Inspur, Baidu launch world’s first OAI-compliant open AI computing solution
-
The world's first Open Accelerator Infrastructure (OAI)-compliant and
liquid cooling rack-scale AI computing product, X-MAN 4.0, is jointly
launched by Baidu and Inspur
- Inspur launches the first 21” reference system supporting the OAI Universal BaseBoard (UBB) specification
Inspur, a data centre and artificial intelligence (AI) full-stack solutions provider, has announced two AI-driven open computing systems. The X-MAN 4.0, developed with Baidu, is the world’s first OAI-compliant and liquid-cooled rack-scale AI computing product optimised for deep neural network applications, the OCP Platinum Member and OCP Solution Provider said.
The hardware needed to run AI is becoming more complex, Inspur says, with AI accelerator integration typically taking six to 12 months. The OAI specification—led by Baidu, Facebook and Microsoft in the Open Compute Project (OCP) community—centralises the technical specifications of the accelerator module and simplifies the design complexity of AI accelerators, thereby shortening time to market.
X-MAN 4.0 is the fourth generation of Baidu’s X-MAN series of full-rack AI computing products and provides performance, flexibility, cost and other advantages over traditional graphics processing unit (GPU) servers.
X-MAN 4.0 has eight AI accelerators in a single box and scales to 32 AI accelerators per rack. The boxes are interconnected through Quad Small Form Factor Pluggable Double Density (QSFP-DD) cables to minimise communication latency across nodes.
X-MAN 4.0 is the first in Baidu’s series of full-rack AI computing products to introduce OCP Accelerator Module (OAM)-compliant accelerators. Accelerator resources can be software-defined. Together, these features mean flexible, multivendor support for supporting AI applications of different workloads.
The newly-launched Inspur OAI UBB system, meanwhile, is a 21" full-rack OAM solution delivering breakthrough efficiency, flexibility and management. The solution supports two out of the three OCP interconnect topologies: Hybrid Cube Mesh (HCM) and Fully Connected (FC) and provides simpler intermodule communication to scale up, and input/output bandwidth to scale out.
“Inspur continues to pioneer the technologies that enable next-generation AI applications,” said Peter Peng, Inspur Group Senior VP.
“As a member of the OCP, Open 19 and ODCC global open computing standards organisations, Inspur has always been an active promoter of open source technology to help users build open data centres. We work to facilitate cooperation among all key open standards to help our customers accelerate business innovation and more efficiently and effectively bring about next-generation AI applications.”
The open source I-Flex standard server design and the approval for another Inspur platform were also announced at the OCP Regional Summit. Inspur's latest OCP-approved platform, the NF8380M5, is a 3U four-socket server based on the Project Olympus normative four-socket server developed for the Intel Purley Platform. Suitable for multi-scenario AI applications, the NF8380M5 can drive GPU acceleration training workloads when serving as a compute head node. It also supports all NVMe solid state drive (SSD) configurations for real-time decision database applications. Other Inspur’s OCP-approved products include the NF8260M5, ON5263M5, ON5266M5and ON5388M5.
Inspur provides data centre infrastructure, cloud computing, and AI solutions, and is a global top-three server manufacturer.
- Inspur launches the first 21” reference system supporting the OAI Universal BaseBoard (UBB) specification
Inspur, a data centre and artificial intelligence (AI) full-stack solutions provider, has announced two AI-driven open computing systems. The X-MAN 4.0, developed with Baidu, is the world’s first OAI-compliant and liquid-cooled rack-scale AI computing product optimised for deep neural network applications, the OCP Platinum Member and OCP Solution Provider said.
 |
| Source: Inspur. The X-MAN 4.0. |
The hardware needed to run AI is becoming more complex, Inspur says, with AI accelerator integration typically taking six to 12 months. The OAI specification—led by Baidu, Facebook and Microsoft in the Open Compute Project (OCP) community—centralises the technical specifications of the accelerator module and simplifies the design complexity of AI accelerators, thereby shortening time to market.
X-MAN 4.0 is the fourth generation of Baidu’s X-MAN series of full-rack AI computing products and provides performance, flexibility, cost and other advantages over traditional graphics processing unit (GPU) servers.
X-MAN 4.0 has eight AI accelerators in a single box and scales to 32 AI accelerators per rack. The boxes are interconnected through Quad Small Form Factor Pluggable Double Density (QSFP-DD) cables to minimise communication latency across nodes.
X-MAN 4.0 is the first in Baidu’s series of full-rack AI computing products to introduce OCP Accelerator Module (OAM)-compliant accelerators. Accelerator resources can be software-defined. Together, these features mean flexible, multivendor support for supporting AI applications of different workloads.
The newly-launched Inspur OAI UBB system, meanwhile, is a 21" full-rack OAM solution delivering breakthrough efficiency, flexibility and management. The solution supports two out of the three OCP interconnect topologies: Hybrid Cube Mesh (HCM) and Fully Connected (FC) and provides simpler intermodule communication to scale up, and input/output bandwidth to scale out.
“Inspur continues to pioneer the technologies that enable next-generation AI applications,” said Peter Peng, Inspur Group Senior VP.
“As a member of the OCP, Open 19 and ODCC global open computing standards organisations, Inspur has always been an active promoter of open source technology to help users build open data centres. We work to facilitate cooperation among all key open standards to help our customers accelerate business innovation and more efficiently and effectively bring about next-generation AI applications.”
The open source I-Flex standard server design and the approval for another Inspur platform were also announced at the OCP Regional Summit. Inspur's latest OCP-approved platform, the NF8380M5, is a 3U four-socket server based on the Project Olympus normative four-socket server developed for the Intel Purley Platform. Suitable for multi-scenario AI applications, the NF8380M5 can drive GPU acceleration training workloads when serving as a compute head node. It also supports all NVMe solid state drive (SSD) configurations for real-time decision database applications. Other Inspur’s OCP-approved products include the NF8260M5, ON5263M5, ON5266M5and ON5388M5.
Inspur provides data centre infrastructure, cloud computing, and AI solutions, and is a global top-three server manufacturer.
Comments
Post a Comment