According to a recent LinkedIn post from MemryX, the company’s SDK v2.2 is being presented as its most significant update since version 2.0, with changes spanning the compiler, driver, and runtime layers. The post highlights expanded operator support for Keras and ONNX models, along with higher frame rates on common architectures such as YOLO, in some cases reportedly reaching up to 4x higher FPS.
Claim 55% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
The post also points to reduced manual setup through automatic detection for newer YOLO versions, higher driver throughput for bandwidth-intensive workloads, and a unified API across C++ and Python with improved multi-stream performance. New runtime controls to balance performance and power are featured as well, suggesting an emphasis on efficiency and ease of deployment for AI inference applications.
For investors, these software enhancements may indicate a strategic push by MemryX to strengthen its AI acceleration platform and lower barriers to adoption among developers. If the reported performance gains and broader model compatibility translate into real-world customer wins, the update could support higher utilization of MemryX hardware, improve competitive positioning against other AI chip and accelerator vendors, and potentially drive revenue growth over time.