SolidRun announces the i.MX 8M Mini SOM Dev Board with Gyrfalcon AI acceleration for next-gen Edge AI applications
Working For Notebookcheck
Are you a techie who knows how to write? Then join our Team! Especially English native speakers welcome!
SolidRun has announced a new System on Module (SOM) to accelerate AI development for the Edge. Dubbed the i.MX 8M, the mini SOM integrates a Gyrfalcon Lightspeeur 2803S Neural Accelerator chip that supports accelerating Edge AI applications created using the TensorFlow, Caffe, and PyTorch frameworks.
The i.MX 8M comes in three variants — Solo, Dual, and Quad based on the number of NXP ARM Cortex-A53 1.8 GHz CPU cores on offer. The SOM supports up to 4 GB LPDDR4 RAM (for the Dual and Quad SKUs) and offers Bluetooth and Wi-Fi (802.11n), PCIe 2.0 interface, 20 32-bit audio channels, MIPI-DSI display interface, and 1080p60 video encoders/decoders on a 47mm x 30mm package.
The Gyrfalcon Lightspeeur 2803S Neural Accelerator offers dedicated AI processing capabilities to the SOM without having to depend on the cloud. The 9mm x 9mm accelerator is based on Gyrfalcon's Matrix Processing Engine (MPE) architecture and is rated at 24 TOPS/W per chip. The Lightspeeur 2803S sips just 700 mW of power at a frequency of 250 MHz and can attain a peak performance of 16.8 TOPS at 300 MHz. The i.MX 8M SOM is designed for fan-less operation and is ideal for digital assistant solutions, self-driving cars, security systems, digital signage, and more.
SolidRun is also offering the HummingBoard Pulse carrier board, a Single Board Computer (SBC) that integrates the i.MX 8M SOM with additional Raspberry Pi-like connectivity features including HDMI, USB Type-C, microUSB, USB 3.0, mPCIe, M.2 expansion slots, microSD slot, SIM card holder, Gigabit Ethernet, and audio I/O.
With SBCs and devices such as smartphones getting more powerful, AI processing on the Edge is becoming popular by the day as it eliminates most of the latency, privacy, and reliability issues associated with cloud-based AI. In 2017, Intel introduced the Movidius Neural Compute Stick that brought end-inference capabilities on a USB 3.0 stick for US$79. In March this year, NVIDIA launched the Jetson Nano AI computer featuring a Maxwell GPU with 128 CUDA cores based on the scalable NVIDIA EGX platform for US$99. Such affordable Edge AI processing capabilities means less reliance on the cloud and more focus on faster AI operations and data privacy.
The SolidRun i.MX 8M Mini SOM is slated to be available soon with prices starting from US$56.
SolidRun Press Release
Top 10 Smartphones
Smartphones, Phablets, ≤5-inch, Camera SmartphonesNotebookcheck's Top 10 Smartphones under 160 Euros