テクノロジー
Qualcomm Dragonwing始動：フィジカルAIの覇権を握る「エッジの巨人」の次なる一手とNVIDIAへの挑戦状
2026年1月、ラスベガスで開催されるCESにおいて、Qualcomm Technologies（以下、Qualcomm）は今年のテクノロジー業界を象徴する重大な発表を行った。同社は、これまでスマートフォン向けSoCで築 […]
Tech Product
Dragonwing Q-8750
Overview
最終更新: 2026年7月21日77 TOPSのAIエンジンを搭載し、クラウド接続なしで最大110億パラメータのモデルを実行可能なハイエンドプロセッサ。自律飛行ドローンの障害物回避や高解像度映像のリアルタイム処理など、ミッションクリティカルな用途を想定している。
Mentioned Articles
1 件Research Papers
5 件- Wavelets and filter banks
- Determinants of microvascular flow.
- Detection of the (3, 0) Phillips band of interstellar C2 toward zeta Ophiuchi
- Arduino UNO Q with Qualcomm AI Chip: Enabling Next-Generation Edge Intelligence for Embedded AI Prototyping
The Arduino UNO Q introduces a novel dual-processor heterogeneous architecture, combining a Qualcomm Dragonwing QRB2210 microprocessor with a real-time STM32U585 microcontroller. The QRB2210 features a quad-core 64-bit Arm Cortex-A53 CPU (2.0 GHz) with an Adreno 702 GPU (845 MHz), delivering significant computational improvements over legacy Arduino platforms. Benchmark analysis reveals that the UNO Q achieves a 12.5× throughput improvement over the Arduino UNO R3 (16 MHz) and a 4.2× improvement over the UNO R4 WiFi (48 MHz). The memory architecture shows a 1,048,576× increase in SRAM relative to the UNO R3, with 2 GB of LPDDR4X enabling complex AI inference. Peak memory bandwidth reaches 2.4 MB/ns, compared to 0.32 MB/ns on the UNO R3. The dual-brain architecture enables real-time deterministic control via the STM32U585 subsystem, while leveraging GPU acceleration for TensorFlow Lite inference with sub-100ms latency. This work examines the architectural innovations and practical implications for edge AI, IoT, and robotics, which require both high-performance computing and real-time response guarantees in resource-constrained environments.
- TRPV1-mediated cough in guinea pigs Protease-activated receptor-2 activation exaggerates