Dnotitia Presents First Live Demonstration of Its VDPU Vector Search ASIC at SC26

Dnotitia Presents First Live Demonstration of Its VDPU Vector Search ASIC at SC26

PR Newswire

  • Building on earlier FPGA-based evaluation, Dnotitia’s ASIC performs real-time natural-language search across approximately 2.6 million video frames
  • Search offloading improves performance and reduces host CPU utilization by 90% or more

CHICAGO, Nov. 19, 2026 /PRNewswire/ — Dnotitia Inc., an AI data infrastructure and semiconductor company, today announced the first public live demonstration of its VDPU (Vector Data Processing Unit) ASIC, a chip designed for AI data retrieval, at Supercomputing 2026 (SC26) in Chicago. The chip performed real-time searches across approximately 2.6 million video frames using natural-language queries.

Attendees visit the Dnotitia booth at SC26 to view its VDPU Vector Search ASIC demonstration.

The live search ran on a VDPU card featuring an application-specific integrated circuit (ASIC). In response to a  user’s natural-language query, the VDPU searches and compares vector representations of video frames to find semantically relevant results. At AI Infra Summit 2026 in September, Dnotitia presented FPGA-based VDPU search performance. At SC26, the company showcased its ASIC-based VDPU in operation.

Compared with processing the same search workload on the CPU alone, offloading search to the VDPU improved search performance and reduces host CPU utilization by 90% or more. The CPU resources freed up can be used to run AI agents or other services.

AI agents repeatedly retrieve relevant information and users’ past interactions as they carry out tasks. Semantic search helps them find information that matches the meaning of a query rather than relying on exact keyword matches. Vector search is a key technology behind semantic search: it represents the semantic features of documents, images and other data as numerical vectors, then compares them to identify related information. As agents take on more tasks and access more data, fast, accurate retrieval becomes increasingly important. The growing volume of traversal and comparison operations also calls for efficient use of computing resources.

Dnotitia is optimizing VDPU performance in server environments, targeting approximately 10 times the search performance of a dual-socket CPU server with a single server equipped with multiple VDPU cards. The company aims to improve computing resource utilization by assigning service execution to CPUs, AI model computation to GPUs and vector search to the VDPU.

Dnotitia develops the Seahorse vector database and AI agent software alongside the VDPU, incorporating requirements from real-world search workloads into the chip’s design. It is also evaluating the VDPU for integration into third-party server and storage systems.

“This demonstration validated the search performance and operation of our first VDPU ASIC,” said MK Chung, CEO of Dnotitia. “We will continue our proof-of-concept work with global server and storage companies to assess performance and reliability.”

SC26 is an international conference and exhibition showcasing technologies and research in high-performance computing (HPC), networking and storage. Dnotitia is exhibiting at Booth 5613 at McCormick Place in Chicago from November 17 to 19.

###

Dnotitia, Inc.

Dnotitia is an AI data infrastructure and semiconductor company developing silicon and software for AI retrieval. The company designs the VDPU (Vector Data Processing Unit), accelerator cards and software that integrates with existing vector databases. Its technology offloads vector search to dedicated hardware, helping free CPU and GPU resources for applications and AI models. Headquartered in Seoul, South Korea, Dnotitia works with customers and partners in North America through its San Jose, California office.

For more information, visit us.dnotitia.com.

Cision View original content to download multimedia:https://www.prnewswire.com/news-releases/dnotitia-presents-first-live-demonstration-of-its-vdpu-vector-search-asic-at-sc26-302945556.html

SOURCE Dnotitia Inc.