Every effort has been made to ensure the visual results of all variations are the same regardless of what vendor of GPU you use. However, there are some considerations to keep in mind.
It’s no secret that AMD is very much in favor of OpenCL and other similar open standards. It’s also no secret that despite Nvidia being on the Khronos board, they are much less in favor of OpenCL and instead prefer their closed standard, CUDA.
This leads to a peculiar situation where both vendors technically support OpenCL on paper, while only one fully does.
Nvidia’s implementation of OpenCL is crippled and therefore you may get strange results when using their cards. Below is a running list of known problems that have no known fix:
- Nvidia’s implementation of OpenCL does not like the synth variation. This is likely because it contains a very large amount of code which contains function calls. Whenever you add it to an xform, you will see Fractorium freeze for roughly one minute and your system will consume an additional 4GB of RAM while compiling. No such problem exists with AMD cards.
- Certain variations will fail to compile on a Mac with an Nvidia card when using double precision. This is due to a bug in their compiler.