PCI Slot 3 / USB Conflicts?

An area for people to discuss Scope related problems, issues, etc.

Moderators: valis, garyb

Post Reply
digitalboy
Posts: 15
Joined: Mon Jan 14, 2002 4:00 pm

Post by digitalboy »

I just read somewhere recently (I think the article was in Electronic Musician) that most PC's share PCI bus 3 with the USB controller and that this can lead to problems for any audio hardware placed in slot 3.

For starters, does anyone know if this is in fact the case? Also, does this just affect certain motherboards and chipsets or is the problem a Windows OS issue? Also, is this issue related only to Windows 98 or all Microsoft OS's?

I'm running Windows XP Pro on a custom PC with a Gigabyte GA-81RXP motherboard with the Intel 845 chipset. I have noticed in the System Information window that I have 23 IRQ's available (all of which are used at the moment - with several of the IRQ's being shared between my audio cards and various other controllers and system components). So maybe this PCI slot 3 / USB conflict isn't an issue with my setup???

If anyone has some definitive information on this matter, I would appreciate the feedback.

Thanks!
digitalboy
Posts: 15
Joined: Mon Jan 14, 2002 4:00 pm

Post by digitalboy »

Oh, I should probably mention that my motherboard features an additional RAID controller as well as onboard USB 2.0. These features might also be hogging the available PCI bandwidth???
Rob van Berkel
Posts: 236
Joined: Sun Jul 29, 2001 4:00 pm
Location: Rotterdam, The Netherlands

Post by Rob van Berkel »

It's all dependant on the motherboard implementation and the chipset. Reading the manuals of the Intel chipsets reveal a lot of info on how the various busses share bandwidth and how IRQ's are shared. As you already stated, RAID and USB2 generally do share PCI bandwidth with e.g. your pulsar board. On the contrary, the 'old' usb interface, the onboard LAN and onboard audio do not share PCI bandwidth, alhough all traffic does travel through the same chip (ICH2).
http://developer.intel.com/design/chipsets/845/ clarifies a lot of the chipset secrets.
Hope this helps.
Cheers,
Rob
User avatar
garyb
Moderator
Posts: 23375
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

raid? i dunno.w/ the 845 usb should be no prob(different busses for usb and pci).watch out for shared irq's tho.(check the mobo's manual about that)
Rob van Berkel
Posts: 236
Joined: Sun Jul 29, 2001 4:00 pm
Location: Rotterdam, The Netherlands

Post by Rob van Berkel »

On 2002-08-21 12:31, garyb wrote:
w/ the 845 usb should be no prob(different busses for usb and pci
Garyb, you're absolutely right when it comes to USB1.1, but we're talking USB2.0 here, which is only integrated in the newer 845E chipset. As this is a std 845 (ICH2) board it uses an external (probably NEC) USB2 chip. This chip shares PCI bandwidth (and interrupts) with the rest of the PCI components. It's a pitty Gigabyte doesn't give much info about this.
I expect USB not to be the real pain when it comes to PCI throughput, as long as you keep data that flows through it to an absolute minimum (don't use USB-lan, DSL, storage). RAID on the other hand definitely influences the bandwidth that remains to the rest of the PCI-bus. I personally would connect my disks to the standard IDE busses and disable the RAID. It is for that reason I bought a Abit TH7-II without raid. It had to be ordered because it wasn't in stock, as everybody buys the raid-version. But we pulsar users have different priorities :wink:
Cheers,
Rob
subhuman
Posts: 2573
Joined: Thu Mar 29, 2001 4:00 pm
Location: Galaxy Inside

Post by subhuman »

The Gigabyte board has been winning a lot of awards, but for our use, it really is not a better board than something without all that onboard stuff that is basically sitting on the PCI bus... it's much better to have things like USB(1.1, and 2), and iDE not sharing the limited PCI time.

A board without onboard RAID, and without "integrated PCI cards" like an NEC USB controller, are much better for our purposes (at least, for mine), less to think about, more PCI for audio! :smile:
User avatar
garyb
Moderator
Posts: 23375
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

yes!
(thanks)
User avatar
valis
Posts: 7673
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Post by valis »

I'd suggest, rather than disabling the extra RAID ports just use them in a non-raid config. for all your optical drives. CD/CDr shouldn't affect pci performance tremendously unless you're pulling an entire piano sample lib off cd. Even at that, it'll free up both of your ide ports for your OS/audio drives. If you wanted to be really tweaky you can put one drive per chain and maximize throughput, but with the speed of today's drives it's not a HUGE deal.
digitalboy
Posts: 15
Joined: Mon Jan 14, 2002 4:00 pm

Post by digitalboy »

Thanks everyone for your suggestions!

I should have mentioned in my original post that I have the onboard "RAID" controllers configured as two extra ATA 133 controllers. In other words, I'm not running a RAID setup.

I have two hard drives, a DVD-ROM, a CD-RW, and a Zip drive. The nice thing about the additional ATA channels is that almost every drive device can be on its own bus.

However, these two extra Promise ATA controllers show up as SCSI devices which is sort of weird. I don't know if that's because I have the Adaptec ASPI driver (version 4.71) installed???
User avatar
garyb
Moderator
Posts: 23375
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

raid and scsi are kinda the same thing(to the os), no?
Post Reply