Intel® Gaudi® 3 AI accelerator for high-performance high-efficiency Gen AI compute.
Learn moreBuilt and Architected for
Deep Learning Training and Inference.
The Intel® Gaudi® AI acceleration platform was conceived and architected to address training, fine-tuning and inference demands of large-scale era AI, providing enterprises and cloud providers with high-performance, high-efficiency deep learning compute.
CHALLENGING INDUSTRY-LEADING PERFORMANCE
Deep Learning Compute Performance
Based on MLPerf results,
Intel® Gaudi® 2 AI accelerators are the ONLY viable alternative to Nvidia’s H100 for training large language models like GPT-3*
Intel® Gaudi® 2 AI accelerators are the ONLY viable alternative to Nvidia’s H100 for training large language models like GPT-3*
*MLPerf GPT-3 Benchmark, Dec, 2023
PRICE-PERFORMANCE EFFICIENCY
Efficient Design
Lowers Cost
Intel® Gaudi® AI architecture has demonstrated up to 40% better price/performance on Amazon EC2 instances.
MASSIVE SCALABILITY
Near-linear Scalability
With 24 200 GbE ports
integrated onto every
Intel® Gaudi® AI accelerator,
system scale-out is massive,
flexible and cost-effective.
integrated onto every
Intel® Gaudi® AI accelerator,
system scale-out is massive,
flexible and cost-effective.
EASY-TO-USE DEVELOPMENT PLATFORM
Migrate Models with Ease and Support
Our Intel® Gaudi® Software optimized for performance and
ease of use, lowers the bar to
model build and migration.
ease of use, lowers the bar to
model build and migration.
Intel® Gaudi® AI Accelerators
Designed at inception for the Large-Scale Era.
Intel® Gaudi® AI accelerator
First-generation delivering
price/performance leadership for both deep learning and inference.
Learn More
price/performance leadership for both deep learning and inference.
Intel® Gaudi® 2 AI accelerator
Second-generation delivering competitive performance and efficiency for deep learning training and inference.
Learn More
Intel® Gaudi® 3 AI accelerator
Third-generation delivering competitive performance for Gen AI enterprise applications.
Learn More
