Innodisk Japan Corporation Official site

As workloads for AI and real-time processing continue to grow, systems are increasingly relying on higher-performance CPUs and accelerators to handle large amounts of data. With this increase in computational power, system architects are facing the "memory wall" earlier than ever. To meet this rising demand, systems require a large amount of high-capacity, low-latency DRAM. However, traditional memory expansion through DIMM slots is limited by the number of memory channels supported by the CPU. Once all the slots are filled, it becomes extremely difficult to expand memory capacity. Please download to continue reading.

We have announced the high-speed 10GbE LAN series, a group of LAN modules built to meet the growing network demand for edge AI applications. Edge AI applications are increasingly reliant on real-time data exchange between sensors, edge servers, and endpoints, making network connectivity a crucial foundation for system performance. This series is available in both M.2 and PCIe form factors, achieving high throughput and low latency connections through DPDK, PTP, and SR-IOV, allowing for efficient handling of data-intensive workloads even in space-constrained environments.

Innodisk has announced the APEX series Edge AI system, which provides scalability and high performance for industrial and enterprise environments. By integrating hardware, software, and system design, the platform integration is simplified, allowing customers to focus on faster deployment and achieving business outcomes without being burdened by the complexity of infrastructure. One of the key features of the series is the APEX-A100, an advanced edge AI system designed for demanding industrial applications. ● Equipped with Qualcomm Dragonwin IQ9 series industrial processors ● Balances high performance and power efficiency ● Long industrial lifecycle Please continue reading.

Not all SSDs require onboard DRAM to achieve high performance. So, how do DRAM-less designs maintain responsiveness? The answer lies in Host Memory Buffer (HMB). By leveraging system memory via PCIe, HMB allows DRAM-less SSDs to cache critical metadata such as L2P mapping tables, significantly improving lookup speed and random read performance without increasing additional hardware costs or footprint. In embedded and industrial systems where space and component cost efficiency are crucial, HMB offers a smarter balance of performance and design flexibility. Want to learn more about how it works and its importance? Be sure to check out the white paper. ✔ Reasons SSDs rely on fast mapping tables ✔ Trade-offs between DRAM-equipped designs and DRAM-less designs ✔ How HMB enhances performance while keeping the system compact

AI object detection applications are rapidly expanding in areas such as healthcare, smart retail, transportation, and industrial environments. As edge deployment grows, a critical challenge arises: how to deliver powerful AI performance within space-constrained systems. APEX-P200: https://inno.pse.is/8mbwqv ◆ High-Performance AI Computing: Powerful performance with #Intel Core i7 and NVIDIA RTX 2000 Ada (PRO 2000 Blackwell compatible) ◆ Ultra-Compact Design: A 182mm MXM Type-A system designed for edge deployments with space limitations ◆ Flexible One-Stop Integration: Supported by Innodisk's in-house components and customization services, simplifying integration and accelerating deployment ◆ Adapted for Industrial Environments: Supports operating temperatures from -20°C to 60°C, ensuring reliable performance in real-world conditions.