HPE ProLiant Compute DL384 Gen12 with NVIDIA GH200 NVL2: Accelerating Generative AI

Enterprises are increasingly turning to Generative AI (GenAI) technologies to drive innovation and efficiency across sectors like finance, healthcare, retail, and manufacturing. The HPE ProLiant Compute DL384 Gen12, equipped with NVIDIA GH200 NVL2 GPUs, is part of the NVIDIA AI Computing by HPE portfolio, designed to tackle the growing demands of large language models (LLMs) and GenAI applications such as text generation, translation, visual content creation, and drug discovery. View the solution brief

Addressing Enterprise AI Challenges

While GenAI offers transformative potential, it also introduces several challenges:

• Data privacy during model tuning and inference
• Scaling computational resources efficiently
• Securing enterprise data throughout AI operations
• Managing AI outputs to ensure quality and compliance

Key Features of HPE ProLiant DL384 Gen12

• Powered by NVIDIA GH200 NVL2 GPUs: This cutting-edge system includes up to 1.2 TB of fast, unified, coherent memory, optimized for compute- and memory-intensive workloads, ensuring 2x the inference performance compared to previous-generation NVIDIA H100 GPUs.
• Optimized for AI and LLM workloads: This system is built for both AI inference and fine-tuning of models like Retrieval Augmented Generation (RAG), delivering up to 3.5x more memory capacity and 3x higher bandwidth.
• Enhanced Flexibility: Whether you’re processing massive datasets or training sophisticated models, the DL384 Gen12’s architecture can handle mixed or memory-intensive workloads, including AI inferencing and RAG model fine-tuning.
• Seamless Integration with HPE GreenLake: By utilizing HPE GreenLake Flex Solutions, enterprises can scale their AI infrastructure on-demand while enjoying the flexibility and cost efficiency of cloud-based services—without sacrificing on-premises control.

Unlocking the Full Potential of GenAI

HPE ProLiant Compute DL384 Gen12 empowers enterprises to harness the full power of generative AI. With 1.2 TB of fast, unified memory, this platform is built to handle the most demanding AI tasks—from LLM-based simulations to real-time inference for smart city management.

processor configurations:

The HPE ProLiant Compute DL384 Gen12 is available with two processor configurations, each tailored to different workloads and budgets:

1P (Single Processor) Configuration:

• 72-core Grace CPU: Ideal for lighter, yet still compute-intensive, AI tasks.
• 4 PFLOPS Hopper GPU: Provides robust performance for AI inferencing and some model fine-tuning.
• 144 GB HBM3e Memory: Optimized for memory-heavy applications with high bandwidth.
• 5 TB/s bandwidth: Handles fast data processing with 900 GB/s NVLink-C2C for seamless communication between CPU and GPU.
• 480 GB LPDDR5X memory: Enhances memory efficiency for AI-driven workloads.

2P (Dual Processor) Configuration:

• 144-core Grace CPU: Suitable for large-scale AI tasks, offering double the compute power.
• 8 PFLOPS Hopper GPU: Maximized performance for handling complex AI models, including large-scale simulation and real-time AI tasks.
• 288 GB HBM3e Memory: Provides extensive memory to support the heaviest data-processing requirements.
• 10 TB/s bandwidth: Doubles data throughput for larger-scale AI applications, with 2x 900 GB/s NVLink-C2C for optimal CPU-GPU coordination.
• 960 GB LPDDR5X memory: Enhanced memory capacity for high-demand AI and machine learning workloads.