Key Architecture Highlights
- GPU Architecture: NVIDIA Ada Lovelace
- GPU memory specifications: GPU memory: 24GB GPU memory Type: GDDR6 GPU memory ECC: Enabled (default; can be turned off by software)
- GPU memory bandwidth: 300GB/s
Applications
- AI reasoning
- Video transcoding and video AI
- Cloud graphics processing
Engineering Parameters
| GPU Architecture | NVIDIA Ada Lovelace |
|---|---|
| GPU memory specifications | GPU memory: 24GB GPU memory Type: GDDR6 GPU memory ECC: Enabled (default; can be turned off by software) |
| GPU memory bandwidth | 300GB/s |
| GPU memory bus width | 192 bit |
| General-purpose compute performance | FP32: 30.3 teraFLOPs |
| Tensor Compute performance | TF32 Tensor Core: 120 teraFLOPS (Sparse) FP16 Tensor Core: 242 teraFLOPS (Sparse) BF16 Tensor Core: 242 teraFLOPS (Sparse) FP8 Tensor Core: 485 teraFLOPs (Sparse) |
| Integer compute performance | INT8 Tensor Core: 485 TOPs (Sparse) |
| PCIe Interface | PCIe Gen4 x16/x8, also denoted as Gen3 x16. |
| NVLink / Interconnection | The networking capacity: PCIe Gen4 x16 64GB/s |
| Video decoding | 2 units NVENC ／ 4 units NVDEC ／ 4 units JPEG Codec |
| Board power | 72W by default/maximum and minimum 40W |
| Card form factor / cooling | card form factor: half high half long slot (HHHL) cooling method: passive cooling (requires server wind tunnel) |
| GPU Frequency | Base: 795 MHz Boost: 2,040 MHz |
| Operating temperature | 0°C to 50°C |
Warranty
manufacturer's public product information does not specify the warranty period for this model in its supply channel; applicable region, channel and actual warranty service confirmed by the supply contract.
Manufacturer information
Manufacturer informationVerified Documentation & Datasheets
Official products page
Download DocumentPB-11316-001_v01 | March 2023 NVIDIA L4 GPU Accelerator Product Brief
Download Document