Overview
The NVIDIA RTX PRO 5000 Blackwell 72GB is the high-memory configuration of the RTX PRO 5000 Workstation Edition. It pairs the same 14,080-core NVIDIA Blackwell GPU, fifth-generation Tensor Cores and fourth-generation RT Cores with 72GB of GDDR7 ECC memory, 50 percent more than the 48GB card, so large language models, high-resolution datasets and complex engineering assemblies stay resident on the GPU.
With 2,064 AI TOPS, 1,344 GB/s of memory bandwidth and Multi-Instance GPU support for two isolated 36GB instances, the 72GB card is built for LLM inference and fine-tuning, generative AI, data science, digital twins and CAD/CAE simulation on a single dual-slot workstation card at 300W.
Key Features
- 72GB GDDR7 ECC: The largest memory in the RTX PRO 5000 line, with error correction for reliable long-running jobs
- 2,064 AI TOPS: Fifth-generation Tensor Cores with FP4 support for generative AI and LLM workloads
- 14,080 CUDA Cores: Blackwell parallel compute for simulation, rendering and data processing
- 1,344 GB/s Bandwidth: 384-bit GDDR7 keeps large models and datasets fed
- Multi-Instance GPU: Partition into 2x 36GB or run as 1x 72GB
- PCIe 5.0 and DisplayPort 2.1b: Four DisplayPort 2.1b outputs, up to 4x 4K at 165 Hz or 2x 8K at 100 Hz
- Video Engines: 2x ninth-generation NVENC and 2x sixth-generation NVDEC
- 300W Dual Slot: Active cooling and a single 16-pin PCIe CEM5 power connector
Technical Specifications
| Specification | Details |
|---|---|
| GPU Architecture | NVIDIA Blackwell |
| CUDA Cores | 14,080 |
| Tensor Cores | 440 (5th Gen) |
| RT Cores | 110 (4th Gen) |
| Memory | 72 GB GDDR7 with ECC |
| Memory Interface | 384-bit |
| Memory Bandwidth | 1,344 GB/s |
| AI Performance | 2,064 TOPS |
| MIG Support | Up to 2x 36 GB or 1x 72 GB |
| Video Engines | 2x NVENC (9th Gen), 2x NVDEC (6th Gen) |
| TDP | 300W |
| Power Connector | 1x PCIe CEM5 16-pin |
| Interface | PCIe 5.0 x16 |
| Display Outputs | 4x DisplayPort 2.1b |
| Max Simultaneous Displays | 4x 3840 x 2160 @ 165 Hz, 2x 7680 x 4320 @ 100 Hz |
| Form Factor | 4.4" H x 10.5" L, dual slot, full height, active cooling |
| Graphics APIs | DirectX 12, Shader Model 6.6, OpenGL 4.6, Vulkan 1.3 |
| Compute APIs | CUDA, OpenCL 3.0, DirectCompute |
Ideal Use Cases
- Local LLM inference and fine-tuning of larger models on a workstation
- Generative AI and multimodal model development
- Data science with large in-memory datasets
- Digital twins, CAE simulation and large CAD assemblies
- 8K video post-production and virtual production
Why Choose This Product?
If your models or scenes no longer fit in 48GB, the 72GB RTX PRO 5000 gives you the headroom without stepping up to the 96GB RTX PRO 6000 and its higher power budget. Same GPU, same drivers and certifications, 50 percent more memory. The 48GB configuration is available as the NVIDIA RTX PRO 5000 Blackwell.
Interested? Contact us for pricing, volume orders and workstation configuration advice.
