Beijing Plink AI Technology Co., Ltd
Beijing Plink AI is an expert of cloud-to-end one-stop solution.
Add to Cart
Ascend Atlas 200 AI Module 8 GB 128 bits Ubuntu Deep Learning Carrier Board Edge Computing
Hua Wei Ascend Atlas 200 AI Accelerator Module
The Hua Wei Ascend Atlas 200 AI accelerator module has eight Cortex-A55 cores and provides common peripheral ports such as I2C, USB, SPI, and RGMII. It can be used as an embedded system CPU.You can burn the OS to the embedded multimedia controller (eMMC) flash or an SD card. After simple configuration, the ARM CPU in the Atlas 200 AI accelerator module can run users' AI service software.Generally, in this application mode, the Atlas 200 AI accelerator module is connected to simple external devices such as IP cameras (IPCs), I2C sensors, and Serial Peripheral Interface (SPI) displays.The coprocessor application mode and main processor application mode of the Atlas 200 AI accelerator module are similar. The ARM processor of the Hua Wei Ascend Atlas 200 AI accelerator module can still run users' AI service software. The difference is that when the Atlas 200 AI accelerator module is used as a coprocessor, the system has a main processor. The main processor controls operations on the Atlas 200 AI accelerator module, such as peripheral access, power-on, hibernation, and wakeup. The external interfaces required by users' AI service software are also transferred by the main processor.The main processor can control the Hua Wei Ascend Atlas 200 AI accelerator module to enter the deep sleep state through the GPIO pins. When necessary, the Hua Wei Ascend Atlas 200 AI accelerator module can quickly wake up to process AI services. This mechanism reduces the power consumption.
Specifications about Hua Wei Ascend Atlas 200
| Feature | Specification |
| AI processor | Ascend 310 AI Processor Two Da Vinci AI cores Eight A55 ARM cores (maximum frequency: 1.6 GHz) |
| AI compute powera | Half precision (FP16): 4/8/11 TFLOPS Integer precision (INT8): 8/16/22 TOPS |
| Memory | 8G 128 bits LPDDR4X |
| Storage | 64 MB eMMC 4.5 |
| Encoding/Decoding capability | H.264/H.265 decoder, 20-channel 1080p (1920 x 1080) 25 FPS, YUV420 H.264/H.265 decoder, 16-channel 1080p (1920 x 1080) 30 FPS, YUV420 H.264/H.265 decoder, 2-channel 4K (3840 x 2160) 60 FPS, YUV420 H.264/H.265 encoder, 1-channel 1080p (1920 x 1080) 30 FPS, YUV420 JPEG decoding at 1080p (1920 x 1080) 256 FPS and encoding at 1080p (1920 x 1080) 64 FPS, up to 8192 x 4320 resolution PNG decoding at 1080p (1920 x 1080) 24 FPS, up to 4096 x 2160 resolution |
| Temperature | Operating temperature: -25°C to +80°C (-13°F to +176°F) Storage temperature: -25°C to +85°C (-13°F to +185°F) |
| Humidity (RH, non condensing) | Operating humidity: 5% to 90% Storage humidity: 5% to 95% |
| Power consumption | Operating voltage: 3.5 V to 4.5 V; recommended typical value: 3.8 V Typical power consumption – 4 GB: 6.5 W – 8 GB: 9.5 W |
| Maximum altitude | < 5000 m When the altitude is between 1800 m (5905.44 ft.) and 5000 m (16404 ft.), the maximum operating temperature decreases by 1°C (1.8°F) for every increase of 220 m (721.78 ft.) in altitude. |
| Dimensions (H x W x D) | 8.5 mm x 52.6 mm x 38.5 mm (0.33 in. x 2.07 in. x 1.52 in.) NOTE The connector model of the Atlas 200 AI accelerator module is fixed. You can select male connectors with different heights to determine the height of the Atlas 200 AI accelerator module. |
| Net weight | 30g |
| Operating system (OS) | Ubuntu 16.04 |
| Deep learning framework | TensorFlow, Caffe |
