Midv-578
A mid‑size delivery drone equipped with a can run a real‑time SLAM (Simultaneous Localization & Mapping) pipeline at 60 fps, merging visual, inertial, and lidar data. Because the chip supports on‑chip learning, the drone can adapt to new obstacles (e.g., temporary construction scaffolding) without ever contacting the cloud , preserving both bandwidth and privacy.
Maud smiled a crooked smile. “Maybe. Or maybe we just traded one lie for a dozen truths.” MIDV-578
| Block | Function | Highlights | |-------|----------|------------| | | 4× RISC‑V RV64GC cores + 2× Vector‑DSP cores | Supports mixed‑precision compute (FP32/FP16/INT8/INT4). | | NeuroMatrix™ Accelerator | 256‑bit systolic array | 12 TOPS peak, 10 TOPS/Watt sustained. | | On‑Chip Learning Engine (OLE) | Gradient‑descent optimizer + micro‑weight storage | Enables continual learning without cloud connectivity. | | Secure Enclave (SE‑V2) | Hardware‑rooted key management, side‑channel hardened | Meets ISO/IEC 27001, GDPR‑by‑design. | | Sensor Fusion Hub | Dedicated ISP + IMU, Lidar, Radar interfaces | 8‑channel 4K video pipelines, 3 D point‑cloud pre‑processing. | | Memory Subsystem | 2 MB on‑chip SRAM + 8 GB LPDDR5X | Zero‑copy DMA for ultra‑low latency. | | Power Management Unit | Dynamic voltage‑frequency scaling (DVFS) + fine‑grained clock gating | Sub‑50 mW idle, 2 W peak for full‑speed inference. | A mid‑size delivery drone equipped with a can
If the initial search returns no results, try loosening the query: “Maybe
