Nvidia has officially introduced a new version of its RTX PRO Blackwell PRO 5000 graphics card with 72GB of memory. This model features 50% more GDDR7 memory compared to Nvidia’s existing 48GB variant, making it a significant development for users dealing with memory-intensive tasks. Techpowerup was the first to notice this GPU, which Nvidia announced without involving the media.
Both variants of the RTX PRO 5000 Blackwell GPUs are reported to have the same 512-bit memory bus and equivalent memory bandwidth levels. This indicates that Nvidia is utilizing an unconventional memory setup for these GPUs. You cannot achieve 72GB of memory using 3GB memory chips on a 512-bit memory bus. Each GDDR7 memory chip connects to a 32-bit memory interface, meaning that sixteen chips are necessary for a 512-bit memory bus. Therefore, 16 x 3GB results in 48GB, which is the memory configuration of Nvidia’s previous RTX PRO 5000 model.

To reach 72GB, Nvidia must implement two 3GB modules on eight of the GPU’s 32-bit memory interfaces. If this is indeed the case, some of the 32-bit interfaces would have 6GB of memory while others would have 3GB. The memory bandwidth and CUDA core count remain the same for both RTX PRO 5000 Blackwell GPUs. The only change is in the memory capacity of Nvidia's GPU, which for many users is the most crucial specification. Both models also maintain a thermal design power (TDP) of 300W.
While the RTX PRO 5000 Blackwell sets a new standard for peak performance, high-end options are not the only offerings available. The RTX 5080 features remarkable next-generation capabilities, swift ray tracing, and DLSS 4 support—all at a more affordable price point. Whether you’re constructing a high-performance system or upgrading your existing setup, the RTX 5080 delivers outstanding gaming and creative performance at a competitive rate. Use our price comparison tool to find the best deals on the GeForce RTX 5080 graphics card now.