Re: New PCI card developing

["Kevin Brace" <kevinbraceusenet@hotmail.com>]

>From: "Sergey Platovskikh" <sergey.platovskikh@vissim.no>
>To: <pci-sig@znyx.com>
>Subject: Re: New PCI card developing
>Date: Mon, 29 Apr 2002 14:53:40 +0400
>
>
>The card was given IRQ1 (that was indicated in BIOS information table,
>before OS loading) in HP Vectra VL420 (Pentium 4, I845 chipset).
>So, the the
>card wasn't working because of crasy IRQ assignment.
>I haven't got this PC yet to analyse the problem, but it seems that the 
>BIOS
>can't correctly recognize the card.


        If I am correct, in an IBM-PC/AT compatible computer, an 8042-based 
keyboard controller is connected to IRQ1.
Therefore, normally nothing else should be connected to this IRQ.
So, it sounds to me it is a BIOS bug than a hardware bug.
Perhaps you may want to obtain a new BIOS that fixes the problem.
Another thing you might want to try is to change the location of the PCI 
card.
Supposedly, if you populate a PCI card that uses INTA# near an AGP slot, the 
PCI's INTA# and AGP's interrupt will get shared, and I have heard about 
problems related to that.




>Also we had problems with BARx assignment with Intel I845BG motherboard
>under WinNT 4.0 . BIOS detected our card properly, but the driver could
>register only 2 BARs out of 4 - we got a message that 2 BAR addresses are
>busy but they were not (according to WinNT resource map). If we ignore this
>error message everything is working Ok. The card was working fine under
>Win2000. I am not an expert in drivers, so might be my explanation isn't
>clear enough...
>


        If the problem occurs with Windows NT 4.0, but not with Windows 
2000, again the problem sounds like a software (device driver) issue, and 
not a hardware issue.
However, why does your application need 4 BARs, and are any of them I/O 
mapped?
Newer host-PCI bridges might have more problems allocating I/O ports 
compared to older host-PCI bridges.



>Anyhow thanks a lot to the group for the information.
>I think I shall go for a PCI controller chip from Cypress or PLX. PCI cores
>implementation seems too complicated for my FPGA knowledge.
>
>Kind regards
>
>Sergey.
>


        In your first posting, you mentioned that even if you used a PCI 
ASSP chip (Those PCI bridge chips from PLX or AMCC.), the backend interface 
will still be complex, so you said you will rather use a PCI IP core if you 
can obtain one for free.
Whichever route you go, you will probably still have to spend significant 
time on the backend, so I feel like there is no huge advantage of using a 
PCI ASSP chip compared to using a PCI IP core assuming that you can get one 
for free or very cheaply.



Kevin Brace

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp.