A200T A3 Box8 SuperPoD/Atlas 800I A3 SuperPoD/Atlas 800T A3 SuperPoD

The node affinity of the A200T A3 Box8 SuperPoD Server, Atlas 800I A3 SuperPoD Server, and Atlas 800T A3 SuperPoD Server is basically the same as that of the Atlas 900 A3 SuperPoD.

Ascend AI processors used on these products are high-performance AI processors developed by Huawei. Every two Ascend AI processors, for example, Ascend AI processors 0 and 1, are connected through SIO to form a HiAM module. Each A200T A3 Box8 SuperPoD Server, Atlas 800I A3 SuperPoD Server, and Atlas 800T A3 SuperPoD Server contains eight HiAM modules. The HiAM modules are connected through HCCS.

The number of Ascend AI processors allocated to a job can be 1, 2, 4, 6, 8, 10, 12, 14, or 16. The allocated Ascend AI processors must occupy the entire node. If the number of allocated Ascend AI processors is an even number, the entire HiAM module must be occupied. For example, if a job needs two Ascend AI processors and the remaining Ascend AI processor IDs on the node are 0, 2, 3, and 4, only Ascend AI processors 2 and 3 can be used, because they are in the same HiAM module. For a distributed job, the number of Ascend AI processors that can be allocated is 2, 4, 6, 8, 10, 12, 14, or 16.