I'm a noob to this list, so I've remained quite. But this is kind of the
reason I joined this list, so I figured now is a good time to speak up.
There are two major functions that graphic cards need to do these days.
The first is to actually get a display out to a monitor. (DUH)
The second has become this GPU processing of 3D elements.
In order to take the load off the CPU and it's RAM, the industry has
been moving more and more of these functions onto the graphic card. This
was important before the days of multi core processors, but my question
is... are any of these GPU's really that much faster than a single Intel
or AMD core at these math functions? If you dedicated an entire core to
just graphics, instead of OS chores or other functions, would it be a
match for the latest GPU hardware?
We've just had an 8 Core Mac released, and we have Quad core chips
around the corner that will turn even simple embedded PC's into quad
core monsters.
One of the reasons for an "Open Graphics" platform is to get away from
proprietary hardware and it's closed source drivers, but the graphic
card industry is moving toward bigger and bigger cards with more
powerful GPU's. Essentially they are just sticking a custom graphic
computer onto the bus of our main system. They're business model is
based on new technology trickling out, and forcing a social need to
upgrade.
Instead of a bunch of custom hardware, and trying to emulate the big
boys... why not move those functions back onto a single, or even dual
core of the main machine.
From what I'm reading, we don't even have an effective way to program
for these multi-cores. Unless you're doing enterprise level spreadsheets
or 3D animation, the bulk of this processing power is going to be
waisted right now on virtualization anyway.
If you design a custom board to do this kid of 3D, it's always going to
be a generation behind the top of the line, and it's going to be more
expensive to produce because of the niche popularity. Less power for
more money is never a good way to do anything.
Instead... use the most generic parts you can. If you have a look at the
current generation of high end cards... they're power hogs with huge
physical requirements.
If we need a CPU/GPU on the card for some reason, let's just go with the
most sold processors on the planet. Then we get economies of scale
working FOR us.
We also get the support of the big processor manufactures as part of
this deal. Ok... not so much AMD since they purchased ATI, but Intel
would LOVE this idea since it would sell more of their products.
The only other concern is sharing processor memory. That could slow the
whole system down. But what if our board just has it's own high speed
RAM like the big boys? Imagine a graphic card that relied on a Core Duo
as it's processor and had it's own high speed RAM. Is anything better
documented than that? At this point we're talking more about a software
graphic OS, than a dedicated piece of hardware.
Since graphic cards are climbing into the $500-$1000 range anyway... it
doesn't sound that far fetched to me to just add a whole computer to get
the job done.
That's my idea of a TRULY Open Graphic platform. A PC dedicated to
graphics with something more like a LinuxBIOS dictating it's functions.
Hope I'm not out of line, and that no one takes office at this line.
It's really kind of a question phrased as a statement.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
