← 返回命令列表

Linux command

amdgpu_top 命令

文本

复制后可按需替换文件名、目录或参数。

常用示例

Start the TUI

amdgpu_top

Launch the graphical interface

amdgpu_top --gui

Launch simple TUI mode

amdgpu_top --smi

Dump AMDGPU info

amdgpu_top -d

Output JSON formatted data

amdgpu_top -d --json

Display GPU processes

amdgpu_top -p

Select a specific GPU

amdgpu_top -i [0]

Dump GPU metrics

amdgpu_top --gpu-metrics

说明

amdgpu_top is an AMD GPU monitoring tool that displays real-time utilization data similar to htop but specifically for AMD graphics hardware. It gathers information from performance counters (GRBM, GRBM2), sensors, fdinfo, and the AMDGPU driver to provide comprehensive GPU metrics. The tool offers three display modes: a full-featured TUI (default), a simplified SMI-style TUI, and a graphical interface. It monitors GPU activity, VRAM usage, temperature, power consumption, fan speed, and per-process GPU utilization.

参数

-d, --dump
Dump AMDGPU info including specifications, VRAM, PCI, ResizableBAR, VBIOS, and video caps.
--list
Display a list of available AMDGPU devices.
-J, --json
Output JSON formatted data. Can be combined with the -d or --gpu-metrics option.
--gui
Launch graphical interface mode.
--smi
Launch simple TUI mode similar to nvidia-smi or rocm-smi.
-p, --process
Dump all GPU processes and memory usage per process.
-i _INDEX_
Select GPU instance by index number.
--pci _PATH_
Specify PCI path in domain:bus:dev.func format.
--apu, --select-apu
Select APU instance.
--single, --single-gpu
Display only the selected APU/GPU.
--no-pc
Disable reading performance counters (GRBM, GRBM2). Useful to avoid deactivating power saving features.
-gm, --gpu_metrics, --gpu-metrics
Dump gpu_metrics for all AMD GPUs.
--pp_table, --pp-table
Dump power/performance table from sysfs and VBIOS for all AMD GPUs.
--drm-info
Dump DRM information.
--json_fifo, --json-fifo _PATH_
Output JSON formatted data to a FIFO (named pipe) for other applications and scripts.
--decode-gm, --decode-gpu-metrics _PATH_
Decode the specified gpu_metrics file.
-s _MS_
Set refresh interval in milliseconds for JSON mode (default: 1000ms).
-n _COUNT_
Set maximum iteration count for JSON mode (0 = infinite).
-u, --update-process-index _SEC_
Update interval in seconds for fdinfo (default: 5s).
--dark, --dark-mode
Enable dark mode for TUI/GUI.
--light, --light-mode
Enable light mode for TUI/GUI.
--gl, --opengl
Use OpenGL API for GUI backend.
--vk, --vulkan
Use Vulkan API for GUI backend.
-V, --version
Print version information.

FAQ

What is the amdgpu_top command used for?

amdgpu_top is an AMD GPU monitoring tool that displays real-time utilization data similar to htop but specifically for AMD graphics hardware. It gathers information from performance counters (GRBM, GRBM2), sensors, fdinfo, and the AMDGPU driver to provide comprehensive GPU metrics. The tool offers three display modes: a full-featured TUI (default), a simplified SMI-style TUI, and a graphical interface. It monitors GPU activity, VRAM usage, temperature, power consumption, fan speed, and per-process GPU utilization.

How do I run a basic amdgpu_top example?

Run `amdgpu_top` in a terminal, then adjust file names, paths, flags, or remote targets for your system.

What does -d, --dump do in amdgpu_top?

Dump AMDGPU info including specifications, VRAM, PCI, ResizableBAR, VBIOS, and video caps.