Table of Contents
Advertisement
9 Monitoring
9.1 Thermal Sensors
The DPU incorporates the DPU SoC, which operates in the range of temperatures between 0°C and 105°C.
Three thermal threshold definitions impact the overall system operation state:
•
Warning – 105°C: On managed systems only: When the device crosses the 105°C threshold, a Warning Threshold message is issued by the
management SW, indicating to system administration that the card has crossed the warning threshold. Note that this temperature threshold does not
require nor lead to any action by hardware (such as DPU shutdown).
•
Critical – 115°C: When the device crosses this temperature, the firmware automatically shuts down the device.
•
Emergency – 130°C: If the firmware fails to shutdown the device upon crossing the critical threshold, the device automatically shuts down upon
crossing the emergency (130°C) threshold.
The DPU's thermal sensors can be read through the system's SMBus. The user can read these thermal sensors and adapt the system airflow following the
readouts and the needs of the above-mentioned SoC thermal requirements.
9.2 Heatsink
The heatsink is attached to the DPU by three screws to dissipate the heat from the SoC.
The DPU SoC has a thermal shutdown safety mechanism that automatically shuts down the DPU in cases of high-temperature events, improper thermal
coupling, or heatsink removal.
Refer to the below table for heatsink details per card configuration. For the required airflow (LFM) per OPN, please refer to Specifications.
Card Configuration
OPN
Single-slot DPUs
900-9D3B4-00EN-EA0, 900-9D3B4-00PN-EA0,
900-9D3B6-00CV-AA0, 900-9D3B6-00SV-AA0,
900-9D3B6-00CC-AA0, 900-9D3B6-00SC-AA0,
900-9D3B6-00CC-EA0, 900-9D3B6-00SC-EA0
Dual-slot DPUs
900-9D3B6-00CN-AB0, 900-9D3B6-00SN-AB0
NVIDIA BlueField-3 DPU User Guide
Maximum Dimensions
Length, Width, Height:
139.6mm x 92.7mm x 10.2mm
Length, Width, Height:
139.6mm x 92.7mm x 29.3mm
| 82
Advertisement
Table of Contents
