Graphics processing unit (GPU) rendering isn’t pushing pixels in the same way that a real-time walk-through with 60 frames per second is. GPU rendering means using the power of the graphics processing unit instead of the central processing unit (CPU) for calculating images. The same is true when using the GPU to calculate the results of a computational fluid dynamics (CFD) simulation.
CFD and other simulation techniques are used across many industries: aerospace, automotive, marine, life sciences, etc. High-end rendering applies to fields as varied as feature films, television, advertising, engineering, design and architecture. Companies use GPUs for faster calculations and faster results because the increase in performance is measured not in percentage performance improvements, but in factors of performance improvements.
GPU computing performance applies to everyday workflows. For example, GPUs, and specifically NVIDIA’s Quadro M6000, are supported in ANSYS’s Release 17.0. Yes, GPU rendering is popular with production studios like Guru Studio. Design and engineering teams can also leverage GPU rendering with plug-ins for popular design applications such as MAYA, 3ds Max and Rhino, as well as with dedicated GPU rendering applications such as Redshift, which is a good example of performance gains on GPUs. Redshift can deliver a 10-fold performance increase in rendering using Quadro GPUs compared to CPU-only rendering. Automotive companies exemplify both needs using GPUs to render prototype styling and designs as well as simulating airflows for future products.
Common to all of these GPU-based workflows is a need for large quantities of graphics memory. The Quadro M6000 2016 release pushed onboard memory from 12 GB to 24 GB. The GPU remains the same as the 12 GB version and is based on NVIDIA's GM200 GPU with 8 billion transistors, over 3,000 compute unified device architecture (CUDA) cores, up to 7 trillion (tera) floating-point operations per second (TFLOPS) single-precision floating point performance and a memory bandwidth of 317 GB per second. The combined processing power, memory bandwidth and onboard memory make the Quadro M6000 the most capable GPU-based computing platform in NVIDIA’s Quadro family.
The Quadro M6000 has all the bells and whistles. It drives four 4K resolution displays. It features 3D stereo connectors. It supports GPUDirect technology for faster GPU communications. It has NVIDIA Quadro Sync support and NVIDIA Mosaic support for building multiple display systems for visualization. All those bells and whistles will cost $5,000.
When GPU performance is critical, and when the performance boost can be multiple times faster than other solutions, the price tag of a Quadro M6000 can look like a real bargain. The 2016 refresh providing double the memory capacity is an improvement on this high-end graphics solution.
About the Author
Tom Lansford is an international marketing consultant and manages the sites Professional Workstation, CADplace France and CADplace UK. He has been living in Europe since 1992, and previously managed workstation marketing in Europe at NVIDIA. Lansford is a professional videographer and his interests include design visualization, simulation, graphics and GPU computing.