PNY Part Number | VCNA800-PB |
|---|---|
Product | NVIDIA A800 40GB Active |
Architecture | NVIDIA Ampere |
Foundry | TSMC |
Process Size | 7 nm NVIDIA Custom Process |
Die Size | 826 mm |
CUDA® Cores | 6912 |
Streaming Multiprocessors | 108 |
Tensor Cores | Gen 3 | 432 |
FP64 Performance | 9.7 TFLOPS |
FP32 Performance | 19.5 TFLOPS |
TF32 Tensor Core | 311.8 TFLOPS* |
BFLOAT16 Tensor Core | 312 TFLOPS | 624 TFLOPS* |
FP16 Tensor Core | 312 TFLOPS | 624 TFLOPS* |
INT8 Tensor Core | 1247.4 TOPS* |
INT4 Tensor Core | 1248 TOPS | 2496 TFLOPS* |
NVLink | 2-way low profile (2-slot and 3-slot bridges), 400 GB/s bidirectional |
NVLink Bandwidth | 400 GB/s |
GPU Memory | 40GB HBM2 |
Memory Interface | 5120-bit |
Memory Bandwidth | 1555.2 GB/s |
Multi-Instance GPU Support | Up to 7 MIG Instances |
System Interface | PCIe 4.0 x16 |
Display Support | None Provided, use companion NVIDIA T1000 or RTX A4000 board for video output |
Thermal Solution | Active |
Form Factor | 4.4" H x 10.5" L, Dual-Slot |
Power Connector | CEM5 16-pin |
Maximum Power Consumption | 240W |
* Structural sparsity enabled. 1 3-year software subscription and enterprise support for NVIDIAS AI Enterprise license. Activation required. | |
High-Performance Data Science and AI Platform
Rapid growth in workload complexity, data size, and the proliferation of emerging workloads like generative AI are ushering in a new era of computing, accelerating scientific discovery, improving productivity, and revolutionizing content creation. As models continue to explode in size and complexity to take on next-level challenges, an increasing number of workloads will need to run on local devices. Next-generation workstation platforms will need to deliver high-performance computing capabilities to support these complex workloads.
The NVIDIA A800 40GB Active GPU accelerates data science, AI, and HPC workflows with 432 third-generation Tensor Cores to maximize AI performance and ultra-fast and efficient inference capabilities. With third-generation NVIDIA NVLink technology, A800 40GB Active offers scalable performance for heavy AI workloads, doubling the effective memory footprint and enabling GPU-to-GPU data transfers up to 400 gigabytes per second (GB/s) of bidirectional bandwidth. This board is an AI-ready development platform with NVIDIA AI Enterprise, and delivers workstations ideally suited to the needs of skilled AI developers and data scientists.