NVIDIA B300
Optimized for AI Reasoning with Breakthrough Attention Performance
- 2X attention performance over B200 GPUs
- 1.5X dense FP4 performance boost vs B200
- 192 petaFLOPS inference / 70 petaFLOPS training per system
- Specifically designed for the era of AI reasoning models (like o1, o3) that require massive attention compute
Massive Memory Capacity for Trillion-Parameter Models
- 2.1TB total GPU memory (263GB per GPU across 8 GPUs)
- Increased from 1.4TB in B200 — a 50% memory increase
- 14.4TB/s NVLink aggregate bandwidth
- Enables handling the largest frontier models and multi-modal workloads without memory constraints
Data Center-Optimized Form Factor with Flexible Power
- First DGX system compatible with NVIDIA MGX standard racks
- Available in both AC/PDU and DC/busbar configurations for deployment flexibility
- 10U form factor (vs 4U for B200) designed for modern hyperscale datacenter layouts
- ~14kW power consumption with industry-leading efficiency
Use Cases
Large Model Inference
Run massive models with predictable latency. Optimize for throughput, batch size, and performance per watt.
Generative AI applications for text, image, and audio.
Scaling ML infrastructure as your customer base grows.
