Key Features
Lossless weight compression
Complementing its low-bit depth support, Series3NX products features a lossless weight compression scheme that reduces network model sizes and bandwidth thus increasing overall performance.
Low power consumption
Series3NX products delivers neural network acceleration with low power consumption.
Flexible bit-depth support
Serving as a flexible solution, Series3NX products support neural network bit depths from 16-bit down to 4-bit, reducing bandwidth and increasing performance without compromising inference accuracy.
Security enablement
IMG Series3NX products integrate with the industry-leading security architectures with a flexible infrastructure that enables integration into custom solutions – allowing rightholders to protect their content where required.
AR / VR
High frame rates and engaging immersive experiences across gaming, socialising, exploring and business applications.
Premium ADAS
Vehicles targeting L2+ “hands free” and beyond ADAS features and flexible coverage of both graphics and compute tasks.
Full Features & Specs
Headline Features
- Security enablement
- Lossless weight compression
API support
- IMG DNN API
- Android NN HAL
Framework Support
- Caffe
- TensorFlow
- ONNX
Bus interface
- AXI4
Memory system
- Virtual Memory
TOPS
- 12.5 TOPS
MACs per clock rate
- 8-bit: 4096 MACs/clk
- 16-bit: 1024 MACs/clk
IMG AX3146 NNA 
