After firsts tests and benchmarks of OpenCL, showing that a GT120 (GeForce 9500GT renamed), an $60 entry-level card is equivalent for pure floating point computing than 2 quad-core Xeon at 2.93Ghz (a beautiful $2000 price tag), critics are writing that CUDA & OPenCL are of limited use.
They only see them for mathematical calculation, financial analysis (that is math too), and some niche markets.
They may have wrote the same from the apparition of the computer in the 50’s or the microcomputer in the 70’s!
Some usage of OpenCL & CUDA:
- Any platform – virus detection accelerator
- MySQL – Sort accelerator/Memory table accelerator
- Apache & https – SSL accelerator
- OS – Encryption accelerator (full encrypted HD or communication)
- Photoshop – Image edition accelerator
- Web – thumbnail generators
- Video encoding – accelerator
- Video decoding – accelerate proprietary format TOO
- Sound Mixing/Editing – DSP simulator, hardware accelerated
- iPhoto – Face detection & face recognition
- iTunes – Sound Equalizer, reverb, etc to enhance sound quality
- Games – Physics computations (aka Physx on nVidia)
- Web browser – Image decompression, accelerated UI
- Chess Game – 1 billion node/second on a PC or Mac!!!
Most of them are actually implemented, in studies, in experiments, in open-source code, in proprietary software, and in the next years, many of these tasks that is heavy for the CPU will rely more and more on a GPU.
