Timothy Miller wrote:
> This isn't a bad idea, and it has been looked into before. Early on,
> we investigated some of the Intel northbridge chips, and what we found
> was that it's just not designed to function without having a CPU
> hanging off of the FSB. That's how it gets initialized properly, and
> basically, your solution puts the CPU on the wrong side of it.
No, you've completely misunderstood my proposal:
---------------------
| motherboard |
---------------------
| PCI Express
| \
--------------------- \
| Spartan 3 FPGA: | |
| PCIe-FSB bridge | |
--------------------- | proposed
| Intel P4/Core FSB | card
| |
--------------------- |
| G965 North Bridge | |
| w/ GMA X3000 | |
--------------------- /
| | | /
DDR DVI DVI
mem 1 2
The *only* interface between the host and the G965 will
be through the Intel FSB, just as the chip is intended to
operate. The PCIe interface of the G965 is not used at
all (nor can it be, since it shares pins with DVI 2).
The FPGA does the *opposite* of a normal FSB-to-PCIe
bridge. It takes PCIe transactions initiated by the
host, and turns them into FSB master transactions to
the G965. The G965 has no way to know that it isn't
connected directly to a P4/Core FSB. Thus there is no
problem with initializing the G965 registers.
Eric
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
