The head background
The head background
Nebula Accelerator X9 High-performance AI Inference Acceleration designed for intelligent computing center
High chip utilization ratio
High chip utilization ratio

Empowered by the first streaming AI chip with 95.4% chip utilization ratio

High cost efficiency
High cost efficiency

Single card with 4 chips, providing 64TOPS peak performance for HPC and intelligent computing center applications

Low Latency
Low Latency

3ms latency, suitable for various scenarios real-time analysis

Nebula Accelerator X9

Empowered by the first streaming AI chip CAISA,X9 provides 64TOPS peak performance, and widely used in intelligent computing center,smart city,smart transportation,etc.
corerainX9
High Performance and Low Latency High Performance and Low Latency
High Versatilit High Versatility
High Cost Efficiency High Cost Efficiency
Easy Deployment Easy Deployment
Product Specifications
Chip: CAISA x 4
PCIe: PCI Express3.0 x16
Architecture: CAISA 3.0
Memory: 32GB DDR4
Chip utilization ratio: Up to 95.4%
Power Supply: PSU power
Peak performance: 64TOPS
Cooling: Passive cooling
Dimension:

313.9mm x 106.65 mm

(Standard full-length, full-height, double-slot)

ResNet-50
2.09 ms 7661.04 FPS 92.3%
ResNet-152
5.92 ms 2698.04 FPS 95.4%
YOLOv3
21.19 ms 737.13 FPS 82.4%
SSD-ResNet50
14.99 ms 1067.80 FPS 77.1%
U-Net Industrial
50.54 ms 316.60 FPS 65%
Note: Batch=16,INT8
Applications
Smart Manufacturing Smart Manufacturing
Smart City Smart City
Production Monitoring Production Monitoring
Remote Sensing Remote Sensing
Smart Grid Smart Grid
Smart Oilfield Smart Oilfield
Resources
AI Computing Center、Cloud Computation Center、HPC AI Computing Center、Cloud Computation Center、HPC

AI Computing Center、Cloud Computation Center、HPC