Oracle Cloud Infrastructure Unifies AI and Simulation Using NVIDIA RTX PRO Blackwell Instances for Enhanced Enterprise Security and Performance
Oracle Cloud Infrastructure has addressed the growing complexity of fragmented AI and simulation systems by launching a new bare metal compute shape powered by the NVIDIA RTX PRO 6000 Blackwell Server Edition. According to Oracle this new infrastructure tier is designed to consolidate multimodal AI, photorealistic rendering, and advanced physical simulations into a single environment. By running on the Oracle Acceleron software suite the platform simplifies complex operations while enhancing system security and performance across large scale enterprise workloads.
The new shape identified as BM.GPU.RTXPRO.8 utilizes the latest Blackwell architecture to support generative AI and real time data processing. The Intel Xeon 6 CPU system delivers processing power with its 144 cores which can achieve Turbo speeds of 4.4GHz. The system includes 8 NVIDIA RTX PRO 6000 GPUs which each contain 96 GB of fast GDDR7 memory. This configuration enables the simultaneous processing of large datasets and multimodal models that include video and text and simulation data. The bare metal approach from the company gives users direct access to hardware resources which enables them to achieve predictable performance levels while experiencing less variable performance results compared to virtualized offerings from other hyperscalers.
The instance exceeds industry cloud storage standards by providing higher memory and storage capabilities. The platform delivers 3 TB of system memory which doubles the memory capacity available on other hyperscaler instances. The system uses 61.44 TB of local NVMe space for local storage which provides four times the storage capacity available in standard GPU virtual machines. The high specifications function to remove data bottlenecks which emerge during the transition from experimental testing to complete production.
The networking stack establishes exceptional performance to ensure successful communication between workloads that operate from different locations. The instance supports up to 400 Gbps front end connectivity and 1600 Gbps RDMA back end bandwidth. The low latency architecture enables organizations to conduct inference and media processing at the same time while keeping their systems from forming isolated operational structures. The consolidation of different workloads onto one GPU platform enables companies to use their resources more efficiently while decreasing their expenses that result from using separate systems for graphics and AI.
The OCI RTX PRO costs 4.50 for each hour of GPU usage. The rate delivers excellent value because it enables mixed enterprise workloads to achieve maximum performance per dollar spent. The bare metal nature of the instance ensures that users do not suffer from the formatting or remote access issues often found in virtualized GPU environments. The system operates best for applications that require low latency including robotics and digital twins that use NVIDIA Omniverse technology and agentic AI systems.
The Blackwell architecture functions as a flexible system which supports various professional needs across multiple industries. Generative AI developers can use the shape to create large scale inference and RAG pipelines. Advanced RT cores in visual computing provide real time ray tracing capabilities which virtual production and photorealistic architectural design use. Scientific computing and engineering teams can use the hardware to create synthetic data and conduct detailed scientific simulations. The unified approach enables AI applications to perform simulation and visualization activities within a production ready environment that can function as a flexible system.
