Reported Statistics

There are two types of thermal sensors:
  • SOC_THERM sensors
  • tmp451 sensors
CPU / GPU / CV / SoC thermal sensors are the on-chip SOC_THERM thermal sensors, which are controlled by BPMP firmware and tegra-bpmp-thermal Linux kernel driver. EXT0, EXT1 are the tmp451 modules, which are from TEXAS INSTRUMENTS. In K5.10, they are controlled by nct1008 Linux kernel driver. In K5.15, the driver is switched to lm90 Linux kernel driver. Each tmp451 module (EXT0 and EXT1) has 2 sensors (a local sensor and a remote sensor). Local: EXT0-local and EXT1-local which measure the board temperature. Remote: EXT0-remote and EXT1-remote which measure the on die/on-chip temperature.

The following table shows the statistics that the tegrastats utility reports.

Statistic

X

Y

Z

 

RAM X/Y (lfb NxZ)

Largest Free Block (lfb) is a statistic about the memory allocator. It refers to the largest contiguous block of physical memory that can be allocated: at most, 4 MB. It can become smaller with memory fragmentation.

The physical allocations in virtual memory can be bigger.

Amount of RAM in use in MB.

Total amount of RAM available for applications.

Z is the size of the largest free block, N the number of free blocks of this size. 


CPU [X%,Y%, , ]@Z

or 


CPU [X%@Z, Y%@Z,...]

X and Y are rough approximations based on time spent in the system idle process as reported by the Linux kernel in /proc/stat.

Load statistics for each of the CPU cores relative to the current running frequency Z, or 'off' in case a core is currently powered down.

Load statistics for each of the CPU cores relative to the current running frequency Z, or 'off' in case a core is currently powered down.

CPU frequency in megahertz. Goes up or down dynamically depending on the CPU workload.

 

GR3D_FREQ X%@[Y,Y,...]

GR3D is the iGPU engine.

Percent of the GR3D that is being used, relative to the current running frequency. Aggregated between GPC0 and GPC1.

GR3D frequency in megahertz for each available GPC (that is, GPC0, GPC1, and so on.)

N/A

 

X1@Y1C X2@Y2C 
X3@Y3C...

X1, X2, X3 denote the list of the available sensors

The sensor name

The sensor's temperature in Celsius

N/A

 

EMC_FREQ @Y

EMC is the external memory controller.

N/A

EMC frequency in megahertz

N/A

 

NVENC @Y
NVENC1 @Y

NVENC is NVIDIA video encoder.

NVENC1 is the second instance of NVENC. NVENC1 may be reported if it is present and enabled.

Load stats is added. If both frequency and load are enabled, the output looks like NVENC X%@Y. If only frequency is enabled: NVENC_FREQ @Y. If only load: NVENC_LOAD X%.

NVENC frequency in megahertz

N/A

 

NVDEC @Y
NVDEC1 @Y

NVDEC is NVIDIA video encoder.

NVDEC1 is the second instance of NVDEC. NVDEC1 may be reported if it is present and enabled.

Load stats is added. If both frequency and load are enabled, the output looks like NVDEC X%@Y. If only frequency is enabled: NVDEC_FREQ @Y. If only load: NVDEC_LOAD X%.

NVDEC frequency in megahertz

N/A

 

VIC @Y

VIC is the video image compositor.

Load stats is added. If both frequency and load are enabled, the output looks like VIC X%@Y. If only frequency is enabled: VIC_FREQ @Y. If only load: VIC_LOAD X%.

VIC frequency in megahertz

N/A

 

NVDLA0 @Y 
NVDLA1 @Y

NVDLA is NVIDIA deep learning accelerator. NVDLA1 is the second instance of NVDEC. NVDLA1 may be reported if it is present and enabled.

 Load stats is added. If both frequency and load are enabled, the output looks like NVDLA0 X%@Y. If only frequency is enabled: NVDLA0_FREQ @Y. If only load: NVDLA0_LOAD X%. NVDLA frequency in megahertz N/A 
 
PVA0_VPU0_LOAD X%
PVA0_VPU1_LOAD X%

PVA is the programmable vision accelerator and VPU is the vector processing unit.PVA0_VPU1 is the second instance of the VPU for PVA0.

 Load statistics N/A N/A