Exciting news: HPE sets multiple AI inference world records and a first for industry

[ComputeExperts]

We’re proud to announce that HPE has achieved multiple outstanding performance benchmark records in recently conducted MLCommons[i] AI benchmarking studies.

Achieving world record results require a high-performance infrastructure. The MLPerf™ Inference: Datacenter Benchmark Suite, part of MLCommons set of measurement and assessment tools, establishes industry standards to ensure fair comparisons of artificial intelligence and machine learning performance.

We are proud to announce that the HPE ProLiant Compute DL384 Gen12 with NVIDIA GH200 NVL, and the HPE ProLiant DL380a Gen11 with four NVIDIA H100 NVL, were both submitted and achieved top results on over 10 of the benchmark scenarios. The HPE ProLiant Compute DL384 Gen12, announced at HPE Discover in June of 2024, is the first server to deliver these results with NVIDIA™ GH200 NVL with 144GB HBM3e memory.

Notable results include the following:

• #1: Text-to-image generative AI: Stable Diffusion XL (SDXL). An advanced image generation model that produces high‑quality, detailed images from text descriptions.
• #1: Deep learning recommendation model: DLRM-v2-99. A model designed for high‑accuracy prediction tasks.
• #1: Large Language Model: Llama 2 70B. A 70 billion parameter large language model (LLM) used for chat and Q&A applications.
• #1: Object detection: RetinaNet. A deep learning model known for its ability to detect objects in images at different scales and sizes.

These achievements represent just a few features that highlight the strength and exceptional capabilities of the HPE ProLiant Compute DL384 Gen12 and the HPE ProLiant DL380a Gen11, part of NVIDIA AI Computing by HPE, a collaboration created to deliver co-developed enterprise AI solutions and joint go-to-market integrations that assist businesses in streamlining the development and deployment of AI applications from pilot to production.

For more detailed information, check out our world record infographic and explore the MLCommons results interactive table.

If you want to find additional information about these powerful AI servers, check out HPE ProLiant Compute DL384 Gen12 and HPE ProLiant DL380a Gen11.

Meet HPE Blogger Sonja Hickey

Sonja is currently a product marketing manager at Hewlett Packard Enterprise (HPE), focusing on HPE ProLiant servers. Working in the IT industry since 1997, Sonja has extensive marketing and product management experience with enterprise software and technology companies, including, but not limited to, HPE, Dell, Sun Microsystems and Zebra Technologies. In 2011, Sonja co-authored a book (IT Operations Management), which discusses best practices associated with IT infrastructure management, especially as they relate to cloud and virtualized environments.  Sonja’s education includes an MBA from University of Chicago’s Booth School as well as a MS and BS in Engineering from the University of Illinois in Champaign-Urbana.

Compute Experts
Hewlett Packard Enterprise

twitter.com/hpe_compute
linkedin.com/showcase/hpe-servers-and-systems/
hpe.com/servers

[i] An Open AI Engineering Consortium: MLCommons is an Artificial Intelligence engineering consortium, built on a philosophy of open collaboration to improve AI systems. Through our collective engineering efforts with industry and academia we continually measure and improve the accuracy, safety, speed, and efficiency of AI technologies–helping companies and universities around the world build better AI systems that will benefit society.