Post a Comment
I'm not sure if I understood it correctly but I got the impression that if you reprogram it yourself you can leave out the VGA compatibility and even make the hardware do something not related to graphics at all. IMHO it sounds pretty interesting and I am quite curious. But since I have almost no experience with so low-level software programming I won't benefit much from such card.. And I really can't think of much use for it either. But it sure sounds neat and I am 100% certain that there will be lots of enthusiasts buying these and coming up with dozens of different interesting use cases.
Because then you are not directly working with a single frame buffer, you either end up shadowing the frame buffer in main memory and have to copy that to the separate buffers in the video cards, or your software has to figure out which card it is drawing to on the fly.
Either method has a lot of overhead.
Well, if your software really has to have direct access to the hardware frame buffer.. But if your app was just an X application X would present all the screens to it as just one large screen and you could render anything you wish and leave the rest to X. Though, I don't have a clue what you're planning so can't really help more than that 
Sad that the original grand project appears to have failed while this smaller project looks much more promising.
Xilinx/Avnet also has a board with Spartan3s 1000 on it and a Philips triple 10b DAC tda8777 (upto 330MHz), runs about $300 or so and can use PCIe 1x connector. The PCIe IP needed for this can be used on trial. I would guess the Spartan3s might have a tough time driving the VGA anywhere near its max spec so the video output would likely be limited. The Xilinx docs mentions a 25MHz clock for driving at basic VGA 640 resolution, if it could atleast get to 1600x1200, I'd be a lot more interested, put my cpu core in as well.
The PCIe IP though is the hurdle there.





