-
-
Notifications
You must be signed in to change notification settings - Fork 170
Description
Geniatech AIM-M2 Accelerator Module M.2 M-Key (4 lane PCIE Gen4) Form Factor (Brand name "Geniatech" / "XMSJ" from Guangdong, China?)
It is just an M.2 module with based on Kinara Ara-2 (40 TOPS AI accelerator SoC) and on-device RAM so should be able to use Kinara SDK and models built for their Kinara Ara-2 Edge AI processor:
Geniatech AIM-M2 Accelerator Module specifications sais "Up to 16GB LPDDR4(X)" onboard storage so assume they maybe mean it comes if different models with different RAM sizes?
Based on Kinara Ara-2 (40 TOPS AI accelerator SoC)
Geniatech website say they charge $188 for a single sample if order direct from them there. Alao a relativly low bulk price on Alibaba, but very high price from the one reseller found on Aliexpress:
Quote:
40 TOPS M.2 AI Acceleration Module
Geniatech AI accelerator module, with an M.2 form factor, integrate seamlessly into PCs and edge devices for real-time deep neural network inferencing with low power consumption across diverse applications.
- 40 TOPS, supports LLaMA 2.0 and YOLOv8
- M.2 (M Key 2280) interface, 4-lane PCIe Gen4
- AI framework compatibility with TensorFlow, TorchScript, PyTorch, ONNX, Caffe, MxNet
- Processor compatibility tested with NXP, Nvidia, Qualcomm, Xilinx
- Optimized for transformers and generative AI models
- Multi-chip support for scalable AI solutions
- Stable Diffusion 1.4 ~10s (20 iterations)
- ResNet50 2ms latency
PS: Note that Kinara also has its own KM-2 M.2 and KU-2 USB 3.0 modules/reference designs and recommends 2GB of RAM for conventional AI, and 8GB of RAM for generative AI.
