𝗪𝗶𝘄𝘆𝗻𝗻 𝗗𝗲𝗺𝗼𝘀 𝗙𝗶𝗿𝘀𝘁 𝗦𝗖𝗔𝗗𝗔 𝗦𝗲𝗿𝘃𝗲𝗿

𝗪𝗶𝘄𝘆𝗻𝗻 𝗗𝗲𝗺𝗼𝘀 𝗙𝗶𝗿𝘀𝘁 𝗦𝗖𝗔𝗗𝗔 𝗦𝗲𝗿𝘃𝗲𝗿

Wiwynn showed the first Nvidia SCADA server at Computex 2026. This machine changes how AI storage works.

Traditional servers use the CPU to manage data transfers. This creates a bottleneck for AI workloads. The SCADA architecture lets GPUs handle storage I/O directly. This removes the CPU from the path.

Technical details: • 2.9 PB of max storage. • 528 million IOPS random read speed. • 96 liquid-cooled E3.S SSDs. • Four RTX Pro 6000 Blackwell GPUs. • Uses PCIe 6.0 hardware.

This shift helps AI tasks like vector search and RAG. These tasks need fast, small data accesses. When GPUs own the I/O path, storage becomes a peer to compute. It is no longer just a peripheral.

The server uses Nvidia Vera CPUs and ConnectX-9 SuperNIC cards. It relies on high-speed components from Micron and Broadcom. This setup targets the massive data needs of modern AI models.

The industry is moving toward storage that matches compute speeds. SCADA represents a new way to design AI clusters.

Source: https://tomshardware.com

Full story: https://dev.to/gentic_news/wiwynn-shows-first-scada-server-29pb-no-cpu-for-io-2mgg

Optional learning community: https://t.me/GyaanSetuAi