Coral TPU Mini PCIe Edge Accelerator Module
Coral TPU Mini PCIe Edge Accelerator Module uses Google's Edge TPU coprocessor to provide local ML inferencing. The half-size card is designed to fit any standard Mini PLIe slot to easily integrate into ARM and x86 platforms running Debian Linux 6.0 or higher.
The Google Edge TPU coprocessor is an ASIC that adds fast ML inferencing to systems to reduce latency, increase data privacy, and remove the need for constant high-bandwidth connectivity. The Edge TPU can execute state-of-the-art mobile vision models such as MobileNet v2 at 400 FPS in a power-efficient manner. It can perform 4 trillion operations (tera-operations) per second (TOPS) using 0.5 watts for each TOPS (2 TOPS per watt). The Accelerator delivers 4 TOPS at 2 Watts of power consumption.
Coral, a division of Google, helps build intelligent ideas with their platform for local AI.
Features
- Google Edge TPU ML accelerator coprocessor
- High-speed TensorFlow Lite inferencing
- Power efficiency at .5 watts for each TOPS of performance
- Supports Debian Linux on host CPU
- Dual A+E keyed
Specifications
- 30.00 x 26.80 x 2.55mm
- 3.6g
- Half-mini PCIE card hardware interface
- PCIe Gen2 x1 serial interface
- 3.3V ± 10% DC supply
- -40°C to 85ºC (storage) temperature
- -20ºC to 70ºC (operating) temperature
- 0 to 90% (non-condensing) Relative humidity
- ESD3 compliance - 1kV HBM, 250V CDM
- Shock
- 100 G, 11 ms (persistent)
- 1000 G, 0.5 ms (stress)
- 1000 G, 1.0 ms (stress)
- Vibration (random/sinusoidal)
- 0.5 Grms, 5 - 500 Hz (persistent)
- 3 Grms, 5 - 800 Hz (stress)
- 11°C/W thermal resistance (junction to the top of shield can)
- 100°C maximum Edge TPU junction temperature
Applications
- Embedded platforms
- Mini-PCs
- Industrial gateways
- IoT devices
- Workstations
Dimensions (mm)
Pin Positions
See all Coral Accelerators
Coral Edge TPU ML Accelerator Modules
Designed to execute neural network processing for machine learning at high speed and low power cost.
Coral USB Accelerator
USB Type-C socket connects to any Linux-based system to perform accelerated ML.