Great news! Nvidia finally has a 72GB graphics card — but since it's the RTX Pro 5000 Blackwell, it will probably cost you just a bit more than a 64GB DDR5 stick

• RTX Pro 5000 Blackwell supports up to 72GB GDDR7 memory with ECC
• Blackwell architecture delivers 14,080 CUDA cores and enhanced Tensor and Ray Tracing Cores
• Multi-Instance GPU support enables splitting a single card into multiple virtual GPUs

The Nvidia RTX Pro 5000 Blackwell increases memory capacity, offering configurations up to 72GB of GDDR7 memory with ECC support.

Its 512-bit memory interface delivers a theoretical bandwidth of 1.344TB/s, supporting high-resolution simulations, large models, and datasets commonly found in workstation environments.

The GPU uses the Blackwell architecture and includes 14,080 CUDA cores, fifth-generation Tensor Cores optimized for FP4 precision, and fourth-generation Ray Tracing Cores that handle complex geometry calculations.

Core architecture and compute specifications

The card includes an AI Management Processor and support for Multi-Instance GPU (MIG), which allows it to partition into multiple virtual GPUs and support parallel engineering and simulation workloads.

This professional graphics board supports up to four DisplayPort 2.1b outputs with maximum digital resolutions exceeding 7680 x 4320 at 60Hz.

The system interface is PCIe 5.0x16, which provides high-speed communication between the GPU and CPU and helps move large datasets without creating bottlenecks in demanding workstation applications.

Nvidia also upgraded its video engines, adding ninth-generation NVENC and sixth-generation NVDEC with 4:2:2 support.

These engines accelerate encoding and decoding of high-resolution video streams used in specialized engineering and visualization tasks.

The RTX Pro 5000 Blackwell includes enhanced streaming multiprocessors paired with AI-enabled Tensor Cores, targeting workloads that depend heavily on GPU computation.

Its single-precision floating-point performance reaches 65TFLOPS, while RT Core performance peaks at 196TFLOPS.

Despite these specifications, real-world performance depends on application optimization, CPU pairing, and how efficiently software uses available memory.

Users running CAD, 3D modeling, or simulation software will likely see the greatest benefit when working with very large datasets or multi-GPU partitions, although many engineering workloads remain limited by CPU performance or software design.

The card draws 300W through a single 16-pin PCIe power connector and uses a dual-slot, full-height form factor.

An active cooling solution maintains thermal stability during sustained workloads, which suits professional environments where systems often run for long periods.

Its full-length design of 10.5 inches requires careful planning when selecting a compatible workstation chassis.

The 48GB RTX Pro 5000 Blackwell lists at $4,199.99, and moving to a 72GB configuration is expected to raise the price further.

Using simple “GB-for-GB” math, the higher-capacity version would land around $6300, placing it above many competing professional cards from Nvidia and other vendors.

Given that Nvidia’s 96GB RTX Pro 6000 Blackwell Workstation Edition now costs just over $7,500 after recent price adjustments, the expected pricing of the RTX Pro 5000 Blackwell raises questions about overall value.

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!

And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.