IMG 4NX-MC6
IMG Series4 Neural Network Accelerator (NNA)
The super-high performance hexa-core solution for neural network acceleration
IMG Series4 next-generation neural network accelerator (NNA) is ideal for advanced driver-assistance systems (ADAS) and autonomous vehicles such as robotaxis. The range of cores incorporates sophisticated technical features including Imagination Tensor Tiling, advanced safety mechanisms and intelligent workload management.
Imagination’s low-power and functionally safe NNA architecture is optimised to run full network inferencing and executes multiple operations in a single pass to maximise performance per watt and deliver its industry-leading energy efficiency.
The cores are assembled on a cluster design with per-core on-chip memory and an internal bus and have been designed in 1, 2, 4, 6 and 8-core multi-core configurations. These can then be used as building blocks for further configurations.
The IMG 4NX-MC6 has been developed from the multi-core codebase and features Imagination Tensor Tiling which can reduce bandwidth by up to 90%. It features 75 TOPS at less than three watts and is available to license now.
Memory system
- Virtual memory
TOPS
- 75 TOPS @1.5 GHz
MACs per clock rate
- 8-bit: 24,576 total, 4096 MAC/clk per core
- 8×16-bit: 12,288 total, 2048 MAC/clk per core
- 16-bit: 6,144 total, 1024 MAC/clk per core
Architecture
Markets
- Automotive:
- ADAS
- Autonomous driving
- Data centre
Features
- Imagination Tensor Tiling
- Lossless weight compression
- Low latency
- High-density neural network acceleration
- Wide and variable bit-depth (4 to 16 bits)
- Dynamic fixed-point data type
- Per filter exponent selection
- Memory Management Unit
- Wide range of elementwise unit operations
- Embedded configurable on-chip memory