As generative AI evolves from simple question-and-answer to complex "agentic AI," developers' demands for computing power and memory on desktop computers are also increasing. NVIDIA recently officially announced that the NVIDIA RTX PRO 5000 professional graphics card, which uses the latest Blackwell display architecture, will now have a 72GB memory version available for immediate shipment.
This new card is primarily designed to fill the gap between the existing 48GB version and higher-end products, freeing data scientists, AI developers, and creative professionals from being constrained by memory capacity when facing massive LLM (Large Language Model) or complex 3D rendering tasks.
Solving the "memory anxiety" in AI development
Why is 72GB needed? NVIDIA points out that as AI development enters the era of "proxy AI," systems often need to simultaneously connect multiple tools, execute RAG (Retrieval Augmentation) processes, and handle multimodal understanding tasks. This means that the GPU must simultaneously store multiple AI models, data sources, and code formats in memory.
The new RTX PRO 5000 72GB features high-speed GDDR7 memory, with a 50% increase in capacity compared to the original 48GB version. This extra space allows developers to train, fine-tune, or prototype larger models directly on their workstations without constantly uploading data to cloud data centers, thus ensuring data privacy, low latency, and cost-effectiveness.
In terms of performance, the RTX PRO 5000 72GB based on the Blackwell display architecture can provide up to 2142 TOPS of AI computing performance, effectively solving the bottleneck under multitasking load.
Performance testing: Image generation is 3.5 times faster.
According to official data released by NVIDIA, the RTX PRO 5000 72GB performs exceptionally well in standard benchmark tests for generative AI.
• Image Generation:Its performance is 3.5 times that of the previous generation of hardware.
• Text generation:Its performance is 2 times that of the previous generation of hardware.
• 3D rendering:Rendering time has been reduced by up to 4.7 times in path tracing engines such as Arnold, Chaos V-Ray, and Blender, as well as in real-time renderers such as D5 Render and Redshift.
• Engineering Design:The graphics performance of computer-aided engineering (CAE) and product design has been improved by more than 2 times.
Practical Applications: From NASA to Virtual Production
Startups have already adopted this new card. Michael Bogomolny, CEO of InfinitForm (a member of NVIDIA Inception), an engineering design generative AI software provider, said that they are currently using the RTX PRO 5000 72GB to optimize their CUDA acceleration software, helping customers such as Yamaha Motor and NASA accelerate product innovation and manufacturing optimization.
In the film and television industry, virtual production company Versatile Media uses its 72GB memory capacity to handle more complex real-time rendering scenes. Eddy Shen, general manager of its production center, pointed out: "For film-level virtual production, memory capacity directly translates into creative freedom."
Listing Information
The NVIDIA RTX PRO 5000 72GB Blackwell GPU is now officially available through global partners, including Ingram Micro, Leadtek, Unisplendour, and xFusion, providing system integrators with a new core option for building AI-ready workstations.
