Press Release | 
March 18, 2024

At NVIDIA GTC 2024: LIQID Unveils UltraStack Whitepaper Showcasing Breakthrough AI Server Performance

LIQID, a leader in composable infrastructure technology, announced today at NVIDIA GTC, a global AI conference, the release of its technical whitepaper for the groundbreaking UltraStack AI server with the latest MLPerf benchmark results showing exceptional performance.
At NVIDIA GTC 2024: LIQID Unveils UltraStack Whitepaper Showcasing Breakthrough AI Server Performance

SAN JOSE, Calif.--(BUSINESS WIRE)--LIQID, a leader in composable infrastructure technology, announced today at NVIDIA GTC, a global AI conference, the release of its technical whitepaper for the groundbreaking UltraStack AI server with the latest MLPerf benchmark results showing exceptional performance.

UltraStack Whitepaper: Setting New Standards in AI Computing

1. Unprecedented GPU Density and Performance:

  • UltraStack’s integration of up to 20x NVIDIA L40S GPUs delivers exceptional computational efficiency for AI applications, surpassing traditional server limitations.

2. Cost Efficiency and Simplified Infrastructure:

  • The UltraStack design significantly reduces the total cost of ownership while enhancing computational performance, making it an economical choice for AI-driven enterprises.

3. Universal GPU Capability for Diverse Workloads:

  • UltraStack is equipped to handle a wide array of data center workloads, including generative AI, compute, 3D graphics, rendering, and video processing. Featuring NVIDIA L40S GPUs, UltraStack delivers powerful performance to accelerate NVIDIA’s full software stack, including the NVIDIA AI Enterprise and NVIDIA Omniverse platforms.

4. Exceptional Benchmark Performance:

  • MLPerf 3.1 Inference LLM (Large Language Model) Results with 16x L40S GPUs: UltraStack scored an impressive 94 queries/sec at ~7,000W.
  • Additional MLPerf 3.1 Inference tests with 16x L40S GPUs showed: Object Detection: 10,104 queries/sec; Medical Imaging: 62.6 queries/sec; Natural Language Processing: 44,730 queries/sec.
  • These impressive single server results highlight UltraStack’s capability in efficiently managing various AI workloads.
  • The full performance data is available in the whitepaper.

Sumit Puri on UltraStack’s Achievement

“The benchmark performance results for our UltraStack AI servers are a testament to our unwavering commitment to providing the most reliable, high-performance AI computing solutions on the market,” said Sumit Puri, President, Chief Strategy Officer & Cofounder, LIQID. “The breakthroughs documented in our latest whitepaper underscore our commitment to pushing the boundaries of AI computing, providing our customers with the most advanced and efficient AI servers available today.”

LIQID and Dell Technologies: Forging the Future of AI

This innovation in AI infrastructure is a result of the strategic partnership between LIQID and Dell Technologies, merging LIQID’s advanced CDI technology with Dell’s robust R series PowerEdge servers. This collaboration is reshaping the AI infrastructure landscape, setting new benchmarks for performance and efficiency.

Availability and Additional Information

UltraStack is now available in 8-, 10-, 16-, and 20-GPU configurations, setting a new standard in AI server performance. For an in-depth look at UltraStack’s capabilities, read Architecting the Future of AI: The UltraStack Approach. Stay connected with LIQID on LinkedIn for the latest news and insights.

Speak With An Expert