EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·28d ago·5 min read
HP ZGX Fury Is Now Orderable: GB300 Superchip, 748GB Unified Memory
TL;DR
HP has launched the ZGX Fury AI station, featuring the GB300 Superchip and 748GB of unified memory, designed for shared AI inference tasks. This system aims to extend enterprise AI capabilities from data centers to edge environments.
✦ Why It Matters
Consider evaluating the ZGX Fury for your department's AI inference needs today.
Key Takeaways
- Explore the ZGX Fury for shared AI inference capabilities.
- Utilize the 748GB memory for large model training and inference.
- Leverage Red Hat AI Factory for streamlined AI deployment.
- Consider the system's edge computing potential for your organization.
How It Works
The ZGX Fury integrates a 72-core Grace CPU with a Blackwell Ultra GPU, allowing direct memory access between CPU and GPU. This architecture supports efficient processing of large AI models, enabling trillion-parameter inference and fine-tuning on a single machine.