Modern GPU applications often need lightweight, repeatable performance checks that can run outside a full profiling session. A developer might want to confirm whether a workload is throttling, whether data is moving across GPU interconnects as expected, whether ECC counters are increasing, or whether the GPU is spending enough time on useful compute work. These checks are especially useful during application tuning, cluster validation, and regression testing.
On AMD GPU platforms, the ROCm Data Center tool (RDC) provides a command-line interface for collecting this type of telemetry. RDC brings together hardware status, reliability data, and runtime profiling metrics so developers and system operators can observe GPU behavior without opening a full profiler interface. This blog introduces commonly used RDC metrics and compares them with similar DCGM metrics used in NVIDIA environments.
The goal is not to provide a one-to-one mapping for every counter. Instead, this post shows how to think about the most common monitoring categories: hardware health, data transport, error monitoring, runtime utilization, occupancy, and data type activity. These categories give developers a practical starting point for evaluating whether an application is using GPU resources effectively.
