[svlug] Multi drive PCI SATA card or 6 port SATA motherboard

[Ivan Sergio Borgonovo]

On Mon, 13 Mar 2006 10:51:49 -0800
Rick Moen <rick at linuxmafia.com> wrote:

> Quoting Ivan Sergio Borgonovo (mail at webthatworks.it):

> > Cables may make the difference... but the mobo I just bought (asus
> > a8n-sli premium) has 8 sata II ports on board and it wasn't terribly
> > expensive (less than 200$). 
> 
> Hmm, interesting.  I see that four of those are Nvidia NForce 4, and
> four of them are SiI 3114R.  Jeff Garzik is quoted as saying that
> "Unfortunately, Nvidia is the only SATA hardware vendor that chooses not
> to give me any hardware information" -- which is sadly typical of that

Nvidia released open source drivers for agp and other stuff that now are integral part of the kernel.
nvidia should perform a bit better than sil on md.

> vendor, generally.  How well is Garzik's sata_nv driver working for you?

At the beginning when I was still understanding which distro to install and how I had some problem. Some of which made my box not bootable (I've a vague suspect that some of the problems where related to nvidia *video* driver that caused a really bad crash).
I'm using Dapper that is a testing version on a "immature" architecture (amd64 x2).
I had some apps crashing *regularly* as sylpheed-claws-gtk2 and amarok not working.
This box has 2Gb ram. At this moment I hardly had the chance to hear hd spinning. Anyway it is working from 8th March (as a desktop).

root at dark:~# hdparm -tT /dev/md0

/dev/md0:
 Timing cached reads:   2880 MB in  2.00 seconds = 1438.78 MB/sec
 Timing buffered disk reads:  360 MB in  3.01 seconds = 119.54 MB/sec

2x WDC WD1200JS-00M Version: 02.0

Let me play longer...

I've heard that nforce 4 has some problems with matrox HD. I think I still have the email reporting the problem and the specific model.

> > Performances doesn't seem particulary different between on board ctl
> > and more expensive external "true" raid ctl, expecially if he is going
> > for soft raid anyway.

> Marc expressed a quite rational preference for the Linux-native "md"
> software-RAID driver, which has generally proven faster and more
> reliable than manfacturer-specific fakeraid schemes (such as Nvidia's
> NVRAID and Silicon Image's Medley), and usually competitive in most
> respects with genuine hardware RAID, too.

What I mean is that if he is going to use md, onboard ctls work quite fast.