Ambarella Launches CV5 AI Vision Processor: 8K at 30fps in <2W, Redefining Power‑Efficient Imaging

Ambarella has unveiled its fifth‑generation AI vision processor SoC, the CV5, at CES 2021. Built on Samsung’s cutting‑edge 5 nm process, the chip delivers full 8K resolution at 30 fps while consuming less than 2 W. This milestone makes it a game‑changer for intelligent camera systems in automotive, consumer, drone, and robotic applications.

The CV5 couples Ambarella’s proprietary CVflow AI engine with a dual‑core Arm Cortex‑A76 CPU operating up to 1.6 GHz. Together they provide the horsepower required to run a wide range of AI algorithms. With an image‑signal processor capable of ingesting up to 14 camera streams, the chip can simultaneously generate high‑quality footage for human viewers and feed machine‑learning pipelines.

“This is our fifth generation of image processors, with separate processing for video and AI,” said Chris Day, Vice President of Marketing and Business Development at Ambarella. “It’s the first 5 nm chip, enabling low power consumption which is critical for many applications, especially in sports cameras.” Day added, “We work closely with the Samsung fab and have one of the best VLSI design teams in the industry. This means we have good early access to the process libraries thanks to our close partnership.”

In automotive video‑telematics, CV5 can encode multiple streams from front‑seat ADAS cameras, driver‑monitoring, cabin‑monitoring and side‑view lenses. Its CVflow engine runs lane‑departure and forward‑collision‑warning algorithms alongside driver‑drowsiness detection, while the 8K video stream delivers higher‑resolution imagery for better detection accuracy over longer distances.

Day noted that initial interest is strongest in the automotive sector, which accounts for roughly 20 % of Ambarella’s revenue.

The chip’s low power envelope also unlocks next‑generation high‑resolution action, 360° and virtual‑reality cameras. With on‑chip image stabilization and stitching, it supports 8K recording and playback in compact form factors.

In robotics and drone platforms, the CVflow engine accelerates simultaneous localization and mapping, path planning and obstacle‑avoidance. For aerial cinematography, the SoC can deliver 8K video at 60 fps while maintaining real‑time flight control and navigation.

“The key benefit of the CV5 is the ability to add 8K recording across all markets,” Day said. “We’re also the first to include a PCIe interface on the chip, opening new connectivity options. While we’re focused on automotive, sports cameras, drones and VR now, we expect security and surveillance variants later this year.”

The CV5 shares a common SDK and CV toolchain with Ambarella’s other CVflow families, simplifying development across a spectrum of price and performance tiers. The toolkit includes a neural‑network compiler, debugger and full support for industry‑standard frameworks such as PyTorch, ONNX, Caffe and TensorFlow, along with detailed CNN optimization guidelines.

The main features of the CV5 SoC include:

• CVflow architecture with DNN support
• Dual‑core Arm Cortex‑A76 up to 1.6 GHz with NEON DSP extensions and FPU
• High‑speed SLVS‑EC, MIPI‑CSI (C/D PHY) interfaces, connecting up to 14 cameras
• Multi‑channel ISP with up to 8KP60 throughput
• Native support for RGGB, RCCB, RCCC, RGB‑IR and monochrome sensor formats
• Multi‑exposure high‑dynamic‑range (HDR) processing
• Real‑time, hardware‑accelerated fisheye dewarping and lens‑distortion correction (LDC)
• AVC and HEVC encoding up to 8KP60 throughput with multiple stream support
• Rich interface set: 4‑lane PCIe, CAN FD, Gigabit Ethernet, USB 3.1 host/device, three SD‑card controllers, MIPI DSI/CSI‑2 and HDMI outputs
• LPDDR4x/LPDDR5/LPDDR5x support, 64‑bit data bus, up to 32 GB capacity
• Security features: secure boot with TrustZone, secure memory, TRNG, OTP, DRAM scrambling, virtualization
• 5 nm process technology
• 16 × 16 FBGA package, 0.5 mm ball pitch