Network Exp

NVIDIA Tesla A100 40GB/80GB GPU - AI Training & Inference

尺寸 40G 80G

This product requires a quote. for pricing and availability.

NVIDIA Tesla A100 40GB/80GB GPU

The NVIDIA A100 is the ultimate data center GPU, delivering unprecedented acceleration for AI training, inference, data analytics, and high-performance computing (HPC) workloads.

Key Specifications

Architecture: NVIDIA Ampere
CUDA Cores: 6,912 (80GB) / 5,832 (40GB)
Tensor Cores: 432 (80GB) / 336 (40GB, 3rd Gen)
GPU Memory: 80GB or 40GB HBM2e
Memory Bandwidth: Up to 2.0 TB/s (80GB) / 1.6 TB/s (40GB)
Interconnect: NVIDIA NVLink / PCIe Gen4
FP64 Performance: Up to 19.5 TFLOPS
FP32 Performance: Up to 19.5 TFLOPS
Tensor Float 32 (TF32): Up to 312 TFLOPS
FP16 Performance: Up to 624 TFLOPS
INT8 Performance: Up to 1,248 TOPS
Multi-Instance GPU (MIG): Up to 7 instances
Form Factor: SXM4 / PCIe
Max Power: 400W (SXM4) / 300W (PCIe)

Ideal For

AI & Deep Learning Training
Large-scale Inference
Data Analytics & Data Science
HPC & Scientific Computing
Virtual Desktop Infrastructure (VDI)

Available in 40GB and 80GB configurations. PCIe and SXM4 form factors. Contact us for volume pricing and availability.

⚠️ Export Compliance Notice

This product is classified under U.S. Export Administration Regulations (EAR) and may require an export license for certain destinations.

Controlled Destinations: Export, re-export, or transfer to China (including Hong Kong & Macau), Iran, North Korea, Russia, Syria, Belarus, Cuba, or other sanctioned countries/entities requires a U.S. government license.
End-User Verification: We verify the identity and location of all buyers prior to shipment for controlled products.
Buyer Responsibility: Purchaser is responsible for compliance with all applicable export, re-export, and transfer laws.
Order Refusal: We reserve the right to refuse or cancel any order that may violate U.S. export regulations.

For pricing and availability, please contact us with your company details and intended use.

View full details