AImotive Achieves 98% Vision Efficiency on Nextchip Apache5 NPU

AImotive, a leading automotive AI software and IP provider, announced that its aiWare3P neural‑network processor achieved up to 98 % sustained real‑time efficiency on Nextchip’s Apache5 imaging edge processor.

This performance represents a 2‑ to 3‑fold improvement over comparable NPUs currently available for automotive inference. The Apache5 combines a high‑resolution ISP capable of 5.7 MP sensors, a quad‑core Arm A53 CPU, and a compact 9 mm × 9 mm package, all meeting AEC‑Q100 Grade 2 reliability standards for demanding edge‑vision workloads.

Within weeks of receiving the first samples, AImotive’s aiDrive team leveraged Nextchip’s Apache5 and the aiWare Studio SDK to deploy multiple well‑optimized neural networks. Only minimal pre‑deployment tuning was required, and the 1.6 TOPS aiWare3P core delivered the claimed performance with real‑time efficiency exceeding 95 % across a broad set of automotive AI workloads.

“We’re thrilled that the close collaboration between Nextchip and AImotive enabled Apache5 to run high‑impact automotive AI applications with exceptional efficiency in just weeks,” said Young Jun Yoo, Nextchip’s Chief Marketing Officer.

Márton Fehér, AImotive’s Senior Vice‑President of Hardware Engineering, added, “Apache5 delivers 2‑ to 3‑fold higher CNN performance at the same TOPS rating. Our aiWare Studio SDK empowered the aiDrive team to bring multiple optimized NNs online in days.”

Apache5 targets a full spectrum of advanced ADAS applications, including driver monitoring systems (DMS), smart rear‑view (SRV), autonomous emergency braking (AEB), lane‑departure warning (LDW), and lane‑keeping assist (LKAS).

The aiWare NPU was engineered in partnership with automated driving teams to meet L2–L4 automotive‑grade real‑time inference demands while minimizing host‑CPU usage. It can be integrated into a SoC or deployed as a standalone NPU, with configurable on‑chip memory, floorplan, and external‑memory bandwidth to meet specific customer needs. The latest aiWare3P IP core is the third generation of this high‑performance design.

aiWare Studio is a production‑ready tool that translates prototype neural networks into fully optimized, high‑performance implementations, ensuring that system performance—not just code efficiency—reaches its maximum potential.