PNY RTX PRO 6000 Blackwell Server Edition, 96 GB GDDR7

The PNY NVIDIA RTX PRO 6000 Blackwell Server Edition is designed for demanding AI, visual computing, and scientific workloads in the data centre. 96 GB of GDDR7 with ECC provides ample space for large models, complex 3D projects, and extensive datasets. The Blackwell architecture combines high compute performance with specialised Tensor and RT cores for a variety of professional tasks.

• 96 GB GDDR7 with ECC offers plenty of memory for large AI models and complex datasets
• 4 PFLOPS FP4 accelerate generative and agent-based AI workloads
• 120 TFLOPS FP32 deliver compute power for simulations and scientific computing
• Up to four MIG instances partition the GPU into isolated workloads
• Four NVENC and NVDEC engines accelerate professional video and streaming pipelines

96 GB GDDR7 for large models and datasets

With 96 GB of GDDR7 and ECC, the PNY RTX PRO 6000 can hold extensive AI models, 3D scenes, and datasets directly in GPU memory. The memory bandwidth of 1,597 GB/s ensures that the compute units can quickly access this data. This reduces bottlenecks in memory-intensive AI inference, data analysis, and visualisation.

Blackwell compute power for AI and simulation

24,064 CUDA cores deliver up to 120 TFLOPS FP32 for highly parallelised compute tasks. Fifth-generation Tensor Cores achieve up to 4 PFLOPS at FP4, supporting especially compute-intensive AI workloads. 188 fourth-generation RT cores additionally provide up to 355 TFLOPS for ray tracing and complex visualisations.

Shared resources for data centres

With Multi-Instance GPU, the PNY RTX PRO 6000 can be divided into up to four isolated GPU instances. This allows you to allocate compute resources specifically to different users or applications and run multiple workloads on the same GPU. PCIe 5.0 provides a fast connection to the host system. The passive cooling is designed for appropriately ventilated servers, and NVIDIA offers air-cooled dual-slot as well as liquid-cooled single-slot configurations.