Be the first to review “NVIDIA® A100 80GB” Cancel reply
Our website uses cookies to give you the best experience possible. By continuing to use our site, you agree with our use of cookies outlined in our Privacy Policy
Home / Components / NVIDIA Data Centre GPUs / NVIDIA® A100 80GB
The NVIDIA A100 80GB Tensor Core GPU is the world’s fastest and most versatile data centre GPU for AI, data analytics, and HPC applications. It is the engine of the NVIDIA data centre platform, and it offers up to 20x more performance than the previous NVIDIA Volta generation. A100 can scale up or be split into seven separate GPU instances with Multi-Instance GPU (MIG), giving elastic data centres the flexibility to adapt to changing workload demands.
A100 is part of the comprehensive NVIDIA data centre solution that includes hardware, networking, software, libraries, and optimized AI models and applications from NGC. It is the most powerful AI and HPC platform for data centres, enabling researchers to achieve real-world results and deploy solutions at scale. It also allows IT to maximize the use of every A100 GPU available.
Manufacturer | |
---|---|
Manufacturer Part Number | NVA100TCGPU-KIT |
Multi-Instance GPU (MIG) Support | Yes, up to seven instances per GPU |
---|---|
NEBS ready | Level 3 |
NVLink | 2-Way Low Profile, 2-Slot |
NVLink Interconnect | 600 GB/s Bidirectional |
Secure and Measured Boot Hardware Root of Trust | CEC 1712 |
vGPU Support | NVIDIA Virtual Compute Server with MIG Support |
Passive Cooling | Yes |
---|
# 3rd Gen Tensor Cores | 432 |
---|---|
# CUDA Cores | 6912 |
# Streaming Multiprocessors | 108 |
FP32 Performance (TFLOPS) | 19.5 |
FP64 Performance (TFLOPS) | 9.7 |
GPU Architecture | |
GPU Slot Size | |
Lithography | 7nm |
Peak BFLOAT16 Tensor Core | 624 | Sparsity |
Peak FP16 Tensor Core Performance (TFLOPS) | 624 | Sparsity |
Peak FP32 Tensor Core Performance (TFLOPS) | 312 | Sparsity |
Peak FP64 Tensor Core Performance (TFLOPS) | 19.5 |
Peak INT4 Tensor Core Performance (TOPS) | 2496 | Sparsity |
Peak INT8 Tensor Core Performance (TOPS) | 1248 | Sparsity |
Memory Bandwidth (GB/s) | 1555 |
---|---|
Memory Capacity (GB) | 80 |
Memory Interface (bit) | 5120 |
Memory Type | HBM2e ECC on by Default |
PCIe Version | PCIe 4.0 |
---|
# 8 Pin PCIe | 1 |
---|---|
Max Power Consumption (W) | 250 |
Supported OS | Linux, Windows |
---|
Mfr Warranty Length (M) | 36 |
---|---|
Mfr Warranty Type | Direct To Manufacturer |
By signing up to the Workstation Specialists newsletter, you agree you have read and agree to our Privacy Policy and consent to us communicating/marketing to you in accordance with the policy. Read the terms and conditions here. You can unsubscribe at any time.
Reviews
There are no reviews yet.