Now Reading
Microsemi enhances Switchtec PCIe switch for fabric connectivity and composability of multi-host GPU and NVMe SSD systems
0

Microsemi enhances Switchtec PCIe switch for fabric connectivity and composability of multi-host GPU and NVMe SSD systems

Posted by Zenobia HegdeMarch 16, 2018

Microsemi Corporation, a provider of semiconductor solutions differentiated by power, security, reliability and performance, announced enhancements to its Switchtec PCIe advanced Fabric Switch (PAX), which now provides high-performance fabric connectivity and composability for multi-host graphics processing unit (GPU) and NVMe solid state drive (SSD) systems. Microsemi will be showcasing a demonstration of the new device during the Open Compute Project (OCP) Summit March 20-21, 2018 in San Jose, California.

As the use of GPUs for deep learning, artificial intelligence and machine learning make groundbreaking improvements in areas of image classification, speech recognition, autonomous driving, bioinformatics and video analytics, the need for efficient parallel GPU computing continues to grow.

Microsemi’s Switchtec PCIe advanced fabric switches provide a scalable, low-latency and cost-effective multi-host interconnect for a network of GPUs, NVMe SSDs and other PCIe endpoints. The Switchtec PAX product can be used to create a flexible fabric interconnect to a shared pool of GPUs and/or NVMe SSDs using standard operating system drivers. System management is simplified through a fabric application programming interface (API), significantly reducing time to market and development cost for multi-host systems.

Andrew Dieckmann

“Our Switchtec PAX switch enables GPU resources to be allocated to match workload requirements and maximise efficiency,” said Andrew Dieckmann, vice president of marketing for data centre solutions at Microsemi. “Using the dynamic partitioning and multi-host SR-IOV sharing features available on our Switchtec PAX switches, GPU resources can be ‘composed’ or dynamically allocated to a specific host or set of hosts in real-time using standard host drivers.”

According to market research firm IndustryARC, the GPU machine learning market is expected to grow at a compound annual growth rate (CAGR) of nearly 38% till 2024. The firm also notes advanced technologies such as low power consumption, high throughput and flexibility will drastically increase the adoption of machine learning applications across the globe.

Microsemi’s PAX advanced fabric PCIe switch will be showcased during the OCP Summit March 20-21, 2018 at the Hilton San Jose.

Comment on this article below or via Twitter: @IoTNow_OR @jcIoTnow

About The Author
Zenobia Hegde

Leave a Response