5 For
- Affordable: Nvidia’s Graphic Cards can be combined to build a super computer.
- Easier parallel coding: Since Nvidia inserts an abstraction level, parallel thread programming burden of the programmers is diminished.
- Forward code compatibility: Previously mentioned abstraction level also enables to source code to be used for future hardware versions without modification. But modifications will make code performance better.
- Extensible architecture: The designed model can be used for different levels of architecture, by adding or subtracting units of GPUs.
- C / C++ syntax
2 Against
- CUDA is strict to Nvidia solutions.
- CUDA is not mature yet.