ADATA's enterprise storage brand TRUSTA and ADATA Industrial showcased AI and edge computing solutions at US SuperComputing 2026 (SC26) and Japan EdgeTech+ 2026. TRUSTA launched an upgraded AI Scaler Toolkit integrating vLLM inference engine, connecting GPU memory, DRAM, and enterprise SSDs through KV cache offloading technology, with prompt preprocessing speeds increasing up to approximately 25% in specific test scenarios when combined with Intel AMX technology. The exhibitions addressed enterprise needs for effective compute resource utilization and data reliability as AI adoption accelerates across deployment environments. Memory and storage technology integration has become critical for enterprise AI inference, data centers, edge computing, and smart terminal applications.
TRUSTA Debuts AI Scaler Toolkit at SC26 Exhibition
TRUSTA participated in SC26 for the first time under the theme "Scaling AI Efficiently," displaying the upgraded AI Scaler Toolkit. As enterprise AI applications transition from model training to inference, large language models in long-document processing, multi-turn conversations, and AI Agent applications face KV cache capacity exceeding GPU memory limits, increasing recomputation time and deployment costs.
AI Scaler Toolkit Integrates vLLM Engine and Intel AMX Technology
TRUSTA's AI Scaler Toolkit integrates the open-source vLLM inference engine, storing KV cache in layers with GPU memory handling current requests, DRAM serving as high-speed cache, and TRUSTA PCIe Gen5 enterprise SSDs providing large-capacity storage space. Long-form content can be read directly from SSDs without recomputation. The system can retrieve previously saved KV cache to continue conversations even after server restarts, reducing AI response wait times.
The AI Scaler Toolkit also integrates Strata inference engine and Intel AMX acceleration technology. In specific Intel Xeon platform tests, prompt preprocessing (Prefill) speeds increased up to approximately 25%. TRUSTA stated the AI Scaler Toolkit provides a one-stop AI deployment and management interface, supporting vLLM, llama.cpp, and Strata inference engines, with GPU, DRAM, and SSD resource allocation and model testing capabilities to help enterprises utilize existing hardware resources and lower on-premises AI deployment barriers.
ADATA Industrial Showcases Edge AI Solutions at EdgeTech+
ADATA participated in Japan EdgeTech+ from 11月18日 to 20日 under the theme "Powering Reliable Edge AI," integrating ADATA Industrial, TRUSTA, and embedded product portfolios to demonstrate product layouts from enterprise AI deployment and edge computing to smart terminals.
ADATA Displays Embedded Products for Smart Devices
ADATA exhibited eMMC, UFS 3.1, and LPDDR5X products addressing stable storage, high-speed data access, and low-power computing requirements respectively. UFS 3.1 can shorten smart device boot and data loading times. LPDDR5X transfer rates reach up to 8,533 MT/s, helping portable and wearable devices balance AI computing performance and battery life under limited power consumption and heat dissipation conditions.
ADATA Industrial Presents Industrial-Grade SSD and Memory Portfolio
ADATA Industrial displayed industrial-grade SSDs covering multiple specifications including PCIe Gen4x4 M.2 2230, M.2 2242, U.2, and 2.5-inch SATA III, supporting AES 256-bit encryption and TCG OPAL data security mechanisms. Product options include PLP, A+SLC, and wide temperature support from -40°C to 85°C. Memory products included DDR5 6400MT/s ECC CUDIMM and DDR5 ECC CSODIMM to strengthen industrial computing platform performance and stability.
ADATA Demonstrates Customization Capabilities Through Japan Cases
ADATA demonstrated customization technology and technical support capabilities through Japan local cases. Services cover product selection, firmware tuning, and hardware design. For usage scenarios involving frequent writes, sudden power outages, and unstable power supply, ADATA provides A+SLC, PLP, A+Power Protect, Conformal Coating, and Anti-sulfuration technologies to enhance industrial equipment long-term operation reliability.
FAQ
What did TRUSTA showcase at the SC26 exhibition?
TRUSTA showcased an upgraded AI Scaler Toolkit at SC26 that integrates the vLLM inference engine, connecting GPU memory, DRAM, and enterprise SSDs through KV cache offloading technology. In specific test scenarios combining Intel AMX technology, prompt preprocessing speeds increased up to approximately 25%.
What products did ADATA Industrial display at EdgeTech+?
ADATA Industrial displayed industrial-grade SSDs in multiple specifications (PCIe Gen4x4 M.2 2230, M.2 2242, U.2, 2.5-inch SATA III), DDR5 6400MT/s ECC CUDIMM and ECC CSODIMM memory, and embedded products including eMMC, UFS 3.1, and LPDDR5X at the EdgeTech+ exhibition from 11月18日 to 20日.
How does ADATA's AI Scaler Toolkit reduce AI deployment costs?
The AI Scaler Toolkit stores KV cache in layers across GPU memory, DRAM, and enterprise SSDs, allowing long-form content to be read directly from SSDs without recomputation. The system can retrieve saved KV cache after server restarts to continue conversations, reducing response wait times and helping enterprises utilize existing hardware resources to lower on-premises AI deployment barriers.