CPU power measures the total power consumption of the entire CPU, reflecting the energy usage during execution. This metric provides insights into the energy efficiency of workloads and the thermal behavior of the processor.
| Metric [W] | Description |
|---|---|
| cpu power | Total power consumed by the CPU |
Balancing performance with power efficiency, such as by optimizing algorithms or leveraging hardware power management features, can improve overall system efficiency.
Core power measures the power consumption of the CPU cores. It provides insight into the energy usage of the core computational units, excluding other components such as uncore (components of the cpu that are not part of the core) or memory subsystems.
| Metric [W] | Description |
|---|---|
| core power | Power consumed by the physical CPU cores only |
Core power is a key metric for analyzing compute-intensive workloads where the focus is on the execution of instructions by the CPU cores. High core power usage typically indicates heavy computational activity, which is expected in compute-bound tasks.
DRAM power measures the power consumption of the main memory subsystem (DRAM). It provides insight into the energy usage of memory-intensive workloads.
| Metric [W] | Description |
|---|---|
| dram power | Power consumed by DRAM |
High DRAM power usage can indicate significant memory activity. To optimize performance and energy consumption, consider improving memory access patterns, reducing unnecessary memory operations, and ensuring efficient data reuse.
System power measures the total power consumption of the entire physical system, including all hardware components such as the CPU, memory, storage, network interfaces, and other peripherals.
| Metric [W] | Description |
|---|---|
| system power | Total power consumed by the entire system |
System power provides a comprehensive view of the energy usage across all components in a machine. High system power may indicate intensive workload activity across multiple subsystems.
GPU power measures the power consumption of the Graphics Processing Unit (GPU) in the system, including its computational cores, memory, and associated components. It also provides the power limit, which represents the maximum allowable power consumption for the GPU.
| Metric [W] | Description |
|---|---|
| gpu power | Power consumed by the GPU subsystem |
| gpu power limit | Maximum allowable power consumption for the GPU |
Monitoring these metrics is useful for workload optimization and ensuring efficient use of system resources. Strategies such as optimizing kernel execution and leveraging GPU power management features can help reduce unnecessary power consumption.
FPGA power measures the power consumption of Field-Programmable Gate Arrays (FPGAs) within the system. This includes the energy used by the FPGA's computational logic, memory, and interconnects. Currently only Xilinx FPGAs are supported.
| Metric [W] | Description |
|---|---|
| fpga power | Power consumed by the FPGAs |
FPGA power is a critical metric for systems leveraging FPGAs for specialized, high-performance workloads such as machine learning, data acceleration, or custom logic processing. High FPGA power usage may indicate intensive computational activity. To optimize FPGA power efficiency, consider refining the FPGA design, minimizing redundant operations, and leveraging low-power configurations or clock gating techniques where applicable.