SiMa.ai and Supermicro team up to boost power-efficient edge machine learning

0
84

SiMa.ai announced a new strategic partnership and integration with Supermicro, a Total IT Solution Provider for AI, Cloud, Storage, and 5G/Edge. The SiMa.ai and Supermicro Edge ML Server integration brings together SiMa.ai’s purpose built Machine Learning System-on-a-Chip (MLSoC) that delivers high performance and power efficient edge ML inference for enterprise, industrial and field deployments, and Supermicro’s SYS-E300-13AD high performance compact edge server. The integration provides customers a scalable approach for increasing video channels without scaling the host CPU performance. This provides for in-field upgrades for additional video channel support without swapping out the server or motherboard, which is an important consideration for deployed servers. 

The SiMa.ai MLSoC Platform is transforming the embedded edge market. SiMa.ai’s MLSoC hardware combined with its Palette Software delivers a purpose-built platform with model compilation in minutes, enabling effortless ML deployment at the embedded edge that offers 10x better performance compared to competitive compiled solutions. The SiMa.ai MLSoC features best in class efficiency inferencing in frames per second per watt (fps/watt) as validated by ML Commons MLPerf benchmarks. 

The Supermicro SYS-E300-13AD is a versatile high-performance IoT/Edge server powered by 13th/12th Generation Intel Core processors. Its small physical footprint and superior acoustics make it an ideal solution for a wide range of edge AI use cases such as smart vision applications like theft management, access control, employee safety, inventory management and more. Integrated with SiMa’s Half-Height Half-Length PCIe card, it can deliver over 50 TOPS of ML inference compute performance in a compact rackmount server platform. SiMa.ai’s Palette Software can be hosted on the server platform as a Docker container running on the popular Ubuntu Linux operating system, enabling updates to the models and pipelines even when not connected to a network. 

SiMa.ai and Supermicro integrated their technologies to deliver a compact Edge ML Server, supporting multiple video cameras via Ethernet and wireless remote video cameras via PCIe. This SiMa.ai enabled Edge ML Server provides for multi-stream video analytics to be processed locally, reducing total cost of ownership, while increasing reliability and security. The Edge ML Server design provides a compute environment to process and analyze multiple video streams and provide edge intelligence for many enterprise use cases, to facilitate automation in retail, manufacturing and smart city solutions. The Edge ML Server can be augmented with additional remote and cloud based compute resources to provide intelligence across the enterprise. 

“The Supermicro – SiMa.ai solution brings together a remarkable fusion of hardware and software innovation,” said Raju Penumatcha, Senior Vice President and Chief Product Officer at Supermicro. “SiMa.ai’s complete ML pipeline on a chip eliminates the host CPU bottlenecks, providing an integrated scalable solution for multi-channel video applications with lower power at the intelligent edge.” 

With this announcement, Supermicro joins the SiMa Partner Program as an OEM partner member and will deliver the SiMa.ai and Supermicro Edge ML Server solution to customers and System Integrators (SIs) as a pre-integrated solution. The integration validates the production quality of the SiMa.ai MLSoC platform in providing edge ML enablement to industry leading remote servers. 

“Partner enablement is a critical step in creating an ecosystem of edge ML design vehicles for customers. With Supermicro now on board as an OEM partner at SiMa.ai, we have successfully demonstrated an integration that can scale across the Supermicro product line based on customer needs,” said Elizabeth Samara Rubio, Chief Business Officer, SiMa.ai. “We will continue to meet customer demand for making AIML accessible by introducing new OEM integrations that solve customers’ edge specific AIML development challenges in an effortless way.”