• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

PCI bus transfer speed

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

ninthebin

Member
Joined
Mar 24, 2002
Location
Liverpool, UK
I have thought that myself...I dunno really, I would like to think they each get its own BUS rather than trying to hog one...

does anyone with pci network card sound card gfx card etc notice a distinct speed increase when taking everything out but the gfx card? if the answer is no, I suppose they are all seperate :|
 

hassmh

Registered
Joined
Jul 1, 2002
Location
SE PA, USA
It is 132MB/s for the entire bus. A lot of server/workstation boards (dualies and up) have multiple busses (as well as faster ones), I don't think any desktop boards do, though.
 

macklin01

Computational Oncologist / Biomathematician / Mode
Joined
Apr 3, 2002
Location
Bloomington, IN
Agreed, IDE master and slave share the speed on each channel. But as I understood it, each of the following were independent:

1) IDE0 (master and slave share)
2) IDE1 (master and slave share)
3) PCI (all slots share)
4) AGP

Is that right? Thanks! -- Paul
 

moorcito

Member
Joined
Sep 19, 2002
Location
Chicago, IL
yeah, that's what I meant. I though about adding the stuff about the channels but decided against it. Sorry about that
 
OP
Henry Rollins II

Henry Rollins II

Member
Joined
Apr 21, 2001
Location
The North Pole
macklin01 said:
Agreed, IDE master and slave share the speed on each channel. But as I understood it, each of the following were independent:

1) IDE0 (master and slave share)
2) IDE1 (master and slave share)
3) PCI (all slots share)
4) AGP

Is that right? Thanks! -- Paul

Im not a 100% sure about this, but I think that the IDE controller is a pci device. If it shares bandwidth with the pci slots also I dont know.
 

macklin01

Computational Oncologist / Biomathematician / Mode
Joined
Apr 3, 2002
Location
Bloomington, IN
Good question -- that's what I wondered, too. I know that they share the same clock speed and divider. (eg., on my board, I have a 1/4 divider, so at my current FSB , they're both running at 37.25 MHz.) But whether or not they share bandwidth is something I'm not sure of.

I do know that the two IDE channels are independent of one another. (This is why they recommend putting your CDRW and HDD on separate channels, for higher transfer rates.)

If somebody could help, I'd certainly appreciate that! ;) -- Paul
 

macklin01

Computational Oncologist / Biomathematician / Mode
Joined
Apr 3, 2002
Location
Bloomington, IN
Thanks! (And BTW, welcome to the forums! :sn: )

The strange thing about that, then, is that one of the the best ways to improve disk I/O may well be to move your video card to the AGP bus. -- Paul
 

ninthebin

Member
Joined
Mar 24, 2002
Location
Liverpool, UK
so if you have a ata166 hdd (even thought they never get that high anyway) but its a lie as it cant get higher than 133MB/s unless you have a better PCI standard?
 

theotherphil

Member
Joined
Nov 18, 2002
Location
England, Near Europe
Thanks :)

The majority of high performance GFX cards sold nowadays are AGP anyway so this isn't a problem. The lower spec gfx cards won't push the PCI bus much. What will though is a couple of WD Special Editions in a RAID0 array....there's ~90Mb/s used straight away on a sustained transfer. If you do any sort of video encoding using dedicated Video hardware and a RAID array than things can get crowded :eek:.
 

macklin01

Computational Oncologist / Biomathematician / Mode
Joined
Apr 3, 2002
Location
Bloomington, IN
That's right. You can never have higher transfer rates on the PCI / IDE bus than 133 MB/s. In reality, though, this isn't what's holding up disk I/O; it's the disks themselves, since they can never handle sustained transfer much above 70-90 MB/s on some of the fastest, fastest SCSI drives. -- Paul
 

macklin01

Computational Oncologist / Biomathematician / Mode
Joined
Apr 3, 2002
Location
Bloomington, IN
theotherphil said:
Thanks :)

The majority of high performance GFX cards sold nowadays are AGP anyway so this isn't a problem. The lower spec gfx cards won't push the PCI bus much. What will though is a couple of WD Special Editions in a RAID0 array....there's ~90Mb/s used straight away on a sustained transfer. If you do any sort of video encoding using dedicated Video hardware and a RAID array than things can get crowded :eek:.

Good points again! :) But ah, what a pleasant form of PCI bus crowding!!! :D

Actually, the one that occurs to me, then, is a 100 MB/s LAN transfer. Say you're transferring data over a network from a fast drive to a fast drive, one of the transfer rates is gonna' get clobbered .... Interesting. I guess PCI-X will be a good thing someday! ;) -- Paul
 

theotherphil

Member
Joined
Nov 18, 2002
Location
England, Near Europe
macklin01 said:
That's right. You can never have higher transfer rates on the PCI / IDE bus than 133 MB/s. In reality, though, this isn't what's holding up disk I/O; it's the disks themselves, since they can never handle sustained transfer much above 70-90 MB/s on some of the fastest, fastest SCSI drives. -- Paul

On a single SCSI drive maybe...the Cheetah X15.3 gives close to 70Mb/s :eek:. Get a couple of those babies RAID'ed and you can easily saturate a 33mhz/32bit PCI bus. I am running a couple of Atlas 10K III's in a RAID0 array and get pretty decent performance (see below). If I had a normal uni-processor machine (that only has a 33/32 bus) and also had a few other high I/O cards, I'd be in trouble.

iwill64dyn.gif