Product Forms in Inference Scenario
Product forms are distinguished by the PCle operating mode of Ascend 310 AI Processors. If the PCle works in master mode and supports peripherals, this mode is called the RC mode. If the PCle works in slave mode, this mode is called the EP mode.
- The following products support the RC mode: Atlas 200 AI accelerator module and Atlas 200 DK developer kit (model 3000).
The processors of such products directly run the AI service software specified by users and connect to peripherals such as network cameras, I2C sensors, and SPI monitors as slave devices.
- The following products support the EP mode: Atlas 200 AI accelerator module, Atlas 300I inference card, Atlas 500 AI edge station, Atlas 500 Pro AI edge server, and Atlas 800 inference server.
In EP mode, the host acts as the master, and the device acts as the slave. Customers' AI service programs run on the host system. The product functions as a device system and connects to the host system through the PCIe channel. The host system interacts with the device system through the PCIe channel and loads AI tasks to Ascend 310 AI Processors on the device.
Figure 2-12 shows the products and architecture of the two modes.
- The host is an x86 server or ARM server that is connected to the hardware equipped with an Ascend AI Processor. The host leverages the neural network (NN) compute capability provided by the Ascend AI Processor to implement services.
- A device is a hardware backend equipped with an Ascend AI Processor. It is connected to the server over the PCIe interface and provides the required NN computing capability for the server.
