The NVIDIA GB200 Grace Blackwell Superchip is a premier AI training and inference module designed for trillion-parameter scale generative AI. It integrates two NVIDIA Blackwell GPUs and one NVIDIA Grace CPU over a 900GB/s ultra-low-latency NVLink-C2C interconnect.
- Architecture: NVIDIA Blackwell with Second-Generation Transformer Engine
- Performance: Up to 20 petaflops of FP4 AI performance per superchip
- Memory: 384GB of high-bandwidth HBM3e memory
- Memory Bandwidth: 16TB/s aggregate GPU memory bandwidth
- Interconnect: Fifth-generation NVLink providing 1.8TB/s bidirectional throughput
- Efficiency: Up to 25x lower TCO and energy consumption compared to previous generations
- Scalability: Designed for integration into GB200 NVL72 rack-scale systems
- Security: Integrated NVIDIA Confidential Computing for sensitive data protection
- Decompression Engine: Dedicated hardware for accelerating data analytics and SQL
NVIDIA
NVIDIA - GB200 - Grace Blackwell AI Superchip Training Module
For a quote, contact us at info@tropical.com
- SKU:
- GB200 AI Training Module
- Weight:
- 18.00 LBS
- Width:
- 7.00 (in)
- Height:
- 2.50 (in)
- Depth:
- 14.00 (in)
Related Products
A relatable product