[rommel]

lol....well is that then the case....say scsi 320u on a pci 66mhz/64bit bus would be the the better performer....i am using LSR (linux software raid 0) with with 160u cheetahs now.

[miguelbarroso]

By what I've read, in some new nForce2 boards (eg.ASUS), the SerialATA controler is not connected to the PCI Bus, so SerialATA RAID in these boards will probably outperform normal PATA RAID controlers...

just my 2 cents...

M.

[docinthebox]

Quote

The PCI bus was moved from the northbridge to the southbridge, and thus the ATA controllers no longer have to share PCI bus bandwidth.

You mean the PCI bus is moved from the southbridge to the northbridge, right? (not the other way round)

True, if you use a separate hub link / v link to connect each PCI-X bus to the northbridge, then the IDE controller connected to the PCI-X bus is just a counterpart to the southbridge; they all talk directly to the northbridge via their own hub link. But right now this architecture is only found in server chipsets like the Intel E7500 chipset for the Xeon. I don't think we'll see a desktop chipset with this architecture for quite a while.

For the upcoming Springdale chipset (Q2 2003), we'll see the SATA controller built into the southbridge (ICH5), like what Jan said. Not as good as a separate hub link for the controller but at least now the controller will have access to the bandwidth of the NB-SB interconnect and not be restricted by the bandwidth of the PCI bus. At this point, SATA RAID will be faster than PATA RAID but only because no one is making a PATA RAID controller integrated into the southbridge.

[docinthebox]

Quote

By what I've read, in some new nForce2 boards (eg.ASUS), the SerialATA controler is not connected to the PCI Bus, so SerialATA RAID in these boards will probably outperform normal PATA RAID controlers... 

just my 2 cents...

M.

From what I've read, the Asus A7N8X deluxe uses a Silicon Image SATA controller hooked up to the PCI bus.

Maybe you're talking about the Abit NF7-S. The spec says it uses an "on board Serial ATA PCI controller" but doesn't provide details. If it's anything similar to Abit's AT7-MAX2, it probably uses Marvell SATA bridges that possibly hook up to regular PATA channels in the southbridge. If this is the case, it could be faster than a PATA RAID controller (not integrated into the southbridge).

[rommel]

i appreciate all of your input but if i wanted this thread...yeah the one i started to be a discussion of the benefits of sata over pata then thats what i would have asked....thanks for nothing

[Jan Kivar]

Sorry about that... Let's fix it:

Todays SATA drives are nothing more than PATA drives with a converter. So basically the performance is similiar to PATA drives and PATA RAID. You can compare the drives in Storagereviews database, even to SCSI drives. Quick reference: 15krpm SCSI drives are fastest, but they cost much more than IDE drives.

The only usable difference between SATA and PATA is four independent channels, if one has four SATA drives and using them in RAID0. That would have greater performance than two channel PATA RAID0.

The "problem" with four drive RAID0 is that standard 32b/33MHz PCI bus bandwidth limits the throughput of the array. It has maximum transfer rate of 132MB/s, and 100+MB/s sustained. This is problem with U160/U320 devices also. You can surely put one 15k.3 to the U160 channel and get the max out of it, but slap in two drives (in HW/SW RAID0 for concurrent access) and You saturate the U160 channel. But doing so You have already capped the PCI bus. U320 helps from drives to SCSI card, but the PCI bus is holding You down.

So what You need here is a motherboard and controller cards with something better than a 32b/33MHz PCI bus in order to see the difference.

Cheers,

Jan

[rommel]

thank you jan...and i agree....waiting for hammer though i think to up from 160u and a mainboard with 66mhz/64bit bus