Week In Review: Design, Low Power

Tools
Aldec extended its TySOM family of embedded prototyping boards with the introduction of TySOM-M-MPFS250, the first in a planned series to feature a Microchip PolarFire SoC FPGA MPFS250T-FCG1152 and to have dual FMC connectivity. The board contains 16Gb FPGA DDR4 x32, 16Gb MSS DDR4 x36 with ECC, eMMC, SPI Flash memory, 64 Kb EEPROM and a microSD card socket. The PolarFire SoC is a five-stage single issue in-order pipeline with RISC-V CPUs from SiFive.

Corigine debuted its MimicPro prototyping systems based on Xilinx UltraScale FPGAs. The prototying system includes automated partitioning with a system-level view that enables capabilities such as monitoring and force/release for debugging and functional safety prototyping.

IP
Rambus completed its acquisition of AnalogX, adding SerDes technology specifically built for ultra-low power and very low latency. “We are excited to welcome the AnalogX team to the Rambus family,” said Luc Seraphin, president and CEO of Rambus. “Their technology and expertise are an ideal fit for Rambus and accelerate our roadmap for next-generation data center interconnect solutions.” Terms of the deal were not disclosed.

In a blog, Arm CEO Simon Segars argued in favor of the company’s acquisition by Nvidia and pushed back against the suggestion from Qualcomm and others that the company go public. “We contemplated an IPO but determined that the pressure to deliver short-term revenue growth and profitability would suffocate our ability to invest, expand, move fast and innovate.” Arm had previously been a public company before its acquisition by SoftBank in 2016.

Sondrel announced an IP reference platform aimed at speeding creation of high-performance, battery-powered IoT devices. The SFA 100 IP reference platform is based on the Arm Corstone-300 subsystem for security and includes an Ethos-U55 Machine Learning processor for applications such as voice activation, image classification, gesture recognition, filtering, inference, and tracking.

Power
Amber Solutions is teaming up with Infineon Technologies to adapt the power management architecture in specific product categories, such as smart circuit breakers, light switches, outlets, to use Amber’s dynamic power management silicon chip and Infineon’s power switch and control technologies. In addition, the two companies will explore the integration of some of Amber’s proprietary architecture with certain Infineon product roadmaps.

EnABLES, a European Research Infrastructure project, is urging EU and industry leaders to improve the power efficiency of IoT devices and boost battery lifetimes so they outlive the devices they power as it warns that 78 million batteries will be dumped every day by 2025. “We need to think about battery life from the outset, in the early stages of product design,” said Mike Hayes, EnABLES project coordinator and Head of ICT for Energy Efficiency at Tyndall. The group is also looking into ways to harvest ambient energy to power IoT sensors. The project is coordinated by Tyndall National Institute with CEA-Leti, CEA-Liten, Fraunhofer IIS, Fraunhofer IMS, imec, Karlsruhe Institute of Technology, Polytechnic University of Turin, University of Bologna, University of Perugia and University of Southampton.

HPC
Nvidia launched a new supercomputer, Cambridge-1, which it says is the UK’s fastest to date. Based on the SuperPOD architecture, it will has 80 DGX A100 systems, 20 terabytes/sec InfiniBand, 2 petabytes of NVMe memory, and requires 500KW of power while providing 400 petaflops of AI performance and 8 petaflops of Linpack performance. The first projects will be run by AstraZeneca, GSK, Guy’s and St Thomas’ NHS Foundation Trust, King’s College London, and Oxford Nanopore Technologies and will focus on life sciences and medical research, including developing a deeper understanding of brain diseases like dementia, using AI to design new drugs, and improving the accuracy of finding disease-causing variations in human genomes.

New Zealand eScience Infrastructure, University of Auckland, University of Otago, and Manaaki Whenua – Landcare Research are planning to double the performance and significantly extend the computational capabilities of the Mahuika HPC cluster, with the assistance of HPE. The upgrades to Mahuika, which came online in 2018, will allow it to be used by a wider range of researchers by using an AMD Milan architecture as well as Nvidia HGX 80GB A100 GPUs.