Deep learning support
FP32, TF32, FP16, BF16, FP8 advanced AI data types
High Performance Acceleration for GAI and LLMs
The Gaudi2 deep learning accelerator enhances deep learning performance and operational efficiency for training and running AI models, from computer vision and NLP to the most complex language and multimodal models. Whether in the cloud or in the data center Gaudi2’s efficient scaling gives the AI industry the flexibility it needs to tackle a growing spectrum of cutting-edge applications.
Gaudi®2: Scalable Performance for AI and Deep Learning
Massive, flexible scale-out
with 24x 100 Gigabit Ethernet (RoCEv2) ports
Independent media engine
decodes and postprocesses compressed media directly
Gaudi®2 Architecture Features
- 7nm process technology
- Heterogeneous compute
- 24 Tensor Processor Cores
- Dual matrix multiplication engines
- 24 100 Gigabit Ethernet integrated on chip
- 96 GB HBM2E memory on board
- 48 MB SRAM
- Integrated Media Control
Aivres AI Servers Supporting Habana® Gaudi®2
Aivres servers amplify deep learning capabilities with the Gaudi2® deep learning accelerator, purpose-designed for high-efficiency, high-performance training and deployment of large-scale workloads.
KR6298V2
Up to 8 Habana® Gaudi®2 OAMs
Supporting 4th Gen Intel® Xeon® Scalable processors
Learn More
