English
You are here:
Low-power, High Performance AI Accelerator
NovuTensor™ IP cores provide highly configurable AI acceleration capabilities with industry-leading performance efficiency. This patented, domain-specific architecture minimizes AI software work to greatly reduce Total Cost of Ownership. NovuTensor IP cores can be configured to consume less than 150mW in 1mm2 (including SRAM) or up to 18TOPS per core running at 1GHz and under 2W
The NovuTensor core is an AI-specific IP block to accelerate Convolutional Neural Network processing. NovuMind’s best-in-class AI compute density, power efficiency, and memory bandwidth utilization makes it ideal for low-power embedded SoCs that require heavy AI workloads. NovuTensor comes with full SDK support for state-of-the-art model topologies and automatic conversion of pre-trained networks for rapid prototyping and production deployment.
The NovuTensor core is an AI-specific IP block to accelerate Convolutional Neural Network processing. NovuMind’s best-in-class AI compute density, power efficiency, and memory bandwidth utilization makes it ideal for low-power embedded SoCs that require heavy AI workloads. NovuTensor comes with full SDK support for state-of-the-art model topologies and automatic conversion of pre-trained networks for rapid prototyping and production deployment.
View Low-power, High Performance AI Accelerator full description to...
- see the entire Low-power, High Performance AI Accelerator datasheet
- get in contact with Low-power, High Performance AI Accelerator Supplier
Block Diagram of the Low-power, High Performance AI Accelerator IP Core
