July 12-16, 2015
Join RapidIO.org along with member companies Concurrent Technologies, Integrated Device Technology (IDT), N.A.T., Prodrive Technologies and Texas Instruments at the 2015 ISC High Performance Conference at booth #622.
- Concurrent will be demonstrating a modular based approach to heterogeneous computing. Current modules include a range based on Intel Core i7 processors with a new module using multiple NVIDIA Tegra K1 processors. Flexible combinations of modules can be used to suit the application and are linked using a RapidIO fabric connection for low latency and low CPU overhead.
- IDT will be demonstrating their High Performance Analytics Lab with Heterogeneous RapidIO based low latency computing.
- N.A.T. will demonstrate high performance platforms which interconnect to external networks through multiple low latency copper/optical RapidIO and 10 GbE interfaces. N.A.T.s own high density and performance optimized processing bricks contain a combination of Multi-Core-PowerPCs and high-range FPGAs. At user’s choice they interconnect via a local backplane through low latency RapidIO and either 10GbE or PCIexpress. Storage applications using NVM Express to access solid-state drives (SSDs) benefit from 32 Gbps PCI Express compared to 6 Gbps of SATA. The open standard also allows to easily build heterogenous systems based ARM/x86/PowerPC-CPU, GPGPU, DSP, FPGA and SoC and optionally connect these to telecom networks by available optical broadband interfaces, i.e. SDH.
- Prodrive will demonstrate their latest building blocks and infrastructure technology for heterogeneous HPC, based on low-latency RapidIO switching technology. Building blocks of a wide variety of silicon vendors will be shown: Xilinx 7-series FPGA and heterogeneous ARM+FPGA SoC, Texas Instruments heterogeneous ARM+DSP SoC, and more. Infrastructure based on IDT RapidIO switching, such as the modular Data Center and Compute Platform and other enclosures, a RapidIO top-of-rack switch box and more.
- TI will demonstrate applications built on convolutional neural networks (CNNs) and graph libraries running on HP Moonshot with m800 cartridges featuring TI SoCs with ARM A15s and C66x DSPs. CNNs and graphs are used in technologies including classification, detection and inference.