Lambda is a Diamond Sponsor at NVIDIA GTC!
Co-authored by Inaki Madrigal, Solutions Architect, CSP, NVIDIA
NVIDIA GTC provides the premier opportunity for ML Engineers, Researchers & Leaders to connect, exchange knowledge, and stay at the forefront of advancements in AI. With over 900 sessions and 300 exhibits and networking events, GTC delivers something for every technical level and interest area.
Lambda at NVIDIA GTC 2024
Lambda is a Diamond sponsor of GTC, taking place March 18-21 in San Jose, CA. This year, Lambda will showcase our latest cloud innovations, our NVIDIA GH200 Grace Hopper Superchip benchmarks, and how Lambda’s ML team is using Retrieval Augmented Generation (RAG).
The Lambda and NVIDIA Partnership
Lambda is a GPU cloud company founded by AI engineers and focused on providing best-in-class infrastructure for AI, powered by NVIDIA technology. Lambda delivers cloud computing at scale while optimizing infrastructure performance across the full AI stack. Companies choose Lambda for our On-Demand Cloud, which features pay-by-the-hour access to NVIDIA GPUs, and our Reserved Cloud that allows ML teams to secure hundreds to thousands of GPUs for large model training. Lambda runs benchmarks on the latest-and-greatest AI infrastructure technology so our customers can have confidence in their AI compute investments. Lambda was also one of the first cloud providers to market with NVIDIA H100 Tensor Core GPUs and NVIDIA GH200 Superchip-powered systems.
Recent NVIDIA GH200 Grace Hopper Superchip benchmark
Lambda’s recent GH200 Superchip benchmark explores the synergy between ZeRO-Inference technology and the GH200, showcasing its remarkable proficiency in handling large language models (LLMs) and substantially enhancing inference throughput. The GH200 achieves 4-5 times the inference throughput of H100 GPUs and 9-11 times that of NVIDIA A100 Tensor Core GPUs for the Bloom 176B model when offloading model weights to CPU memory in GPU-constrained systems. This is attributed to GH200‘s significantly higher chip-to-chip bandwidth with NVIDIA NVLink-C2C, effectively eliminating the communication bottleneck for CPU offloading. This study also underscores the key role of GH200’s coherent memory in facilitating inference with larger batch sizes. GH200 helps open the way to advanced AI models, offering new possibilities for computational efficiency and scalability. At GTC, Lambda’s presentations will provide deeper insights into our GH200 deployments and findings, as well as our endeavors in developing RAG applications on the NVIDIA AI platform.
Lambda’s Booth at NVIDIA GTC
Come to Booth #616 to meet the Lambda team of ML experts, get a firsthand look at our AI and Machine Learning news bot, learn about building RAG applications using open source models, and delve into discussions on quality, safety, and speed. You can also grab some Lambda swag and enter our giveaway for a chance to win Lambda On-Demand Cloud credits.
Lambda presentations at NVIDIA GTC
Lambda team members are presenting on NVIDIA GH200 Superchip deployments and results, as well as the use of Retrieval Augmented Generation (RAG).
Crafting user experience for RAG
Tuesday, March 19th
Join Corey Lowman, David Hartmann, and Chuan Li from 9 - 9:25am at the Hilton Winchester for a talk on our journey of developing RAG applications using open-source LLMs powered by NVIDIA GPUs, exploring an often overlooked aspect: the art of crafting user experiences.
Deploying one of the first NVIDIA GH200 Grace Hopper Superchip Clusters in Lambda Cloud
Wednesday, March 20th
Join David Hall and Maxx Garrison from 9 - 9:25am in room 211A at the San Jose Convention Center, as they discuss Lambda’s GH200 Superchip-powered, AI training and inference-optimized cluster, including Lambda’s findings on performance.
NVIDIA GTC ancillary events
Lambda is participating in a variety of ancillary events, including presentations during NVIDIA GTC and exclusive after-hours gatherings with other leaders in the AI community. Stop by Booth #616 to learn more about Lambda sessions and events.
Monday, March 18th
Generative AI Welcome Lunch with NVIDIA, Weights & Biases, and Run:ai
Run:ai, Weights & Biases, and Lambda are excited to host an exclusive Generative AI GTC welcome lunch exploring methods for extracting value from LLMs using fine-tuning and RAG on Monday from 11am - 1pm at the Farmers Union.
Evening reception with VAST, Run:ai, & PNY
Kick off NVIDIA GTC with VAST Data, Lambda, and Run:AI, 7pm at Noite. Indulge in handcrafted cocktails, hors d’oeuvres, and engaging conversations with fellow industry pioneers, visionaries, and tech enthusiasts.
Tuesday, March 19th
AI Data Summit: Faster and Safer GPU ROI Acceleration in Data Centers and the Cloud
Join DDN with NVIDIA, Lambda, and others at this special event on Tuesday at 9am at the Hilton San Jose in the San Pedro Room, Floor 2, and explore how to quickly deploy and scale GPU infrastructure and how those deployments can be made more efficient and cost-effective. Attendees will also learn about case studies across multiple markets and use cases deployed on premises and in the cloud.
Wednesday, March 20th
NVIDIA GH200 MGX presentation with QCT
Lambda’s VP of NVIDIA Solutions, David Hall, joins QCT to discuss valuable insights into groundbreaking technologies. See firsthand how QCT’s collaboration with NVIDIA is helping shape the future of top-tier data center excellence from 2 - 2:25pm at the Hilton Winchester.
Happy Hour with Gradient + CentML
Ready to debrief on a few days full of cutting-edge AI developments? Join Gradient, Lambda, and CentML on Wednesday at District San Jose from 5 - 8pm for an exclusive happy hour to unwind, connect, and continue the exciting discussions.
We look forward to seeing you at NVIDIA GTC!