Compute · Universal Pro
NVIDIA RTX PRO 6000
96 GB of memory and serious throughput at an accessible price — the universal card for inference, simulation and physical-AI workloads.
Specifications
Big memory, small bill
| GPUs per node | 4–8 × RTX PRO 6000 |
| GPU memory | 96 GB GDDR7 per GPU |
| Node interconnect | PCIe 5.0 |
| Cluster interconnect | 200–400 Gb/s Ethernet / InfiniBand |
| Form factor | Standard rack servers, single- or dual-CPU hosts |
| Availability | On-demand, spot & reserved |
Why teams choose it
The value play
When you don’t need NVLink-scale training, RTX PRO 6000 nodes deliver excellent tokens-per-dollar for serving, batch inference, rendering and simulation — and they’re available by the hour.
Suited workloads
Production inference
Batch & offline inference
Scientific simulation
Physical AI & robotics
Rendering & transcode
Start your journey today
Tell us about your workload — an engineer, not a sales bot, will get back to you within one business day.