The new NVIDIA RTX PRO 4500 Blackwell Server Edition is now in stock

Video and streaming acceleration

Ninth-Generation NVENC Engine

NVIDIA RTX PRO 4500 Blackwell features 3 ninth-generation NVIDIA encoders, enabling nearly 2x the streaming capacity of the previous generation. They support 4:2:2 encoding for H.264 and HEVC. The Blackwell architecture also improves encoding efficiency with roughly 15% better H.265 quality and 10% better AV1 quality.

It also introduces accelerated MV-HEVC encoding for multi-view applications and a new Ultra High Quality AV1 encode mode, boosting AV1 quality by an additional 10%.

Sixth-Generation NVDEC Engine

Blackwell enhances hardware decoding capabilities with support for 4:2:2 color format decoding for H.264 and HEVC as well as 3 NVDEC engines on RTX PRO 4500. Added hardware-accelerated decoding for Motion JPEG improves performance in applications such as surveillance systems and professional video equipment.

AI and data center features

Multi-Instance GPU (MIG)

MIG allows partitioning into fully isolated GPU instances, each with its own high-bandwidth memory, cache and compute cores. The RTX PRO 4500 Blackwell can support up to 2 MIG instances with 16 GB of memory each. This allows professionals to run multiple smaller workloads concurrently with guaranteed quality of service.

PCIe Gen 5

Support for PCIe Gen 5 doubles PCIe Gen 4 bandwidth to up to 64 GB/s at x16. This enables faster data transfer for AI, data science and 3D modeling from large datasets, and improves CPU-GPU communication.

2nd-Generation Transformer Engine

Blackwell Tensor Cores add new precisions including modern micro-scaling formats. With advanced dynamic range management and micro-tensor scaling, the architecture optimizes performance and enables FP4 AI.

Graphics, ray tracing and memory

NVIDIA Blackwell CUDA Cores

Blackwell CUDA cores deliver up to 50 TFLOPS of FP32 performance, accelerating professional graphics workflows as well as simulation and compute tasks.

Fifth-Generation Tensor Cores

The fifth generation of Tensor Cores brings a major leap in performance and supports FP4, TF32, BF16, FP16, FP8, INT8 and FP6.

Fourth-generation RT Cores

The fourth generation of RT Cores delivers more than 2x the performance of the previous generation, accelerating rendering, product design and virtual prototyping workloads.

Enhanced memory subsystem

With 32 GB of GDDR7 and 800 GB/s bandwidth, the card offers more than 50% additional memory capacity and about 2.5x the memory bandwidth of an NVIDIA L4. This benefits data processing, AI inference, content creation, CAD and compute-intensive video tasks.

Technical overview