Mathematica 7 Performance Boost Thanks To CUDA

Mathematica 7 with CUDA

The new version of Mathematica now integrates CUDA technology. Thanks to the computational horsepower of modern GPUs like the GT200 (GeForce GTX 260/280, Quadro FX 5800, Tesla), Mathematica gets an boost in performance of 10-100X in many situations like numerical computing or simulation.

I hope other famous software like Mapple will follow this path.


For french readers only: Mathematica 7 Améliore sa Puissance de Calcul grâce à CUDA @ Geeks3D VF

One comment

  • Jonah

    I have been interested in CUDA for years, but the various hassles of loading the toolsets and the obligatory programming in yesterday’s third-generation languages (like C and Fortran) deterred me from experimenting. What end-user has the time to program in Fortran these days? Not me…

    Mathematica to the rescue! With Mathematica 8, you have full programming access to CUDA in Mathematica’s 4GL, as well as transparent use of CUDA in many Mathematica functions and commands. To my knowledge, Mathematica is the best and easiest way for people like me (I’m a chemist) to use — and have fun — with CUDA.

    I’d go a step further and say that Mathematica would be a very labor-efficient door to massively parallel programming for mathematics as well as other domains, such as parallel search, for software developers and project managers.

    The only small “gotcha” is that you have to pony up for a Mathematica license (you need a Commercial or Student license; the home user license is seriously shortchanged on multicore/CUDA support).