What’s All the Excitement Over Software Defined Visualization and in Situ Visualization?

istock 528616902

Software Defined Visualization (SDVis) has become a mainstream idea: visualization on processors (CPUs) has enormous advantages in flexibility, cost, and performance for large visualizations. While that sounds like a great marketing pitch for Intel, the facts actually check out on this for a number of reasons, including trends in memory and algorithms.

These advantages are overwhelming for truly big data (referred to as “exascale” data).  In the paper “An Image-based Approach to Extreme Scale In Situ Visualization and Analysis,” the data movement challenge is quantified as being 1,000,000,000X as much data for extreme scale simulation data (where CPUs win) vs. that of image processing (where GPUs have dominated).

Eliminating data movement with in situ visualization is a hot topic in the scientific literature and is viewed by experts as a technology requirement for visualization in the face of exascale data. The need for in situ visualization makes SDVis even more important.  “In situ,” meaning “in the natural or original position or place,” makes sense when you consider that moving data around in a computer consumes power and uses up time. For large sets, the power and time can be staggering. SDVis offers an alternative: do it all on the CPU.

Visualization is more important than ever given the massive data sets in scientific computing, big data problems, and artificial intelligence (including machine learning).  We want to understand this big data, a task GPUs were simply not designed to handle.

A few years ago, the biggest visualization solutions started emerging that showed better results on CPUs than on GPUs. A nice summary of this trend, with data, was published in 2015 as “Contrary View: CPUs Sometimes Best for Big Data Visualization.” It showed that a single Intel Xeon processor E7 v3 workstation was able to render a 12-billion particle, 450 GB cosmology data set at seven frames per second. For that input data set, it would take more than 75 GPUs to perform the same visualization task.

In the two years since, the widespread adoption of SDVis has moved this to a more mainstream reality rather than a radical contrarian opinion.  The Best Visualization and Data Analytics Showcase award was won by the Los Alamos Data Science at Scale Team at Supercomputing 2016, highlighting the fact that CPU-based rendering is now at the forefront of visualization technology. You can see the LANL team’s award winning asteroid impact visualization online (very cool).

visualization Intel

Visualizations of LANL/TACC “Asteroid/Earth Collision” data set


Necessity gave birth to OSPRay and OpenSWR

It’s been said that “necessity is the mother of invention.” The author of the 2015 “Contrary View” paper revisited the situation a few months ago in a follow-up piece titled “CPU-based Visualization Positions for Exascale Supercomputing.” In it, Jim Jeffers shares that CPU-based SDVis has seen rapid uptake as the result of two developments born from the necessity to visualize such enormous data:

1.     The general availability of highly optimized CPU-based rendering software (such as the open-source OSPRay ray tracing library and the high-performance OpenSWR raster library in Mesa3d) integrated into popular visualization tools like Kitware’s Paraview and VTK, as well as the community tool, VisIt; and

2.     SDVis filling the big data visualization community’s need for software that uses runtime visualization algorithms that can handle giga-scale and larger data (e.g., exascale).

The transition from OpenGL-targeted hardware rasterization to CPU-based rendering has meant that algorithm designers can exploit large-memory (100s of GBs or larger) visualization nodes to create logarithmic runtime algorithms. The importance of logarithmic runtime algorithms is tremendous in the face of orders-of-magnitude increases in data sizes.

A ‘dream machine’ for visualization work

If you attended ISC 2017, you may have seen a “dream machine for SDVis” in the Intel booth. If not, you have a second chance to see it at SIGGRAPH 2017. Of course, if this article has really excited you, you can just order one (prices start around $79,000). The machine is designed for Software Defined Visualization with a special eye toward in situ visualization development. It’s ready to support visualization of data sets up to 1.5TB in size, and it addresses the needs of the scientific visualization and professional rendering markets.


Image processing looks easy compared to the 1,000,000,000X size challenge of truly big data visualization in the “Exascale Era.”  CPUs have emerged as the winner for visualizing such data because of their memory capacity and ability to support O(log n) algorithms.  This is fueling a rapid movement to in situ visualization and Software Defined Visualization with CPUs at the center of this truly enormous data.

For more information

I recommend the following sites for more detailed information:

SDVis Appliance website – includes detailed data sheet and information on how to order

Click here to download your free 30-day trial of Intel Parallel Studio XE