VPUs poised to become indispensable to image processing, machine intelligence, augmented reality and other cutting edge applications.
Jon Peddie Research (JPR), the industry’s research and consulting firm for graphics and multimedia, today released its latest quarterly report on the Visual Processing Unit (VPU) market.
The second quarterly edition of this new report analyzes the suppliers, the technology, the processors, and the market opportunities. “We have identified 20 suppliers of VPU, and eight IP suppliers, plus ten start up that haven’t produced any silicon yet,” says JPR president Dr. Jon Peddie. “Thirty-eight companies, with industry giants like Intel, TI, Nvidia, AMD, and Qualcomm, don’t get into a market and invest millions of dollars in R&D and acquisitions unless they see a big return. Therefore, no semiconductor, system builder, or software tool, or application developer can afford to ignore this emerging, maybe even explosive, market.”
JPR says this new market is already robust, but there will be inevitable consolidation. “We think there will be just a half dozen suppliers, three major companies, and three niche players by 2020,” says Peddie.
VPUs are at the crossroads of image-processing, convolutional neural nets (CNN), machine intelligence, and the emerging augmented reality market. More than just an image processing algorithm co-processor, and more like a powerful subsystem that can take multiple streams of high-speed pixel data and feed a GPU for display, while simultaneously doing data analysis and extraction.
The VPU is a relatively new device, in fact only one company is making a complete standalone VPU core, Verisilicon, but others are on the way. Cadence has three processor cores, the P5 and P6 which are being used in classic VPU roles, and we expect them to also produce a CNN optimized VPU. which can be used in classic VPU roles, and we expect them to also produce a CNN optimized VPU. Wave Computing’s Coarse Grained Reconfigurable Array, and Google’s Tensor Processing Unit represent new approaches to neural network training and inferencing respectively. For cloud-based deep learning applications those two are at the top of the class and featured in this edition of the VPU report.
The applications for VPUs ranges from super-smart prosumer cameras to automobile license readers at bridges and gateways, to airport security and nozzle monitoring of a satellite launching rocket. With high-resolution cameras being employed in every aspect of our lives, making autonomous vehicles of all types possible, drone surveillance and crop assessment affordable and reliable, and face recognition at ATMs a new normal in our lives, the demand for high-performance front-end processing of the myriad of image-processing functions has never been greater.
Other VPU capable devices like Ceva’s XM4, Inuitive’s NU 4000, and the VPU capable processors Intel has acquired recently from Movidus, Silicon Hive, Nervana, and its own remarkably capable Gen9 GPUs are among the 38 companies JPR has identified that are making VPU capable processors using GPUs, DSPs, and dedicated engines.
