On 10/7/05, Attila Kinali <[EMAIL PROTECTED]> wrote:
> On Thu, 6 Oct 2005 17:41:00 -0400
> Timothy Miller <[EMAIL PROTECTED]> wrote:
>
> > First, there's the Lattice:
> >
> > (1) PCI 32-bit
> > There are 84 pins defined, but lots of them are reserved or ground,
> > and some others don't connect to the chip.  We can itemize them here
> > just for general interest.  I won't do that just yet, because I
> > believe Andy has that worked out.  I also need to check to see if
> > PCI-X has any extras, because for that, I have only been considering
> > clock rate, not how the card indicates it can do the speed.
> > The TROZ spec lists 49 (plus the 64-bit extension), but it only did
> > target, so we'll have to add like 5 more pins to get mastering.
>
> Do you mean here the "interface" between PCI-* and the Lattice chip?

Yes, that's right.

>
> > (3) Local bus to the Xilinx
> > (3.1) Address bus (out):  32 pins
> > (3.2) Data bus (bidirectional):  32
>
> How about using here 64bit for both address and data ?
> Or 64bit data and 48bit address (though i doubt that these 16bit
> would make much difference) ?

Why do we need that?  Is there something special about PCI Express? 
Ok, I see your point on the address, but keep in mind that OGA's
internal address space really only needs 28 bits, and if we have a
PCIe controller that can do 64-bit addressing, it's going to translate
down.  The only reason for 64-bit data is speed.  If we had a 64-bit
PCI interface, it might help, but even then, we could just double the
data rate on the local bus.

One thing I forgot to mention:  The interface to control and program the Xilinx.

>
> It would enable us to keep up with the hardware development
> (like going to PCI-Express) by just changing the Lattice chip.
> I think we should be at least prepared for the developments
> that can be seen at the horizon, but leave them out of the
> design for the moment as we first need a (simple) product.

Yeah.  I'm going for doing double datarate if we need more bits. 
Also, Andy's going to provide more than just those 80 between the two
chips.  These are just the NAMED ones.

> > Then there's the Xilinx:
> >
> > (1) Memory
> > There are eight 256 megabit chips with 16-bit busses.  There are eight
> > data busses and four address busses (data busses run at 2x speed, so
> > we want point-to-point).  These are regular DDR chips.
> > (1.1) Memory address bus
> > (1.1.1) Address: 13
> > (1.1.2) Bank: 2
> > (1.1.3) Chip select: 1
> > (1.1.4) Clock enable: 1
> > [17 pins each, total 68]
> > (1.2) Memory data bus
> > (1.2.1) Data: 16
> > (1.2.2) Data mask: 2
> > (1.2.3) Clock: 2
> > (1.2.4) Data strobe: 2
> > [22 pins each, total 176]
> > The whole memory bus then uses 244 pins.
>
>
> If we have excess pins, how about using them to feed two
> "banks" of DRAM that can be independently accessed?
> I can imagine that this would give some performance,
> though i can not put it into numbers.

Actually, there are four memory controllers.  Memory is arranged in
stripes of pixel pairs interleaved with adjacent pairs over adjacent
memory controllers.

>
> > (3) Misc
> > Additional signals will
> > (3.1) Peripheral bus
> > Usually 8 to 32 data pins, plus maybe 10 address pins, plus some sync
> > stuff.  TROZ defined 24 pins for this, and I think that interface was
> > modeled after the interface required for most RAMDACs.  There will be
> > one on either the Xilinx or the Lattice or both.  But OGA will have
> > only one, so we'll probably define two and then enable only one in the
> > design.
> > (3.2) I2C
> > This is 5 pins:  Data in, data out, mode, clock, enable.
> > (3.3) Clock pins
> > There will be some additional clock signals that are used around the board.
> > (3.4) General I/O
> > Whatever is left
>
> What about JTAG ?

Yeah, that's part of what I was missing above with the control
interface.  Oh, and we need to not forget some headers to reprogram
the Lattice.

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to