Blumind's all analog AMPL™ neural signal processor IP for edge AI applications is designed on standard advanced node CMOS technology and can be integrated into high volume SoC and MCU products.
Ideal for always on voice detection, key word spotting. speech-to-intent commands (10), time series health or industrial sensor data etc.
Input from analog (or digital sensors) our IP has a fall-through neural network architecture that results in only active neuron consumming power.
Our NN core is tiny taking ~1mm2 and about 2x that area including analog preprocessing and 2 second audio buffer options.
AI Inference IP. Ultra-low power, tiny, std CMOS. ~ 100K parameter RNN
Overview
Key Features
- Ultra Low power
- Standard CMOS
- Small area
- Standard software tools Pytorch, Tensorflow
- Silicon proven
Benefits
- Ultra low power
- Compute-in-transistor tiny form factor
- standard CMOS ptocess (22nm/28nm)
Applications
- Smart watches
- True wireless stereo headphones
- Fitness bands
- wearables
- security
- smoke detectors
- remote controls
- smart sensors
Technical Specifications
Foundry, Node
GF 22nm, TSMC 22nm
Maturity
Test Chip Verified
Availability
Contact Blumind
GLOBALFOUNDRIES
Pre-Silicon:
22nm
FDX
SMIC
Pre-Silicon:
28nm
HK
TSMC
Pre-Silicon:
22nm
,
28nm
UMC
Pre-Silicon:
22nm
,
28nm
Related IPs
- AI inference processor IP
- RISC-V ARC-V RMX-100 Ultra-low Power 32-bit Processor IP for Embedded Applications
- ARC NPX Neural Processing Unit (NPU) IP supports the latest, most complex neural network models and addresses demands for real-time compute with ultra-low power consumption for AI applications
- IP cores for ultra-low power AI-enabled devices
- 2D (vector graphics) & 3D GPU IP A GPU IP combining 3D and 2D rendering features with high performance, low power consumption, and minimum CPU load
- Ultra-Low Power Fractional-N digital PLL for IoT Applications in 40nm CMOS