PCI cards in PM G5...do slots matter?

[kugino]

i have a few PCI cards (Miglia Alchemy DVR, SIIG SATA, Koutech FW/USB) that i've tried in the different slots in my Rev. A PM 2.0. they all work in slot 4, the one that runs at 133 MHz. however, none of them work in slots 2 and 3. i've tried them individually, in tandem, etc., but only one card at at time works, and it's only in slot 4. i took the PM to the genius bar today and the "genius" said that individual cards are hit and miss as to whether they'll work or not. i understand that. we all know of the deep sleep issues with some of the cards and that some cards don't seem to work as well as others. BUT, am i missing something by thinking that IF the card works perfectly fine in slot 4, it SHOULD also work in slots 2/3? are slots 2/3 and 4 so different from one another? all the cards i have run at much less than 100 MHz, probably at 66MHz...

i left the PM with the apple store and they're going to check it out, but the "genius" doesn't think they'll find anything (on a side note, he was the worst genius i've ever talked to/dealt with...not very friendly...also had to correct him on some of the issues). can anyone clarify things for me? thanks!

 

[risc]

I work with 64 bit PCI cards all the time on x86 servers and PCI-X 133 / 100 PCI 66 are all backwards compatible (you can even run some 32 bit PCI cards in 64 bit PCI slots), if I have PCI-X 133 card and no PCI-X 133 slots I'll dump it in a 100 or 66 mhz slot with no problems at all, I even put 66 mhz cards in 133 mhz slots depending on the situation.

I'm sorry to hear about your problem, but other than a design fault I can see no logical reason for it.

 

[maxvamp]

The way that PCI-X works is that there are several independent channels.

In a normal PCI configuration, all slots have to share a total of 133MB/s to send data, but in a 133 MHz slot on a PowerMac PCI-X slot, that one slot gets all of the available bandwidth.

The main reason for this, and why PCI-X is more expensive to produce, is that that slot/bus has it's own controller. The 100 MHz slots share their own controller, and if there were an standard PCI slots, they would have their own controller.

It sounds like the controller for the 100MHz bus is bad.

Max.

 

[kugino]

The way that PCI-X works is that there are several independent channels.

In a normal PCI configuration, all slots have to share a total of 133MB/s to send data, but in a 133 MHz slot on a PowerMac PCI-X slot, that one slot gets all of the available bandwidth.

The main reason for this, and why PCI-X is more expensive to produce, is that that slot/bus has it's own controller. The 100 MHz slots share their own controller, and if there were an standard PCI slots, they would have their own controller.

It sounds like the controller for the 100MHz bus is bad.

Max.

i agree. i think the controller is bad, but the "genius" didn't even want to test it - he thought i just had bad cards. hopefully they'll find something.