Energy-efficient multi-workload accelerator for AI inference, data science, video and visual computing.
The NVIDIA RTX PRO 4500 Blackwell Server Edition is an energy-efficient, single-slot 165 W GPU that brings Blackwell capabilities to mainstream servers. It accelerates AI inference, data science, video and visual computing without special power or cooling.
With 32 GB of GDDR7, MIG support and NVIDIA vGPU, it is a practical choice for dense inference, virtual workstations and edge deployments.
Up to 3x the performance of the previous generation with FP4 and DLSS 4.
800 GB/s of bandwidth for larger models and creative workflows.
Up to 2x the ray-tracing performance of the previous generation.
Two fully isolated 16 GB instances with guaranteed quality of service.
Twice the bandwidth of PCIe Gen 4 for data-intensive work.
Passive cooling that fits standard enterprise and edge servers.
| Architecture | NVIDIA Blackwell |
|---|---|
| CUDA cores | 10,496 |
| RT Cores | 82 (4th generation) |
| Tensor Cores | 5th generation |
| FP4 Tensor Core | 1.6 PFLOPS |
|---|---|
| FP8 Tensor Core | 811 TFLOPS |
| FP16/BF16 Tensor Core | 406 TFLOPS |
| TF32 Tensor Core | 203 TFLOPS |
| FP32 | 51 TFLOPS |
| Peak RT Core | 154 TFLOPS |
| GPU memory | 32 GB GDDR7 |
|---|---|
| Memory interface | 256-bit |
| Memory bandwidth | 800 GB/s |
| Power | 165 W |
| Form factor | Single-slot FHFL (4.4" H × 10.5" L), passive |
| Interconnect | PCIe 5.0 x16 |
| Power connector | 1× PCIe CEM5 16-pin |
| Media engines | 3× NVENC, 3× NVDEC |
| MIG | Up to 2 × 16 GB instances |
| Confidential computing | Supported |
Universal data center GPU for agentic AI, physical AI, scientific computing, rendering and video.
Hopper-generation PCIe GPU with large HBM3e memory for LLM inference and HPC in air-cooled racks.
Ada Lovelace universal GPU for generative AI inference, graphics and Omniverse workloads.
Tell us your workload and timeline for the NVIDIA RTX PRO 4500 Blackwell Server Edition. We reply within one business day with a configuration, pricing and lead time.