Terryl Dickey
29 Aug 2024 15:10
CoreWeave becomes the first cloud provider to offer NVIDIA H200 Tensor Core GPUs, boosting AI infrastructure performance and efficiency.
According to PRNewswire, CoreWeave, an AI Hyperscaler™, has announced a pioneering move to become the first cloud provider to bring the NVIDIA H200 Tensor Core GPU to market. This development represents a significant milestone in the evolution of AI infrastructure, promising improved performance and efficiency for generative AI applications.
Advancement of AI infrastructure
The NVIDIA H200 Tensor Core GPU is designed to push the boundaries of AI capabilities, boasting 4.8 TB/s of memory bandwidth and 141 GB of GPU memory capacity. These specifications deliver up to 1.9x higher inference performance compared to the previous H100 GPU. CoreWeave leverages these advancements by integrating the H200 GPU with Intel’s 5th generation Xeon CPUs (Emerald Rapids) and 3200 Gbps NVIDIA Quantum-2 InfiniBand networking. This combination can be deployed in clusters with up to 42,000 GPUs and accelerated storage solutions, dramatically reducing the time and cost required to train generative AI models.
CoreWeave’s Mission Control Platform
CoreWeave’s Mission Control platform plays a critical role in managing AI infrastructure. It simplifies the complexity of deploying and maintaining AI by providing high reliability and resiliency through software automation. The platform features advanced system validation processes, proactive vehicle health checks, and extensive monitoring capabilities to ensure customers experience minimal downtime and reduced total cost of ownership.
“CoreWeave is committed to pushing the boundaries of AI development,” said Michael Intrator, CEO and co-founder of CoreWeave. “Working with NVIDIA allows us to deliver a high-performance, scalable, and resilient infrastructure using NVIDIA H200 GPUs, enabling our customers to process complex AI models with unprecedented efficiency.”
Expanding Data Center Operations
CoreWeave is rapidly expanding its data center operations to meet the growing demand for advanced infrastructure services. Since early 2024, the company has completed nine new data centers, with 11 more in progress. By year-end, CoreWeave expects to have 28 data centers worldwide, with plans to add 10 more in 2025.
Industry impact
Rapid deployment of NVIDIA technology on CoreWeave will give customers access to the latest advancements in training and running large-scale language models for generative AI. Ian Buck, vice president of Hyperscale and HPC at NVIDIA, emphasized the importance of this partnership: “With NVLink and NVSwitch and enhanced memory capabilities, the H200 is designed to accelerate the most demanding AI workloads. When paired with the CoreWeave platform powered by Mission Control, the H200 provides customers with an advanced AI infrastructure that will be the backbone of innovation across the industry.”
Introducing CoreWeave
CoreWeave, the AI Hyperscaler™, provides a cloud platform for cutting-edge software that powers next-generation AI. Since 2017, CoreWeave has operated a growing number of data centers across the US and Europe. The company has been recognized as one of the TIME 100 Most Influential Companies and is on the 2024 Forbes Cloud 100. For more information, visit www.coreweave.com.
Image source: Shutterstock