NVIDIA has officially announced the general availability of its upgraded RTX Pro 5000 Blackwell GPU, now featuring a substantial 72GB of GDDR7 memory. This launch comes just two months after the initial October announcement and aims to address the growing demands of memory-intensive tasks such as training large language models (LLMs) and complex AI data science.
The new 72GB model serves as a high-capacity alternative to the original 48GB variant, providing a 50% increase in video memory. This extra overhead is specifically designed for AI developers and creative professionals who need to juggle multiple models or handle massive datasets that exceed the limits of standard professional hardware.
Technically, the GPU utilizes a 512-bit memory bus to achieve up to 1,344GB/s of bandwidth. While its memory clock sits slightly lower than the consumer-focused RTX 5090, the Pro 5000 series includes Error Correcting Code (ECC) memory, a critical feature for professional stability and accuracy in long-term compute projects.
Under the hood, the Blackwell-based architecture packs 14,080 CUDA cores, delivering 2,064 AI TOPS and 64 TFLOPS of single-precision performance. NVIDIA claims this new configuration offers up to 3.5x the performance of the previous Ada Lovelace generation for image generation and double the performance for text-based AI tasks.
For creative professionals, the benefits extend into rendering speeds. NVIDIA reports that the 72GB model can slash render times by up to 4.7x in popular engines like Arnold, Blender, and V-Ray. It also promises a 2x graphics performance uplift for engineering applications, making it a versatile tool for product design and computer-aided engineering.+1
The RTX Pro 5000 Blackwell 72GB is currently available through select NVIDIA partners, including Ingram Micro, Leadtek, and xFusion. For those looking for pre-built solutions, broader availability through global system builders is expected to begin in early 2026.