Wednesday, October 20, 2010

Is the future of graphics FPGAs or CPLDs?

Well, it is the ultimate conclusion for pixel/vertex/geometry shaders...


They provide a difficult to program interface, but provide a way to approach the fastest possible solution to any massively repeating logical manipulation (such as a shader).

I'm an advocate of turning your game code into streams of data, and this seems like a really good way of writing data transforms in a super efficient manner.

2 comments:

Funcan said...

The general experience from the HPC end is that FPGAs are always 10 years off... the cost/performance of them compared to more conventional technologies, particularly when you factor in the significant increase in development costs, has kept them very, very niche - a few cryptos and other pure integer types use them, but generally if you wait a few months you can do things faster and easier on CPU or GPU...

Richard Fabian said...

But, what if it was specialised towards graphics pipelining, where the basic building blocks were madd and other non single bit logic operations? (like the CPLDs) I'm just wondering if the next step in making shaders go fast is to have the shaders try to run in single or very few cycles. Also, why are things faster (I can get the easier bit just fine) ina few months on CPU or GPU? is it just that they bring out hardware that surpasses it or maybe that they assimilate knowledge gained by using FPGAs and incorporate it into the next hardware rev?