(base) root@db4eb9b72711:/data# mx-smi -hUSAGE: mx-smi [-hjLrsvy] [--count-ecc] [--show-ap-usage] [--show-apusage-toggle] [--show-board-power] [--show-clk-tr] [--show-clock] [--show-core-usage] [--show-eeprom] [--show-fan-speed] [--show-hbm-bandwidth] [--show-hwinfo] [--show-memory] [--show-metaxlink-bandwidth] [--show-pcie-bandwidth] [--show-pcie-slot] [--show-pmbus-power] [--show-power-state] [--show-pptable-version] [--show-sn] [--show-sysinfo] [--show-temperature] [--show-unavailable-reason] [--show-usage] [--show-vbios] [--show-version] [--show-vpu] [--set-pci-speed <[1, 5]>| --show-pcie] [--kill-all-process|--show-all-process| --show-process] [--flr|--vfflr] [--set-ecc-state <[0, 1]>| --show-ecc-state] [-U <bin>|-u <bin>|--force-vbios-flash <bin>] [--set-op-mode <[0, 1]>|--show-op-mode] [--set-persistence-mode <[0, 1]>|--show-persistence-mode] [--set-dpm-max <DPM IP,DPM level>|--show-dpm-max] [--set-power-mode <[0, 1]>| --show-power-mode] [-i <deviceId>] [-l <ms>] [-o <output.csv>] [-t <[60, 36000]>] [--die <dieId>] [--dump-vbios <>] [--set-apusage-toggle <[0, 1]>] [--show-clocks <all|clock IP>] [--show-dpm <all|cur>] [--show-event <aer_ue|aer_ce|synfld|dbe| mmio|all>] [subsystem]Where: --show-temperature Display temperature --show-pmbus-power Display pmbus voltage, current and power --show-board-power Display board voltage, current and power --show-sn Display chip serial --show-memory Display memory information in KBytes --show-eeprom Display eeprom information --show-version Display maca, bios, driver, firmware version --show-usage Display VPUE,VPUD,GPU usage --show-power-state Display power state -L, --list List all devices discovered on the host --show-unavailable-reason Display device unavailable reason --show-hbm-bandwidth Display hbm bandwidth in MBytes/s --show-vpu Display encoder and decoder codec status --show-pcie-bandwidth Display pcie bandwidth in MBytes/s -s, --summary Display device summary information --show-core-usage Display single core usage --show-vbios Display vbios information --show-pcie-slot Display the device pcie slot --show-fan-speed Display fan speed information --show-hwinfo Display hardware information -j, --json Display gpu information in json format -y, --yes Automatic yes to prompts for PF warm reset and run non-interactively --show-clock Display brief clock frequency --show-sysinfo Display device system info --show-clk-tr Display current clocks throttling reason(only support C-class device) -o <output.csv>, --output <output.csv> Write show commands' data to file in csv format supported show commands: show-temperature, show-board-power, show-version, show-usage, show-memory, show-pcie, show-dpm cur, show-pmbus-power, show-vpu, show-fan-speed, show-metaxlink-bandwidth, show-pcie-bandwidth --show-dpm <all|cur> Display DPM performance levels -l <ms>, --loop <ms> Display until Ctrl+C at specified millisecond interval -i <deviceId>, --index <deviceId> Target specific devices. The specified id may be the GPU's index in the natural enumeration, chip serial number, uuid, or PCI bus ID(as domain:bus:device.function in hex). Multiple devices can be combined as '0,1,5', '0-2' '0-4,6', 'GPU-2f8ab36b-882a-0248-494e-0da5d8ae9698, 0000:03:00.0', 'all' --die <dieId> Target specific die index. --dump-vbios <> Dump vbios data Usage: mx-smi --dump-vbios ./mxvbios-xxx.bin -t 90 mx-smi --dump-vbios ./mxvbios-xxx.bin # default time limit is 60s -t <[60, 36000]>, --timelimit <[60, 36000]> Time limit for upgrade or dump command, unit: second --show-event <aer_ue|aer_ce|synfld|dbe|mmio|all> Display kmd events info --show-clocks <all|clock IP> Display target IPs' clocks clock IP: csc, xcore, mc0, mc1, vpue, vpud, soc, dnoc, ccx clock IP(MXC600): csc, xcore, soc, dnoc multiple IPs can be combined with comma --set-apusage-toggle <[0, 1]> Disable/enable collecting xcore ap usage state: 0(disable) 1(enable) Note: The device must be specified by -i, --index. -r, --reset Trigger warm reset of the gpu. Reset operations are not guaranteed to work in all cases and should be used with caution. --show-metaxlink-bandwidth Display MetaXLink bandwidth in MBytes/s --show-pptable-version Display pptable version --show-ap-usage Show xcore ap usage --show-apusage-toggle Show xcore ap usage toggle states --count-ecc Display Ecc error count Either of: --show-pcie Display pcie speed and width --set-pci-speed <[1, 5]> Set pci speed by changing pci generation(only support N-class device) Either of: --show-process Show process currently running --show-all-process show all process opening device file --kill-all-process Kill all process opening device file Either of: --flr Function level reset --vfflr Perform VF function level reset from PF Either of: --show-ecc-state Display ecc's state --set-ecc-state <[0, 1]> Disable/enable ecc ecc state: 0(disable) 1(enable) Note: The device must be specified by -i, --index. Either of: -u <bin>, --vbios-upgrade <bin> Upgrade vbios for all devices Use -U/--force-vbios-upgrade to update specified devices Usage: mx-smi -u ./mxvbios-xxx.bin -t 90 mx-smi -u ./mxvbios-xxx.bin # default time limit is 60s -U <bin>, --force-vbios-upgrade <bin> Force upgrade vbios Usage: mx-smi -U ./mxvbios-xxx.bin -t 90 mx-smi -U ./mxvbios-xxx.bin # default time limit is 60s --force-vbios-flash <bin> Force flash vbios for not available devices Usage: mx-smi --force-vbios-flash ./mxvbios-xxx.bin Either of: --show-op-mode Display system operation mode --set-op-mode <[0, 1]> Set system operation mode mode: 0(Normal mode) 1(Maintenance mode) Either of: --show-persistence-mode Display persistence mode --set-persistence-mode <[0, 1]> Disable/enable persistence mode persistence mode: 0(disable) 1(enable) Note: The device must be specified by -i, --index. Either of: --set-dpm-max <DPM IP,DPM level> Set dpm IP max performance level format: <DPM IP>,<DPM level>; (e.g. "mc,1") DPM IP: <xcore|mc|soc|dnoc|vpue|vpud|ccx> DPM level: get from cmd: mx-smi --show-dpm all Note: The device must be specified by -i, --index. (N-class device is not support) --show-dpm-max Display dpm max performance level (N-class device is not support) Either of: --show-power-mode Display power mode --set-power-mode <[0, 1]> Set power mode Power mode: 0(Normal) 1(High) Note: The device must be specified by -i, --index. --, --ignore_rest Ignores the rest of the labeled arguments following this flag. -v, --version Displays version information and exits. -h, --help Displays usage information and exits. subsystem <ras|topo|dmon|mxlk|misc|om|cxl|eth|vm|sgpu> The desired subsystem to be accessed. Subsystems available: ras Device RAS management [mx-smi ras -h for more info] topo Display device topology [mx-smi topo -h for more info] dmon Display device stats in scrolling format [mx-smi dmon -h for more info] om Display optical module info [mx-smi om -h for more info] vm Device virtualization management [mx-smi vm -h for more info] sgpu Device slice gpu interface [mx-smi sgpu -h for more info] mxlk Device metaxlink management [mx-smi mxlk -h for more info] cxl Display cxl device info [mx-smi cxl -h for more info] eth Display eth info [mx-smi eth -h for more info] misc Various utility commands [mx-smi misc -h for more info] MetaX GPU System Management Interface(base) root@db4eb9b72711:/data#