Visualization in Software using Intel Xeon Phi processors

Print Friendly, PDF & Email

Sponsored Post

Visualization of very large data sets is an important part of many data scientists workflows. The ability to make sense of the terabytes of data that a simulation can produce leads to new insights and can quicken decision making. While there have been tremendous advances in the performance of hardware graphics chips and systems, there are situations where rendering the data in software is beneficial.

For organizations and research labs that are pushing high performance computing technologies, tremendous amounts of data are being generated from more complex simulations. Interactive visualization of this data can lead to better decision making faster. While most visualization software systems have been tuned for the latest GPUs over the past few decades, there are some challenges that can be addressed by using software rendering techniques on the latest Intel Xeon Phi processors. These are:

  • The amount of memory needed to hold terabytes to petabytes of data far exceeds what can be addressed with off the shelf graphics cards.
  • The transfer of the data that is required from the CPU to the graphics cards exceeds that of standard I/O capabilities.
  • Visualization techniques that are needed in some cases can be handled better through highly programmable CPU architectures as compared to traditional rasterization focused GPUs.
  • The power needed for additional hardware such as GPUs can break the power budget for systems and many racks of these systems.
  • The Intel Xeon Phi processor contains the highly parallel architecture that can scale as data grows with increasingly powerful system architecture.
  • New visualization techniques can be more quickly developed using software than waiting for hardware to be designed, verified, and tested.

[clickToTweet tweet=”Software defined visualization is alive and well.” quote=”Creating state of the art visualizations can be done in software, giving scientists and engineers more insight into their massive amounts of data.”]Intel has been at the forefront of working with software partners to develop solutions for visualization of data that will scale in the future as many core systems such as the Intel Xeon Phi processor scale. The Intel Xeon Phi processor is extremely capable of producing visualizations that allow scientists and engineers to interactively view massive amounts of data. This is due to:

  • Quite a large amount of memory for these tasks, about 400 GB memory.
  • MCDRAM, high speed memory, which can be configured for each processor.
  • Many cores – up to 72 for rendering purposes.
  • The AVX-512 instructions which can support complex ray tracing and performance gains of the graphics pipeline.
  • Low power consumption while rendering data sets.
  • Intel Omni-Path fabric which speeds up multimode communication for distributed visualization and other techniques when more than one system is needed for performance.

Rendering in software is not new, but with the performance and memory capabilities of the Intel Xeon Phi processor, visualization in software can be done at interactive rates.  Since new and more complex visualization techniques can be developed on the Intel Xeon Phi processor as compared to hardware implementations, scientists and engineers can view their data in new ways, leading to new discoveries.  While the Intel Xeon Phi processor can enhance the performance of a software defined visualization package,  similar techniques could also be performed on Intel Xeon CPUs. This flexibility of running an interactive visualization on the Intel Xeon processors or the Intel Xeon Phi processors addresses the needs for a range of solutions that can fit the requirements of a wide range of those who require these features.

Download your free 30-day trial of Intel® Parallel Studio XE 2018