900-5G153-2200-000 — RTX PRO 6000 Blackwell Max-Q Workstation Edition 96GB GDDR7

Specifications

Overview

The NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (part 900-5G153-2200-000) is the 300 W member of the RTX PRO 6000 Blackwell family. It carries the same GB202 GPU and the same 96 GB of GDDR7 ECC as the 600 W Workstation Edition, but is power-capped to 300 W with a single-fan blower in a standard-height dual-slot body.

That power envelope is the point of the Max-Q variant. A 600 W card dictates chassis airflow, PSU sizing and slot spacing; a 300 W card fits an ordinary workstation or a 2U-adjacent enclosure, and lets a system carry more cards per kilowatt. The cost is clock headroom - peak FP32 drops from 125 TFLOPS (Workstation Edition) to 110 TFLOPS, and FP4 sparsity from a higher figure to 3.511 PFLOPS.

What is preserved is capacity and connectivity: 96 GB of GDDR7 at 1,792 GB/s over a 512-bit bus, 24,064 CUDA cores, 752 fifth-generation Tensor cores and 188 fourth-generation RT cores. For workloads that are memory-capacity bound - large-model inference, Omniverse scenes, scientific simulation - that combination buys more usable throughput per rack unit than an uncapped card that cannot be densely deployed. Note the Max-Q part carries a controlled-GPU export status per Lenovo's ordering documentation.

Key Benefits

96 GB GDDR7 ECC at 300 W — full-capacity Blackwell in a deployable power budget. 3.511 PFLOPS FP4 — fifth-generation Tensor cores for AI inference. 1,792 GB/s memory bandwidth — 512-bit GDDR7 interface. Standard-height dual-slot blower — fits dense multi-GPU workstations.

Applications

Large-model inference and fine-tuning at the workstation tier, NVIDIA Omniverse and digital-twin simulation, professional rendering and ray tracing, scientific and engineering simulation, and multi-GPU deskside AI systems.

Related Products