NVIDIA RTX PRO 5000 72GB Blackwell
NVIDIA Graphics Card
The 72GB version of the NVIDIA RTX PRO 5000 Blackwell keeps the same 14,080 CUDA cores but raises GPU memory to 72GB of GDDR7 ECC - 50% more than the 48GB model - so larger language models, higher-resolution simulation meshes and very large scenes fit entirely on a single GPU without the complexity of model sharding. The 384-bit interface still delivers 1,344 GB/s of bandwidth within a 300W dual-slot air-cooled design, making it a strong choice where single-card memory capacity matters most.
Specifications
| Architecture | NVIDIA Blackwell architecture (GB202) 5th-generation Tensor Cores (FP4) 4th-generation RT Cores |
|---|---|
| Compute Cores | 14,080 NVIDIA CUDA Cores |
| AI Performance | Up to 2,142 AI TOPS |
| GPU Memory | 72GB GDDR7 with ECC 384-bit interface, 1,344 GB/s bandwidth |
| System Interface | PCIe 5.0 x16 |
| Display Outputs | 4x DisplayPort 2.1 |
| Power | 300W max power consumption |
| Cooling & Form Factor | Active cooling Dual slot, 4.4" (H) x 10.5" (L) |
| Software Support | NVIDIA vGPU, RTX PRO Sync, CUDA, OptiX, DLSS 4 AV1 encode / decode |
| Use Cases | Large-parameter model inference, long-context inference, large-scene rendering, high-resolution CAE simulation |