Penguin Computing Relion 1900e, Technical Manual

Page: 80 / 174

Relion 1900e/2900e Manual
65 Revision 1.3
7.3.10 Thermal Monitoring
The BMC provides monitoring of component and board temperature sensing devices. This monitoring
capability is instantiated in the form of IPMI analog/threshold or discrete sensors, depending on the nature
of the measurement.
For analog/threshold sensors, with the exception of Processor Temperature sensors, critical and non-critical
thresholds (upper and lower) are set through SDRs and event generation enabled for both assertion and de-
assertion events.
For discrete sensors, both assertion and de-assertion event generation are enabled.
Mandatory monitoring of platform thermal sensors includes:
• Inlet temperature (physical sensor is typically on system front panel or HDD back plane)
•
Board ambient thermal sensors
• Processor temperature
•
Memory (DIMM) temperature
• CPU VRD Hot monitoring
•
Power supply inlet temperature (only supported for PMBus*-compliant PSUs)
Additionally, the BMC FW may create “virtual” sensors that are based on a combination of aggregation of
multiple physical thermal sensors and application of a mathematical formula to thermal or power sensor
readings.
7.3.10.1 Absolute Value versus Margin Sensors
Thermal monitoring sensors fall into three basic categories:
•
Absolute temperature sensors – These are analog/threshold sensors that provide a value that
corresponds to an absolute temperature value.
•
Thermal margin sensors – These are analog/threshold sensors that provide a value that is relative to
some reference value.
• Thermal fault indication sensors – These are discrete sensors that indicate a specific thermal fault
condition.
7.3.10.2 Processor DTS-Spec Margin Sensor(s)
Intel ® Server Systems supporting the Intel ® Xeon ® processor E5-2600 v3, v4 product family incorporate a DTS
based thermal spec. This allows a much more accurate control of the thermal solution and will enable lower
fan speeds and lower fan power consumption. The main usage of this sensor is as an input to the BMC’s fan
control algorithms. The BMC implements this as a threshold sensor. There is one DTS sensor for each
installed physical processor package. Thresholds are not set and alert generation is not enabled for these
sensors. DTS 2.0 is implemented on new Intel board generation DTS 2.0 incorporates platform-visible
thermal data interfaces and internal algorithms for calculating the relevant thermal data. As the major
difference between the DTS1.0 and DTS 2.0 is that allows the CPUs to automatically calculate thermal
gap/margin to DTS profile as input for Fan Speed Control. , DTS2.0 helps to further optimize system
acoustics. Please refer to iBL #455822(Platform Digital Thermal Sensor (DTS) Based Thermal Specifications
and Overview – Rev. 1.5) for more details about DTS2.0.
7.3.10.3 Processor Thermal Margin Sensor(s)
Each processor supports a physical thermal margin sensor per core that is readable through the PECI
interface. This provides a relative value representing a thermal margin from the core’s throttling thermal trip
point. Assuming that temperature controlled throttling is enabled; the physical core temperature sensor
reads ‘0’, which indicates the processor core is being throttled.