The main computing resources included in the Da Vinci architecture computing unit are?

Question

Options

• AVector calculation unit
• BScalar Computing Unit
• CTensor computing unit
• DMatrix calculation unit

• A
  72% (28)
• C
  28% (11)

Explanation

The Da Vinci architecture (Huawei's Ascend AI processor core) contains three distinct compute resources: a Vector Calculation Unit (A) for element-wise vector operations like activations, a Scalar Computing Unit (B) for control flow and scalar arithmetic, and a Matrix Calculation Unit (D) - often called the Cube unit - which accelerates matrix multiplications central to neural network inference.

Option C ("Tensor computing unit") is the distractor; while the Matrix/Cube unit effectively processes tensor data, "Tensor Computing Unit" is not the correct architectural term for Da Vinci - it conflates Huawei's naming with Google TPU terminology. The three units (Vector, Scalar, Matrix) map directly to the three classes of operations in deep learning: control logic (Scalar), activation/normalization (Vector), and convolution/matmul (Matrix).

Memory tip: Think V-S-M - "Very Smart Matrix" - Vector, Scalar, Matrix. The odd one out is "Tensor," which sounds plausible but is a naming trap borrowed from a different vendor's architecture.

No community discussion yet for this question.