NVIDIA CUDA-X AI and HPC Software Stack Now Available on Marvell ThunderX Platforms

January 11, 2020

Marvell (NASDAQ: MRVL) today introduced the supply of NVIDIA GPU help on its ThunderX® own family of Arm®-based totally server processors.  Following NVIDIA’s June announcement to convey CUDA to the Arm architecture, Marvell has collaborated with NVIDIA to port its CUDA-X AI™ and HPC libraries, GPU-extended AI frameworks and software improvement equipment to the ThunderX platform.  The computational overall performance and memory bandwidth of ThunderX2®, Marvell’s latest sixty four-bit Armv8-A based totally server processor, combined with the parallel processing competencies of NVIDIA GPUs offer a compelling route to energy-efficient exascale computing.

Artificial intelligence (AI) and device gaining knowledge of (ML) hold to turn out to be crucial technology components to records center server requirements on the cloud and community side.  To deal with those evolving AI and ML workloads, in addition to the maximum hard and complicated issues in technology and research, supercomputers need processors which might be optimized to offer contemporary throughput, utility latency and electricity.

With an preliminary consciousness on computational technology packages which include GROMACS, NAMD, MILC and LAMMPS, the ThunderX2 configurations are demonstrating compelling performance with an greater ability to pressure better and more efficient blended software effects in a GPU-enabled device.

“NVIDIA GPU help for our ThunderX2 server processor brings clear, differentiated price to fulfill the extraordinary overall performance and energy requirements of the exascale computing era,” stated Gopal Hegde, vice president and popular manager, Server Processor Business Unit at Marvell Semiconductor, Inc.  “The availability of NVIDIA GPU acceleration on the Arm architecture will similarly expand the ThunderX2 ecosystem for HPC, cloud computing and part markets, spurring innovation across low level firmware through machine software to commercial ISV applications.”

Read More : >“The availability of CUDA acceleration for ThunderX2 processors marks a big milestone in bringing the electricity efficiency and excessive performance of the Arm architecture to the infrastructure market,” stated Chris Bergey, senior vice president and widespread supervisor, Infrastructure Line of Business at Arm.  “The breadth and intensity of innovation across the surroundings allows an easy migration route and strong support for current and destiny GPU workloads from the brink to the cloud.”

“NVIDIA GPU-improved computing on Arm offers customers international with greater choice in constructing next-gen AI-enabled supercomputers,” stated Ian Buck, trendy supervisor and vp of Accelerated Computing at NVIDIA. “Combining NVIDIA’s unmatched platform for AI and HPC with Marvell’s powerful ThunderX2 Arm-based server processors is already delivering marvelous utility performance.” 

Ecosystem Support

ThunderX2 is the maximum widely supported Armv8-A server processor with an ecosystem of over one hundred companions throughout industrial, open supply and industry standards engagements.  NVIDIA’s full software suite support is permitting the acceleration of more than 600 HPC packages and AI frameworks on ThunderX2 structures.

“Our collaboration with Marvell allows us to assist servers with the industry-main performance of ThunderX2 with our SC8000 compute acceleration expansion platform, bringing information center AI abilties to a host of aspect applications,” said Steve Cooper, CEO at One Stop Systems.  “The SC8000 is the industry’s first answer that consists of NVIDIA Tesla GPUs with NVLink and Arm servers. The addition of Arm-primarily based structure into our solutions extends the cost of use instances for AI at the Fly area appliances for our customers.”

“Red Hat and Marvell have a long records of participating inside the Arm server surroundings, helping to carry open, enterprise-extensive standards to business enterprise Arm structure,” said Chris Wright, senior vp and chief era officer at Red Hat.  “Enabling NVIDIA GPUs on ThunderX-based totally systems paired with the CUDA-X SDK and libraries supports purchaser desire in terms of structure for going for walks HPC, AI and ML packages on pinnacle of Red Hat systems.”

“SUSE and Marvell have a a success track file of sturdy HPC collaboration as seen maximum these days inside the Catalyst UK deployment,” said Daniel Nelson, vp, Products & Solutions Management at SUSE.  “We are excited to build upon this collaboration with the addition of NVIDIA’s GPUs and suite of HPC, AI and ML applications guide on Marvell’s ThunderX2.   SUSE looks forward to riding additional optimizations and software infrastructure support within the Arm environment and might be demonstrating GPU support at the SC19 industry convention in Denver.”

The Marvell® ThunderX platform with NVIDIA GPU help can be verified on the Marvell sales space (#349) at The International Conference for High Performance Computing, Networking, Storage and Analysis (SC19), walking via November 22, 2019 in Denver, Colorado.

Tags