German board maker congatec has launched a modular starter kit for industrial IoT developers working with the new PICMG COM-HPC high performance computing standard.
The kit is based on congatec’s Computer-on-Module conga-HPC/cTLU and Intel’s 11th Gen Core processor technology (code name Tiger Lake) along with high-speed interface technologies such as PCIe Gen4, USB 4.0 and 2x25 GbE connectivity and integrated MIPI-CSI vision capabilities
Target markets include medical, automation, transportation and autonomous mobility, as well as vision based inspection and video surveillance systems.
Congatec was instrumental in the development of the COM-HPC standard, ratified last month, which allows the higher performance processors such as Tiger Lake to be used.
“Our new COM-HPC starter set puts engineers in the fast lane to Gen4 interface technology designs and further ultra fast connectivity,” said Martin Danzer, Director Product Management at congatec speaking at the Embedded World exhibition. “PCIe Gen4 effectively doubles the throughput per lane compared to Gen3, which has massive effects on system designs as it enables engineers to double the number of connected extension devices. Handling all this under more complex design rules to achieve signal compliance makes it even more important to have an evaluation and benchmark platform for own system designs.”
The Ethernet configuration options range from 8x 1GbE switching options and 2x 2.5 GbE including TSN support up to up to dual 2x 10 GbE connectivity congatec’s comprehensive AI support for MIPI-CSI connected cameras from Basler adds further application readiness to industrial IoT and Industry 4.0 connected embedded systems.
Edge AI and inferencing acceleration can be achieved with Intel DL Boost running on the CPU vector neural network instructions (VNNI), or with 8bit integer instructions on the GPU (Int8). Congatec supports Intel’s Open Vino ecosystem for AI, which comes with a library of functions and optimized calls for OpenCV and OpenCL kernels to accelerate deep neural network workloads across multiple platforms.