hdparm -tT /dev/sda2

/dev/sda2:
Timing buffer-cache reads: 128 MB in 0.36 seconds =355.56 MB/sec
Timing buffered disk reads: 64 MB in 1.15 seconds = 55.65 MB/sec

Thats with RAID 1

When just using the driver on the RedHat 9.0 CD redhat shows two drives instead of only one - thats the reason i use the silicon driver - cause it shows correctly only one drive (RAID 1 = mirror)

Nope I'm not using RAID (RAID 0).

cool - so works on just a controller too

To get get good sata in linux u need 2.4.21 or newer.
see the kernel.org change log.
I run gentoo & 2.6.0 test4 on my a7n8x dx.

what does reveal?

hdparm -tT /dev/hdx is a hard drive benchmark

yeah exactly..

I'm interested in the results your'e getting from 2.6.0 test4 kernel..

can u post them?

20G ibm desk star 7200rpm I think it is a ata100 but it my be ata66 being only a 20g drive.
Timing buffer-cache reads: 128 MB in 0.27 seconds =474.07 MB/sec
Timing buffered disk reads: 64 MB in 1.67 seconds = 38.32 MB/sec

Main drive is building abiword.

k...

I think the kernel driver is restricted to ata66, even if u have an ata100 drive..

I don't have a pc at home at, but I can check it sometime..

cheers

taskara

That is not true see this

http://www.linux-ide.org/chipsets.html

A friend of mine and I discovered yesterday one thing: The supplied SATA driver (for Silicon Image 3112a chipset, anyway) locks the system up (on an Nforce A7N8X Deluxe PCB 2.0 motherboard) depending on how much activity the SATA drive has to do. The discovery was made in the following manner:

We tried to copy from his computer to mine (his has the SATA) a bunch of .wav files (a total of 790 Mb, a whole custom-made Audio CD) via Network connection (SAMBA) The problem arose when the copy process tried to copy one of the largest files (117 Mb) the first time. Then the system froze again when tring to copy the rest of files that did not copy. The strange thing is that if we copied the files one at a time (letting the drive around 5 seconds of inactivity) the computer never froze!

Our theory: For some strange reason (niether of us is an electronic engineer, nor a systems programmer) we suppose the drive(r) fails when reading from the internal disk caché, provoking a Segmentation fault on the driver, problem is the driver is compiled into the kernel, causing the system to freeze. This problem was seen when the reading stress on the disk was high (like when transferring 0.78 Gb). Strangely enough the problem is not obseverd (up until now) while writing on the disk, just when reading.

Can anybody else confirm this??

yeah.. this is what I was saying.. the current driver is not good... although it is an ata150 controller, linux support is limited to ata66

Well, according to Alan Cox anyway.

I'm not sure the situation has improved.

enyawix, can u run [quote]hdparm -i /dev/hdx[.quote] and and post the results?

Thetargos, this is the problem everyone has been having. there was a work around to do with max kb requests, which I think is posted in here somewhere.

I have yet to see the kernel silicon image driver working properly.

I don't know if your question is directed to me, but anyways I'll post what I have (well, my friend's machine ):

hdparm -i /dev/hde reports supported ATA UDMA modes from UDMA 2-UDMA 6, current setting at 6 (ATA 150?)
hdparm /dev/hde only reports PIO @ 16-bit and DMA 1 (on), the rest is off.

I forgot to mention that in this machine, there are two Seagate Barracuda SATA drives, one of 120 Gb and the other 80 Gb. Strangely enough the secondary SATA drive (80 Gb, /dev/hdg) is detected as if it was only using udma2 (the other modes are not detected). The RAID array is set to RAID 0 (no RAID)... But I wonder: Why is the drive using only udma2 on this drive and 6 on the primary (/dev/hde)?

BTW @ nforcershq.com, there is the problem reported some work arounds (or at least one that may help) is to set the IDE chipset (any drive connected to the primary IDE controller) to UDMA 4 (default setting is set to 5 [ATA 100]). We experienced some strange errors while trying out this setting: From within the system I could set the udma mode to 4, so I thought (and many at nfrocershq did also) to put the parameter in /etc/rc.d/rc.local. The problem was that the drive (hda, where the Linux system is installed) was giving up, complainting about an IRQ error, entering an endless loop. So if the only mode supported by this chipset driver (currently) is ATA 66, theoretically it could be set during boot up to use udma 3, right? I have yet to try this, but this could be a solution for the lock ups... in the mean time.

In other news, Silicon Image recently released some drivers, which I've seen are pre-compiled for Red Hat (the distro we use) 9, kernel 2.4.20-8 (i.e. The boxed one) and anounced that they are not going to release any sources for the drivers to the community (or so I understood). My question here is how is ever going to be the support for this chipset under linux good, if they at least don't release some technical papers so the community (in particular the linux-ide project) can get to work on the driver?. Anyway... Has anybody been successful to set an hdparm option to abolish these lock ups? Is this idea of mine to set the udma mode to ATA 66 (udma 3) any good or reasonable at least? Or I got everything mixed up?

2.6.0test5 is out
when i get it built i will post what i get

I recently got a new pC

I have a bit of understanding in linux. but sheez it took me long enough to realise that winxp doesnt have native support for sata. so i had to load up driver during install in order for it work.

im using teh silicon image chipset. but using amd and gigabyte motherboard.
forgot which motherboard extactly ;/ ill hafta check dat when i git back home
;/
*loves his trusty manuals*

but im wondering, will it be easy to install redhat9?
i maen sheez everyone seems to ahve problems
an *coughs nervously* i think i have a wd hdd. not t00 sure th0
ill hafta l00k.

i only got it yesterday ;p
gimme sum credit for not remembering :P
*i was fiddling with bios for 2 hours trynna udnerstand wtf sata is*

anyways. does any1 have a gigabyte mobo with silicon sata? oh an im using sata as base. not raid.

thanks

-seps-