Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ

Bài kiểm tra về tuổi thọ pin điện thoại và tính năng AI

Hiểu lý do tại sao mức sử dụng pin AI của điện thoại lại khác nhau tùy theo khối lượng công việc, thiết bị, đường dẫn xử lý, cảm biến, mạng và lập lịch chạy nền; học cách đo lường nó.

Đường dẫn hướng dẫn liên quanTính năng AI Tuổi thọ pin điện thoại
Câu hỏi 1 của 8

Why can two AI features have different battery impact?

Xem trước mọi câu hỏi8 câu hỏi

Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.

  1. Why can two AI features have different battery impact?
    • On-device inference always consumes less battery than cloud inference
    • They can differ in compute duration, sensors, display use, network, and scheduling
    • Battery use is determined only by the model’s parameter count
    • A cloud request requires all model inference to run locally
  2. Which testing design helps compare an AI task’s battery effect?
    • Use the same device and task while also changing screen brightness and network conditions
    • Treat response time alone as a direct energy measurement
    • Compare one run with uncontrolled background apps against another
    • Repeat a defined test against an idle or non-AI baseline on the same device
  3. What can Android profiling help a developer inspect on the device being tested?
    • System activity such as CPU, memory, timing, and power behavior
    • A universal energy cost per model call, independent of device and workload
    • The factual accuracy of generated model responses
    • The remote data center’s power draw for a cloud inference request
  4. Which set gives the most complete phone-side battery comparison for a cloud AI request?
    • Server compute and model size, without phone activity
    • Radio and display time alone
    • Network, screen time, request duration, and any local work
    • Local inference only, excluding network and display activity
  5. What do Android standby-bucket limits describe?
    • A fixed energy cost for each model token
    • How system state can restrict scheduled app work; they are not AI energy estimates
    • A guarantee that an app’s background job runs immediately
    • The maximum size of a model loaded into RAM
  6. Which statement about an NPU and CPU is supported by this guide?
    • An NPU always uses less energy for every task
    • A CPU cannot run model inference
    • An NPU removes sensor and network costs
    • Efficiency depends on workload, chip, model, acceleration, and duration
  7. What should a report include when it claims an AI feature drains a certain amount?
    • Device and workload, but no software version or measurement method
    • Software and network conditions, but no device or task definition
    • Device, workload, software, network conditions, and measurement method
    • A single run without stating the phone, workload, or measurement method
  8. If battery use rises during a feature, what is a useful next diagnostic step?
    • Check diagnostics for foreground compute, sensors, radio, and background work
    • Assume background sensors caused the drain without checking diagnostics
    • Disable every permission before identifying the source
    • Replace the battery before comparing the feature against a baseline