5)
using_dma = 1 (on)
kafre kafre # hdparm -tT /dev/hde1

/dev/hde1:
Timing buffer-cache reads: 1720 MB in 2.00 seconds = 858.41 MB/sec
Timing buffered disk reads: 50 MB in 3.04 seconds = 16.47 MB/sec
kafre kafre # hdparm -tT /dev/hda1

hda1 is a ata100 drive.

/dev/hda1:
Timing buffer-cache reads: 1736 MB in 2.00 seconds = 867.26 MB/sec
Timing buffered disk reads: 126 MB in 3.05 seconds = 41.33 MB/sec

motherboard is asus a7n8x deluxe ( i think ) nforce2 chipset, sata controller is sil3112, 512 ddr333, gentoo 1.4 if i dont remember bad.

fs in sata drive is reiserfs with notail enabled. Tried with ext3 and results were the same, a little faster. Ide drive is ext3.

what can i do? if this perfomance does not increase i'll change the diskdrive for an ata133 one.

thanks.

srs5694 · Posted: Mon Apr 12, 2004 10:11 pm Post subject:

First off, "hdparm -t" tests raw disk read performance, so changing the filesystem won't affect the results. (Any change you saw was due to random fluctuations.) It's typically run against the raw disk device, as in "hdparm -t /dev/hda" rather than "hdparm -t /dev/hda1", but the latter should work and return results that are nearly identical, if /dev/hda1 begins at the start of the disk.

Second, you may want to look into using the libata driver (in the SCSI configuration area) rather than the ATA driver (in the ATA kernel configuration area). Doing so will make your SATA disk "look like" a SCSI disk, so you'd probably access it as /dev/sda. More important, the driver is simply different, and with any luck won't run into whatever's causing sluggish performance as it's configured now. I don't know offhand if the relevant drivers are included in the Gentoo 2.4.x kernels, though. If you're not already using one, you may need to upgrade to a 2.6.x kernel. Even in most of these, the driver may be marked "experimental," although I believe that's changed with 2.6.5.

linux_girl · Apprentice Joined: 12 Sep 2003 Posts: 287

||kafre||
if you find a solution let us know

i am goging to buy a motherbord with sata+RAID :lol:

i forgot to ask you if your motherboard has RAID turned on ?

_________________

||kafre|| · Tux's lil' helper Joined: 15 Sep 2003 Posts: 114

srs5694 i'm using kernel 2.6.2 and dont see that option under scsi options

can u tell me more about where's that option? thx

pbx06 dont know now, will have a look and post it here, no problem

linux_girl · Apprentice Joined: 12 Sep 2003 Posts: 287

||kafre|| · Tux's lil' helper Joined: 15 Sep 2003 Posts: 114

i found more information, any ideas plzZz????

kafre kafre # hdparm -i /dev/hde

/dev/hde:

Model=ST3160023AS, FwRev=3.05, SerialNo=3JS1YFQS
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
BuffType=unknown, BuffSize=8192kB, MaxMultSect=16, MultSect=16
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=268435455
IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: mdma0 mdma1 mdma2
UDMA modes: udma0 udma1 udma2
AdvancedPM=no WriteCache=enabled
Drive conforms to: ATA/ATAPI-6 T13 1410D revision 2:

* signifies the current active mode

kafre kafre # hdparm -i /dev/hda

/dev/hda:

Model=SAMSUNG SP6003H, FwRev=QV100-60, SerialNo=0441J1FT918313
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
RawCHS=16383/16/63, TrkSize=34902, SectSize=554, ECCbytes=4
BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=16
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=117304992
IORDY=yes, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: mdma0 mdma1 mdma2
UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5
AdvancedPM=no WriteCache=enabled
Drive conforms to: ATA/ATAPI-6 T13 1410D revision 1:

* signifies the current active mode

why is my sata drive only udma2 capable? i'm a bit lost.

||kafre|| · Tux's lil' helper Joined: 15 Sep 2003 Posts: 114

thank u all, a new topic is opened with goooood results :lol:

srs5694 · Posted: Tue Apr 13, 2004 12:58 am Post subject:

pbx06 posted the correct part of the kernel configuration for the libata driver. It's under SCSI, under the SCSI hardware subsection when you do "make menuconfig" or "make xconfig". You'd set the CONFIG_SCSI_SATA_SIL option. As I said, you need to tell the system to accept experimental drivers, at least if you're using a pre-2.6.5 kernel (and maybe that one, too; I haven't checked it myself). It's reportedly experimental for a reason, but I see a lot of posts from people who don't seem to have major problems with it. (I've got a different SATA controller, myself, so I can't comment from personal experience.)

pbx06, you should not be trying to enable RAID in the BIOS. The Silicon Image chipset doesn't provide true hardware RAID; it's a software RAID that requires OS drivers (the BIOS just helps out a little, but I don't know the details). Those drivers don't officially exist for Linux, although IIRC, there's some work to support the protocol used. The consensus I've seen is that if you want to use RAID under Linux with this hardware, you should disable it in the BIOS and use Linux's own software RAID features.

As to the output you posted, ||kafre||, it looks like the driver just isn't using the higher UDMA modes of the controller and/or disk. That model number looks like a Seagate model number. If so, I seem to recall hearing something about Seagate's SATA implementation being a bit wacky, and some drivers deliberately cut its speed down to avoid problems. I don't recall the details, though, so I recommend you do a search on "Seagate SATA" or some such to find the details. There may be a workaround, but if not, you may need to replace the drive.

tomk · Posted: Tue Apr 13, 2004 8:13 am Post subject:

hybrid · Posted: Tue Apr 13, 2004 10:51 am Post subject:

hi all,
there is an other solution, to this problem since i have the same SIImage SATA chipset, i found this solution on this forum and it was posted long ago by someone called Heretic (if i remember) thanks to him i get a very high speeds with my sata disk :
in you kernel sources go check for the file drivers/ide/pci/siimage.c and fo check for the function called siimage_tuneproc you'll find a variable declaration:
u32 speedt = 0;
to
u16 speedt = 0;
this is the bug in the siimage driver also you need to set hwif->rqsize = 128; instead of hwif->rqsize = 15; if this haven't been set yet

you can find his complete patch here : http://www.kernel.org/pub/linux/kernel/people/bart/
of course you need to merge the differences manually since this patch is destined to for the 2.6.0-test11

(but the bugs are still here)
here are my tests after the patch :