Adaptive Thermal Monitor
The purpose of the Adaptive Thermal Monitor is to reduce processor IA core power consumption and temperature until it operates below its maximum operating temperature. Processor IA core power reduction is achieved by:
- Adjusting the operating frequency (using the processor IA core ratio multiplier) and voltage.
- Modulating (starting and stopping) the internal processor IA core clocks (duty cycle).
The Adaptive Thermal Monitor can be activated when the package temperature, monitored by any Digital Thermal Sensor (DTS), meets its maximum operating temperature. The maximum operating temperature implies maximum junction temperature TjMAX.
Reaching the maximum operating temperature activates the Thermal Control Circuit (TCC). When activated the TCC causes both the processor IA core and graphics core to reduce frequency and voltage adaptively. The Adaptive Thermal Monitor will remain active as long as the package temperature remains at its specified limit. Therefore, the Adaptive Thermal Monitor will continue to reduce the package frequency and voltage until the TCC is de-activated.
TjMAX is factory calibrated and is not user configurable. The default value is software visible in the TEMPERATURE_TARGET (0x1A2) MSR, bits [23:16].
The Adaptive Thermal Monitor does not require any additional hardware, software drivers, or interrupt handling routines. It is not intended as a mechanism to maintain processor thermal control to PL1 = Processor Base Power. The system design should provide a thermal solution that can maintain normal operation when PL1 = Processor Base Power within the intended usage range.
Adaptive Thermal Monitor protection is always enabled.
TCC Activation Offset
TCC Activation Offset can be set as an offset from TjMAX to lower the onset of TCC and Adaptive Thermal Monitor. In addition, there is an optional time window (Tau) to manage processor performance at the TCC Activation offset value via an EWMA (Exponential Weighted Moving Average) of temperature. For more information on TCC Activation offset.
TCC Activation Offset with Tau=0
An offset (degrees Celsius) can be written to the TEMPERATURE_TARGET (0x1A2) MSR, bits [29:24], the offset value will be subtracted from the value found in bits [23:16]. When the time window (Tau) is set to zero, there will be no averaging, the offset, will be subtracted from the TjMAX value and used as a new maximum temperature set point for Adaptive Thermal Monitoring. This will have the same behavior as in prior products to have TCC activation and Adaptive Thermal Monitor to occur at this lower target silicon temperature.
If enabled, the offset should be set lower than any other passive protection such as ACPI _PSV trip points
TCC Activation Offset with Tau
To manage the processor with the EWMA (Exponential Weighted Moving Average) of temperature, an offset (degrees Celsius) is written to the TEMPERATURE_TARGET (0x1A2) MSR, bits [29:24], and the time window (Tau) is written to the TEMPERATURE_TARGET (0x1A2) MSR [6:0]. The Offset value will be subtracted from the value found in bits [23:16] and be the temperature.
The processor will manage to this average temperature by adjusting the frequency of the various domains. The instantaneous Tj can briefly exceed the average temperature. The magnitude and duration of the overshoot is managed by the time window value (Tau).
This averaged temperature thermal management mechanism is in addition, and not instead of TjMAX thermal management. That is, whether the TCC activation offset is 0 or not, TCC Activation will occur at TjMAX.
Frequency / Voltage Control
Upon Adaptive Thermal Monitor activation, the processor attempts to dynamically reduce processor temperature by lowering the frequency and voltage operating point. The operating points are automatically calculated by the processor IA core itself and do not require the BIOS to program them as with previous generations of Intel processors. The processor IA core will scale the operating points such that:
- The voltage will be optimized according to the temperature, the processor IA core bus ratio and the number of processor IA cores in deep C-states.
- The processor IA core power and temperature are reduced while minimizing performance degradation.
Once the temperature has dropped below the trigger temperature, the operating frequency and voltage will transition back to the normal system operating point.
Once a target frequency/bus ratio is resolved, the processor IA core will transition to the new target automatically.
- On an upward operating point transition, the voltage transition precedes the frequency transition.
- On a downward transition, the frequency transition precedes the voltage transition.
- The processor continues to execute instructions. However, the processor will halt instruction execution for frequency transitions.
- If the P-state target frequency is higher than the processor IA core optimized target frequency, the P-state transition will be deferred until the thermal event has been completed.
- If the P-state target frequency is lower than the processor IA core optimized target frequency, the processor will transition to the P-state operating point.
If the frequency/voltage changes are unable to end an Adaptive Thermal Monitor event, the Adaptive Thermal Monitor will utilize clock modulation. Clock modulation is done by alternately turning the clocks off and on at a duty cycle (ratio between clock “on” time and total time) specific to the processor. The duty cycle is factory configured to 25% on and 75% off and cannot be modified. The period of the duty cycle is configured to 32 microseconds when the Adaptive Thermal Monitor is active. Cycle times are independent of processor frequency. A small amount of hysteresis has been included to prevent excessive clock modulation when the processor temperature is near its maximum operating temperature. Once the temperature has dropped below the maximum operating temperature, and the hysteresis timer has expired, the Adaptive Thermal Monitor goes inactive and clock modulation ceases. Clock modulation is automatically engaged as part of the Adaptive Thermal Monitor activation when the frequency/voltage targets are at their minimum settings. Processor performance will be decreased when clock modulation is active. Snooping and interrupt processing are performed in the normal manner while the Adaptive Thermal Monitor is active.
Clock modulation will not be activated by the Package average temperature control mechanism.
As the processor approaches TJMax a throttling mechanisms will engage to protect the processor from over-heating and provide control thermal budgets.
Achieving this is done by reducing IA and other subsystem agent's voltages and frequencies in a gradual and coordinated manner that varies depending on the dynamics of the situation. IA frequencies and voltages will be directed down as low as LFM (Lowest Frequency Mode). In relatively rare cases, the processor may take throttle actions on the IO domain, which includes IO fabrics and device throttling, that are designed to avoid shutdown of the system. Further restricts are possible via Thermal Trolling point (TT1) under conditions where thermal budget cannot be re-gained fast enough with voltages and frequencies reduction alone. TT1 keeps the same processor voltage and clock frequencies the same yet skips clock edges to produce effectively slower clocking rates. This will effectively result in observed frequencies below LFM on the Windows PERF monitor.