| Products |
NVIDIA A100 80GB PCIe |
NVIDIA A100 PCIe |
NVIDIA A30 |
NVIDIA A40 |
NVIDIA A10 |
NVIDIA T4 |
NVIDIA A16 |
| PNY Part Number |
NVA100TCGPU80-KIT |
NVA100TCGPU-KIT |
NVA30TCGPU-KIT |
NVA40TCGPU-KIT |
NVA10TCGPU-KIT |
TCSCT4-KIT |
NVA16TCGPU-KIT |
| Workload |
Description |
Highest Performance Compute |
Highest Performance Compute |
Mainstream Compute |
Highest Performance Graphics |
Mainstream Graphics |
Small Footprint Low Power |
Small Footprint Low Power |
| |
Recommended Number of GPUs per Server |
| Deep Learning (DL) Training and Data Analytics |
For the absolute fastest model training and analytics |
4-8 GPUs 80GB: Bn+ parameter models (DLRM, GPT-2) |
4-8 GPUs 40GB: Bn+ parameter models (DLRM, GPT-2) |
|
|
|
|
|
| DL Inference |
For batch and real-time inference |
1-2 GPUs w/ multiinstance GPU (MIG) 80GB: large batch size constrained models (RNN-T) |
1-2 GPUs w/ multiinstance GPU (MIG) 40GB: large batch size constrained models (RNN-T) |
2–4 GPUs with MIG |
|
4-8 GPUs |
4-8 GPUs |
|
| High-Performance Computing (HPC) / AI |
For Higher Education Research and scientific computing centers |
1-4 GPUs with MIG |
1-4 GPUs with MIG |
2–4 GPUs with MIG |
|
|
|
|
| Render Farms |
For batch and real-time rendering |
|
|
|
4-8 GPUs |
4-8 GPUs |
|
|
| Graphics |
For the best graphics performance on professional VDI |
|
|
|
2–4 GPUs for high-end virtual workstations* |
2-8 GPUs for mid-range virtual workstations* |
2-8 GPUs for entry-level virtual workstations* |
2–4 GPUs for highest virtual desktop user density** |
| Cloud Gaming |
For 4K resolution / Android |
|
|
|
4-8 GPUs (4K resolution) |
4–8 GPUs (4K resolution) |
1-2 GPUs (Android) |
|
| Enterprise Acceleration |
For mixed workloads, including graphics, ML, DL, analytics, training, and inference |
1-2 GPUs with MIG for compute workloads |
1-2 GPUs with MIG for compute workloads |
1-2 GPUs with MIG for compute workloads |
1-2 GPUs for graphics-intensive workloads* |
1-2 GPUs for graphics-intensive* and compute workloads |
1-4 GPUs for balanced workloads* |
|
| Edge Acceleration |
For differing use cases and deployment locations |
1-2 GPUs with MIG |
1-2 GPUs with MIG |
1-2 GPUs with MIG |
1-4 GPUs for graphics-intensive workloads & AR / VR* |
1-8 GPUs for inference and video workloads |
1-8 GPUs for inference and video workloads |
|