A number of you are very keen on having a "graphics card" with some
kind of CPU or DSP on it.  While I don't think that that's necessarily
the most effective approach, there are reasons why it might be AN
approach quicker than some others.

So here's my suggestion:  If you all want to have a card like this, we
are capable of building one.  We just don't have the resources to
design it.  So you design it.

Obviously, there are gaps that we can fill in.  But going from what
you see in the OGD1 schematic, BOM, and artwork, you can see what
level of detail we need.  Provide that.  The more you provide, the
greater the likelihood we can make it.  Really, the best thing would
be a complete PCB design in Veribest or gEDA.

While Howard, Andy, and I can't do this work (any time soon), we can
guide others through the process.  Time to get grease on your hands,
folks.

Oh, and don't shoot yourselves in the foot by trying to pack in too
many features.  Figure out what is the minimum to make it work and
design that.  If you can get that to work, then make another version
with more stuff.

On 4/18/07, Nicholas S-A <[EMAIL PROTECTED]> wrote:


Note that this only supports VGA (640 x 480).
Which means it is not an option :-(. If only the 2700g was still being
sold...

This is another useful chip:
http://www.semicon.toshiba.co.jp/eng/product/micro/mep/document/index.html
it appears to be a configurable media processor with DSP interface, user
defined single-cycle external
interface, etc. It runs at 1Ghz, 1W!

As an alternative, this is the atmel DSP I mentioned earlier:
http://www.atmel.com/dyn/products/product_card.asp?part_id=3395
100Mhz, 1GFLOP:
For instance, four floating point multiplications, four floating point
additions and two
floating point subtractions, six memory accesses and six address updates can
be exe-
cuted simultaneously in a single instruction.

That is some serious number cruching.

It doesn't have a rating, but the Ipeak leads to around 2W in my
calculations.

I think this discussion is turning more to the SIMD approach, which is more
efficient.

It seems to me that the best way would be to design a small SIMD CPU into
the FPGA
fabric. This makes it much easier to interface and we get to specify it.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)



--
Timothy Normand Miller
http://www.cse.ohio-state.edu/~millerti
Open Graphics Project
_______________________________________________
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