Wiwynn Unveils Nvidia SCADA Server with 2.9 PB Storage and PCIe 6.0 Support

At Computex 2026, Wiwynn unveiled one of the first servers based on Nvidia SCADA (Scaled Accelerated Data Access), a platform designed for AI data centers. The solution is optimized for handling extreme workloads in machine learning training and inference, where rapid access to large datasets is critical.

The server supports up to 96 liquid-cooled E3.S SSDs, enabling storage capacities of up to 2.9 petabytes when using Micron 9650 Pro drives with 30.72 TB capacity. The system delivers up to 528 million IOPS for random 4 KB block reads, while sequential read/write speeds are limited only by PCIe switch and network card capabilities.

At the core of the SCADA architecture is the elimination of bottlenecks common in traditional CPU-centric systems. The platform allows GPUs to directly initiate and control I/O operations, which is crucial for tasks requiring high parallelism. The Wiwynn server is built on Nvidia Vera processors, four RTX Pro 6000 Blackwell GPUs, and Broadcom PCIe 6.0 switches, ensuring maximum efficiency for AI workloads.

Nvidia positions SCADA as an intermediate layer between local SSDs and remote storage (Tier 3.5). Servers in this class are designed to transfer data to compute nodes at high speeds with minimal latency. In the future, the platform will integrate into the Storage Next ecosystem, aimed at unifying storage with GPU memory to accelerate AI applications.

Wiwynn has not disclosed the SCADA server's pricing, but it is clearly targeted at large data centers and enterprise customers. The cost will depend on configuration, including storage capacity and components such as GPUs and Nvidia processors.