The 32-bit OS limits the LuxRender process to 2GB. As already mentioned, you are likely running out of memory. Even with just one figure and minimal background, Lux usually floats around 2.5GB minimum memory usage when I render at 2560x1440. And it’s not uncommon for Lux to require 6-8GB of RAM for more complex scenes.
Things you can try doing: Check the ‘memory conservative’ option in the Options tab (which enables QBVH, which is better/faster than the kdtree default AND uses less memory). Reduce the resolution you are rendering at. Reduce or turn off outlier rejection (the “FF” option on the Options tab that defaults to 5—I forget exactly what Reality’s “friendly” name is for this). Enable ‘collect textures’ and have it reduce the textures to say 25% of original size. But when you say 5 or 6 figures in a scene, I think you’re really just gonna be out of memory no matter what. If some of the figures are more in the background, you could try using one of the lower Level of Detail versions of the Gen4 figures to reduce their poly count (I use DS4, so not sure where the load LOD option is in DS3). (Or if using Genesis, disable subdivision for the figure in Studio to reduce the poly count exported to Reality.) Also, anything not visible, hide it in the Scene view in Studio. For example, if a figure has long boots on, hide their foot, toes, and even the shin if the boots/pants cover it. This will also reduce the amount of geometry exported to Reality.
With the exception of changing the outlier rejection setting, all of those options will also reduce the quality of the render. Reducing outlier rejection won’t change the quality per-se; it will just require many more samples/pixel to achieve the same results as a render done with the default outlier rejection. (Edit to add that hiding invisible bones will also not impact render quality, and should really always be done as it will speed up LuxRender as it has less geometry to consider when determining ray intersections.)
But the real solution is to upgrade to a 64-bit machine with a minimum of 8GB RAM (12GB would be much better). But I realize that’s not always an option.