Optimizing AI server performance requires targeted tuning across GPU compute settings, memory management, data pipe