Sumit Gupta works for Nvidia's Tesla division, in which some of the company's brightest minds work on supercomputers and advanced computing. The company aims to shape the future of embedded computing with their latest product, The NVIDIA Jetson TK1 development kit which boasts of a192-core graphics processor kit built on the same architecture that powers the Titan supercomputer, allowing you to have your own personal supercomputer for just 192$. Sumit shares details in an interview with Krishna Bahirwani.
Would you call the Jetson TK1 dev-kit potential personal supercomputer?
Jetson TK1 is a supercomputer for embedded applications. It is powered by the NVIDIA Tegra K1 mobile processor, the highest-performance processor ever designed for mobile and embedded form factors. Tegra K1 has 4 ARM CPU cores and 192 GPU CUDA cores based on the same Kepler architecture that powers the U.S.'s fastest supercomputer, the Titan supercomputer at Oak Ridge National Laboratories, as well the world's top 10 most energy efficient supercomputers.
The Tegra K1 processor has a 192 fully programmable CUDA cores, delivering over 320 GFLOPs of performance -- nearly three times more than any similar embedded platform -- yet has a power and thermal characteristics that allow it to be put into portable embedded devices.
As robots become able to navigate complex environments, as medical devices become more portable, and as cars learn to see pedestrians, the computational demands for these applications is increasing exponentially. The Jetson TK1 delivers the levels of supercomputing-class performance needed for these markets.
How does this board help budding Data Scientists and students of Machine Learning?
The Jetson TK1 board enables data scientists and students of machine learning to test their algorithms to guide robots and other systems. GPU accelerators can be run first to train the machine learning models and then to classify objects, for example, that a robot or a car “sees” with their cameras.
Raspberry Pi and Arduino have a large community and educational resources backing it? Do you see something like that happening for Jetson?
Yes, it is our goal to foster a large and vibrant community around the Jetson TK1 dev kit, as that is key to its broad, longer term success. We want to make the computational power of the Tegra processor available to the community, and we need to give them all of the tools they need to do amazing things in computer vision and the embedded space.
We already provide extensive documentation to the community on the dev kit and Tegra SoC, software downloads for the BSP, CUDA, and other packages our developer site right now (developer.nvidia.com/jetson-tk1) and. We will continue to build this out, providing more opportunities for community input and participation moving forward.
Can the bundled VisionWorks present exciting applications in computational photography such as face detection across the Internet?
Absolutely.VisionWorks is a toolkit for computer vision and image processing. We’ve taken many of the core algorithms from these fields and ported them to CUDA. VisionWorks provides an infrastructure to link these algorithms together into pipelines to solve real problems. It’s a very flexible architecture; developers can add their own algorithms, or leverage algorithms from different sources, like OpenCV.
If I would want to give the Jetson TK 1 a try as a hobbyist where should I start?
The best place is to go to developer.nvidia.com/jetson-tk1 to learn more about the board and purchase.
What are the potential security applications of the board?
A lot of interesting work is being done apply computer vision and image processing technology to security. Jetson TK1 works in a small power envelope, making it possible to run data-processing intensive applications like facial and object recognition in small, low-power devices like security cameras. It’s going to change the way that people think about deploying security installations.
·Is the future of supercomputing personal?
Even today, when we search for something online or use mobile apps like Shazam that recognize a song or check the weather on our phones, we are really using a massive supercomputer in the cloud. So, supercomputing is becoming more and more personal every day.
Jetson TK1 delivers incredible computational performance in a compact form factor, that enables building all sorts of new handheld and portable devices with supercomputing performance. As processors like the Tegra K1 continue to push the envelope of performance and energy efficiency, they bring similar levels of computing power to “personal” systems that were only available in large scale supercomputers just a few years ago. And, CUDA makes it simple to leverage all of that power. It’s easier than ever for developers to create and inspire the next generation of innovations.