On my most recent research I found a very interesting study made for a Thesis in a Sweden university. They had studied, implemented and compared three different methods for 2D fluid simulations in games. Although all of them were particle based, the document has given me a different point of view, a new perspective. One of them got my atention when they proposed in the paper to use both particle and grid methods, but having the second one just to keep track of the nearest neighbours. That way, the algorithm was very optimized.
However, the rendering of the fluid still has the problem of being composed by metaballs. An option could be changing the way that particles are drawn, but that might be very complex.
As it can be seen in the image above, although many balls can be used (thanks to a better performance), the look is still very unrealistic due to the static shapes of the particles.
