SambaNova Systems Ushers in New Era of Computing with Availability of SambaNova DataScale — Built for AI
Announces world record-breaking performance, and the industry’s first Dataflow-as-a-Service offerings
PALO ALTO, Calif., Dec. 9, 2020 — SambaNova Systems, the company building the industry’s most advanced systems platform to develop and deploy next-generation artificial intelligence (AI) and data-intensive applications from the data center to the cloud and the edge, today announced the availability of SambaNova Systems DataScale™. Built on SambaNova Systems Reconfigurable Dataflow Architecture™ (RDA), SambaNova DataScale is optimized for dataflow from the algorithms to the silicon, enabling enterprises to bring new services and products to market faster than today’s state-of-the-art solutions. Also being announced is the industry’s first Dataflow-as-a-Service offering, making the record-breaking performance and accuracy of the SambaNova DataScale system available via a monthly subscription service.
“We are experiencing one of the biggest transitions in computing history since the internet,” said Rodrigo Liang, co-founder and CEO, SambaNova Systems. “This transition is already pushing the boundaries of what is possible with AI and giving us a clear view of what the future of computing will look like. AI has forced the industry to relook at how we design next-generation infrastructure for these complex machine learning workloads. There is a shift beyond transactional processing to AI that is redefining compute requirements beyond raw performance.”
SambaNova DataScale is a completely integrated software and hardware platform delivering unrivaled performance, accuracy, scale and ease of use. The platform’s out-of-the box performance and efficiency uses open source software and open standards to eliminate the need for low-level tuning. SambaNova DataScale is already empowering organizations to quickly and simply advance their AI innovations across a broad range of use cases, including natural language processing (NLP), high-resolution computer vision, recommender systems, AI for science, and more.
World Record-Breaking Performance Metrics in Four Key Areas
SambaNova DataScale achieves world record-breaking performance metrics at multi-rack scale when compared to the latest A100 GPUs in four key areas:
Performance: World record DLRM inference 7x better throughput and latency than A100. World record BERT-Large training 1.4x faster than DGX A100 systems. Accuracy: World record state of the art accuracy of 90.23% out-of-the-box for high-resolution computer vision compared to DGX A100 systems. World record state-of-the-art accuracy of 80.46% for DLRM recommendation engines compared to NVIDIA A100 GPUs. Scale: World record BERT-Large training and state-of-the-art accuracy at multi-rack scale. Ease of Use: From loading dock to data center, SambaNova DataScale quickly and easily integrates into any existing infrastructure running customer workloads in about 45 minutes. Download thousands of pre-trained Hugging Face Transformer models in seconds on SambaNova DataScale at state-of-the-art accuracy with no code changes required.
Born in the Era of AI — Built for Current AND Future Innovations
A Software-First Approach with SambaFlow
SambaNova DataScale features SambaFlow™ the industry’s most advanced software. Taking a software-first approach, SambaFlow eliminates months of tuning and optimizing to allow customers to focus exclusively on the applications. Designed to take input from standard frameworks such as PyTorch and TensorFlow, SambaFlow automatically extracts, optimizes and executes the optimal dataflow graph of any models on SambaNova’s Reconfigurable Dataflow Units™ (RDUs). SambaFlow also eliminates the complexity associated with scaling workloads across multiple processors, enabling users to scale seamlessly from one to any number of RDUs using the same programming model.
Organizations including Argonne National Laboratory, the Department of Energy’s National Nuclear Security Administration (NNSA), Lawrence Livermore National Laboratory (LLNL), and Los Alamos National Laboratory (LANL) are already successfully using SambaNova DataScale to accelerate AI computing initiatives in life sciences, such as cutting years off of core COVID-19 research time.
“At Argonne National Laboratory, we’re working on important research efforts including those focused on cancer, COVID-19, and many others, and using AI to automate parts of the development process is key to our success,” said Rick Stevens, associate laboratory director, Argonne National Laboratory. “The SambaNova DataScale architecture offers us the ability to train and infer from multiple large and small models concurrently and deliver orders of magnitude performance improvements over GPUs.”
Zero to AI with a Simple Subscription
Along with SambaNova DataScale, SambaNova is also announcing the launch of the industry’s first Dataflow-as-a-Service offerings, available in three distinct subscriptions for natural language processing, high-res computer vision, and recommender systems. All three service offerings deliver the same performance and accuracy as the SambaNova DataScale system, but are acquired as a monthly subscription service. The three Dataflow-as-a-Service offerings have all the cost and quick start benefits of cloud consumption models, run completely behind the subscriber’s firewall and are managed by SambaNova. With Dataflow-as-a-Service, SambaNova is delivering on its promise to make AI accessible to organizations across all industries.
Developing, Deploying Next-Generation AI Applications from Data Center to the Cloud Also announced today as part of the SambaNova DataScale launch is SambaNova AI Cloud Platform, a new AI systems platform in the cloud aimed at universities and research laboratories. Through this new offering, university research labs can gain access to resources on the SambaNova AI Cloud Platform to do more advanced research into unique problems and begin building innovative new solutions sooner. Research proposals are being accepted now.
SambaNova DataScale features include:
SambaNova Systems SambaFlow™, a complete and open software stack that provides no lock-in and ease of use to improve developer productivity. SambaNova Reconfigurable Dataflow Unit™ (RDU), the industry’s next-generation processor built from the ground up to offer native dataflow processing. SambaNova Systems RDU-Direct™, high-speed fabric that provides a low-latency, high-bandwidth direct connection between SambaNova Systems RDUs for maximum system throughput. 8-32 SambaNova Cardinal SN10 RDU™ processors per rack. SambaNova DataScale features terabyte-level memory capacities and low-latency interconnect to hundreds of petaflops of compute capability to process massive amounts of complex data models. Additionally, it has built-in virtualization to securely run various training and inference workloads, single large workloads, or multi-tenant workloads.