So, as it stands i've been shot down or unawnsered by nvidia,physxinfo,epic games,stack exchange and various other developer sites. Basically what I want to know is why is GRB in UDK and in other applications limited to 255/256mb of mem?you can see the settings names here and these exact settings apply for both udk 2014-08 and the apex physx sdk.
if i try to set this to 256 or beyond the apex sdk or udk revert back to cpu physics mode.i want to know if this is some sorta cuda limitation since the code is based off of CUDA radix or that maybe its the physx gpu heap allocation size conflicting with settings or possibly this is some sorta driver or kernel limitation?also i would like to know exactly which file the limit cap resides in and what is this memory exactly based off of? is it based off of total warps/register/thread counts consuming the total of 255 or is this some sorta gpu vram based consumption. if anyone knows the awnser to this question and yes i'm aware that nvidia has updated their gpu rigid solvers and that this is outdated code from 2012 but also its still one of the best alternatives to cpu based physx has it has massive speed ups on fermi architecture and beyond.