Performance Specifications

This section describes only the performance metric reference data of AI processor corresponding to each product. The performance specifications vary according to the memory and CPU configurations of each product. For details about the performance specifications of each product, see the official description on the Ascend community.

This section takes a single device as an example to describe the performance specifications in basic scenarios, as shown in the table below. The performance at other resolutions can be estimated in the same way.

Performance Specifications ( Atlas A2 training product / Atlas A2 inference product )

Table 1 AI processor performance indicator reference data of Atlas A2 training product / Atlas A2 inference product

Scenario

Total Frame Rate

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 10)

n*1440fps

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (n > 10)

14400fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 10)

n*360fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n > 10)

3600fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 10)

n*90fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n > 10)

900fps

Performance Specifications ( Atlas 200I/500 A2 inference product )

In the single-device scenario, the performance specifications are as follows (one channel corresponds to one thread; n channels correspond to n threads).

Table 2 AI processor performance indicator reference data of Atlas 200I/500 A2 inference product

Scenario

Total Frame Rate

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (n < 2)

n*800fps

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (n ≥ 2)

1600fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n < 2)

n*200fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n ≥ 2)

400fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n < 2)

n*100fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n ≥ 2)

200fps

A VPC batch processing API (whose name contains "batch", for example, acldvppVpcBatchCropAsync) allows parallel image processing of multiple VPC hardware units in the DVPP. Therefore, a single channel can reach the maximum total frame rate.

Performance Specifications ( Atlas inference product )

The following performance data is obtained when the aclrtSynchronizeStream interface is executed once after 64 asynchronous media data processing tasks are delivered on a stream. The host server is configured as follows: The CPU dominant frequency is Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz, and the number of CPU cores is 14.

In the single-device scenario, the performance specifications are as follows (one channel corresponds to one thread).
Table 3 AI processor performance indicator reference data of Atlas inference product

Scenario

Total Frame Rate

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 12)

n*360fps

  • Input resolution: 1080p (1920 × 1080)
  • Output resolution: 1080p (1920 × 1080)
  • Input/Output format: YUV420SP
  • n channels (n > 12)

4320fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 12)

n*90fps

  • Input resolution: 4K (3840 × 2160)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n > 12)

1080fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (1 ≤ n ≤ 12)

n*22.5fps

  • Input resolution: 8K (7680 × 4320)
  • Output resolution: 4K (3840 × 2160)
  • Input/Output format: YUV420SP
  • n channels (n > 12)

270fps

A VPC batch processing API (whose name contains "batch", for example, acldvppVpcBatchCropAsync) allows the parallel image processing on multiple VPC hardware units in the DVPP. Therefore, a single channel can reach the maximum total frame rate.

Performance Specifications ( Atlas training product )

The following performance data is obtained by invoking the aclrtSynchronizeStream interface once after 10 asynchronous media data processing tasks are delivered on a stream.

  • For non-8K images, the performance specifications in basic scenarios with a single device are as follows (one channel corresponds to one thread):
    Table 4 AI processor performance indicator reference data of Atlas training product

    Scenario

    Total Frame Rate

    • Input resolution: 1080p (1920 × 1080)
    • Output resolution: 1080p (1920 × 1080)
    • Input/Output format: YUV420SP
    • n channels (n < 16)

    n*360fps

    • Input resolution: 1080p (1920 × 1080)
    • Output resolution: 1080p (1920 × 1080)
    • Input/Output format: YUV420SP
    • n channels (n ≥ 16)

    5760fps

    • Input resolution: 4K (3840 × 2160)
    • Output resolution: 4K (3840 × 2160)
    • Input/Output format: YUV420SP
    • n channels (n < 16)

    n*90fps

    • Input resolution: 4K (3840 × 2160)
    • Output resolution: 4K (3840 × 2160)
    • Input/Output format: YUV420SP
    • n channels (n ≥ 16)

    1440fps

  • For 8K images, the performance specifications in typical scenarios of a single device are as follows (one channel corresponds to one thread):
    Table 5 AI processor performance indicator reference data of Atlas training product

    Scenario

    Total Frame Rate

    • Input resolution: 8K (7680 × 4320)
    • Output resolution: 1080p (1920 × 1080)
    • Input/Output format: YUV420SP
    • n channels (n < 16)

    n*25fps

    • Input resolution: 8K (7680 × 4320)
    • Output resolution: 1080p (1920 × 1080)
    • Input/Output format: YUV420SP
    • n channels (n ≥ 16)

    400fps

    • Input resolution: 8K (7680 x 4320)
    • Output resolution: 4K (3840 × 2160)
    • Input/Output format: YUV420SP
    • n channels (n < 16)

    n*25fps

    • Input resolution: 8K (7680 x 4320)
    • Output resolution: 4K (3840 × 2160)
    • Input/Output format: YUV420SP
    • n channels (n ≥ 16)

    400fps

    A VPC batch processing API (whose name contains "Batch", for example, acldvppVpcBatchCropAsync) allows the parallel image processing on multiple VPC hardware units in the DVPP. Therefore, a single channel can reach the maximum total frame rate.