QNAP Adds Local AI Acceleration with Its Plug-and-Play QAI-M100, QAI-U100 Add-Ons

Network attached storage (NAS) specialist QNAP is getting in on the edge artificial intelligence (edge AI) boom, announcing add-in modules that connect to compatible QNAP NAS units to add three tera-operations per second (TOPS) of compute for on-device computer vision and other machine learning workloads.

"With the rapid growth of edge AI, the demand for real-time data processing and analysis at the edge has increased," says QNAP product manager Andy Chung in support of the company's latest launch. "The combination of QNAP NAS and the QAI edge AI accelerators not only provides a high-performance storage platform but also significantly boosts the computing efficiency of AI applications, improving overall cost-effectiveness."

The two accelerator modules — the M.2 B+M-key QAI-M100 and USB 3.2 Gen. 1 Type-A QAI-U100 — are, effectively, single-board computers based around a Rockchip RK1808 system-on-chip featuring two Arm Cortex-A35 cores running at 1.6GHz and including an on-board neural processing unit (NPU) rated at a claimed three tera-operations per second (TOPS) of minimum-precision compute. Their design has been tailored for compatibility with QNAP's own AI Core software running on its QTS or QuTS Hero operating systems. Both units include 1GB of dedicated memory and a passive heatsink for cooling.

According to the company's internal testing, adding a QAI edge accelerator to a QNAP TS-673A NAS boost the performance of object recognition in the QuMagie Smart Album application by up to 36 per cent, with facial recognition speed boosted by up to 22 per cent. For those who need even more power, the company says a second accelerator can be added — though it has not stated whether this provides a linear doubling in performance or if there are overheads involved.

More information on the QNAP QAI-M100 and QAI-U100 is available on their respective product pages; both devices were showing as out-of-stock on the official QNAP store at the time of writing, with pricing confirmed by resellers at $59 for the QAI-M100 and $69 for the QAI-U100.