overcal systemperformance impact from using PCI controller card vs onboard controller

[theodork]

I have three hard drives and two optical drives. I have a shuttle an35 ultra (nforce 2 based) and my PCI controller card is an Abit Hot Rod 100, based on the Highpoint 370 chipset.

Westerrn Digital 80gb 8mb 7200k Primary Master
Western Digital 80gb 8mb 7200k Primary Master (pci card)
Western Digital 80gb 2mb 7200k Secondary Master (pci card)
LiteOn 52/24/52 CDRW Secondary Maser
LiteOn 16x DVD rom Secondary Slave

I'm thinking of getting rid of one of my hard drives (i just noticed that one whines, and now that i've noticed it, its started to get to me), and when i do so, i was wondering if i would suffer a performance hit from using my primary drive off of the PCI card rather than using the onboard controller. Mainly im thinking about stuff like extra PCI bandwidth being used and the CPU having extra load because of it and stuff. (as of right now the drives connected to the pci card have mostly been used for storage, not as my main apps drive so i dont know if i would get performance hit from heavy bandwidth)

In other words, would the following setup hurt me more than my current?
Westerrn Digital 80gb 8mb 7200k Primary Master (pci card)
Western Digital 80gb 8mb 7200k Secondary Master (pci card)
LiteOn 52/24/52 CDRW Primary Maser
LiteOn 16x DVD rom Secondary Master

 

[Snugglebear]

There are usually three problems you can run into by doing this:

• The controllers may fight during POST if their BIOS order is screwed up or if their BIOS is written improperly (e.g. one may load and stay resident, using up memory the space the other needs).
• If your onboard controller is tacked onto the PCI bus already, and with NF2 it is, adding another controller will fight for the same bandwidth. Though, really, this isn't a big issue with the number and speed of drives you have.
• The drivers for your controller may really suck.

 

[FLECOM]

If your onboard controller is tacked onto the PCI bus already, and with NF2 it is, adding another controller will fight for the same bandwidth. Though, really, this isn't a big issue with the number and speed of drives you have.

really unless you are talking about high end server boards everyone's IDE is attached to the PCI bus...

 

[Snugglebear]

There are a few desktop chipsets that run the onboard controller on a seperate line off the southbridge. I know via introduced one (KT800? - their competitor to the NF2, I think), and Intel has a few floating around. Granted they are rather few and far between, they do exist and if you do your homework you can find the right ones.

 

[Ice Czar]

and its far more common on serverboards and workstations,
but then they are typically 64bit buses

and then there is the Hypertransport schemes like on my board where there is a "southbridge" IO Bus (Firewire, USB, single PCI 32bit 33MHz slot, IDE channels and 4 SATA ports) and then the PCI-X Bus Chipset with dual buses, and then the AGP Tunnel Chipset, of course the Memory controllers are on die

theodork considering the likely throughput probably not much impact
is that a dual channel controller? or 2 single channel controllers

as Snugglebear mentioned, its not that easy to saturate the bus, I would however avoid getting the primary controller card assigned to the same PIRQ as your network card\or onboard controller if you can avoid it, unfortunately most desktop boards dont come with PIRQ routing tables
here is a review of the basics > PIRQ routing
that also controls which card the BIOS sees first in the boot order (it will be set to SCSI as IDE controller cards, IDE RAID cards and onboard controllers use SCSI Drivers)
thus which BIOS of a number of cards would load first, as mentioned some have larger bootup requirements than others, and getting them in an order that works isnt always easy