Systems and methods for detecting thermal runaway
In one embodiment, a method of temperature control comprises receiving temperature readings from a temperature sensor on a chip, calculating one or more second derivatives of temperature with respect to time based on the temperature readings, and determining whether to perform temperature mitigation on the chip based on the one or more calculated second derivatives of temperature.
1. A temperature management system, comprising:
a low-pass filter configured to low-pass filter temperature readings from a temperature sensor on a chip to obtain filtered temperature readings;
a calculation module configured to calculate second derivatives of temperature with respect to time based on the filtered temperature readings; and
a temperature control module configured to detect a start of thermal runaway on the chip by detecting a transition of the calculated second derivatives of temperature from a negative value to a positive value and to determine to perform temperature mitigation on the chip if the start of thermal runaway is detected,
wherein, if a determination is made to perform temperature mitigation, then the temperature control module is configured to reduce at least one of an adjustable clock frequency of a clock signal provided to a circuit on the chip and an adjustable supply voltage provided to the circuit on the chip.
2. The system of claim 1 , wherein the temperature sensor is configured to generate a voltage or a current that is sensitive to a temperature of the chip, and to generate the temperature readings based on the voltage or the current.
3. The system of claim 1 , wherein the temperature control module is configured to receive dynamic power estimates for the circuit on the chip from a power meter, to determine whether there is an increase in a dynamic power of the circuit based on the dynamic power estimates, and to determine whether to perform temperature mitigation based also on the determination of whether there is an increase in the dynamic power of the circuit.
4. The system of claim 3 , wherein the temperature control module is configured to determine an increase in the dynamic power of the circuit if a current one of the received dynamic power estimates is greater than a previous one of the received dynamic power estimates.
5. A method of temperature control, comprising:
receiving temperature readings from a temperature sensor on a chip;
low-pass filtering the temperature readings to obtained filtered temperature readings;
calculating second derivatives of temperature with respect to time based on the filtered temperature readings;
detecting a start of thermal runaway on the chip by detecting a transition of the calculated second derivatives of temperature from a negative value to a positive value;
determining to perform temperature mitigation on the chip if the start of thermal runaway is detected; and
if a determination is made to perform temperature mitigation, then reducing at least one of an adjustable clock frequency of a clock signal provided to a circuit on the chip and an adjustable supply voltage provided to the circuit on the chip.
6. The method of claim 5 , further comprising:
receiving dynamic power estimates for the circuit on the chip; and
determining whether there is an increase in a dynamic power of the circuit based on the dynamic power estimates;
wherein determining whether to perform temperature mitigation is based also on the determination of whether there is an increase in the dynamic power of the circuit.
7. The method of claim 6 , wherein determining whether there is an increase in the dynamic power of the circuit comprises determining there is an increase in the dynamic power of the circuit if a current one of the received dynamic power estimates is greater than a previous one of the received dynamic power estimates.
8. The system of claim 1 , wherein:
the system comprises a clock generator configured to output the clock signal with the adjustable clock frequency to the circuit on the chip; and
the temperature control module is configured to instruct the clock generator to reduce the clock frequency if the determination is made to perform temperature mitigation.
9. The system of claim 1 , wherein:
the system comprises a power source configured to provide the adjustable supply voltage to the circuit on the chip; and
the temperature control module is configured to instruct the power source to reduce the supply voltage if the determination is made to perform temperature mitigation.
10. The system of claim 1 , wherein if the determination is made to perform temperature mitigation, then the temperature control module is configured to reduce both the clock frequency and the supply voltage.
11. The system of claim 1 , wherein:
the system comprises a clock generator configured to output the clock signal with the adjustable clock frequency to the circuit on the chip;
the system comprises a power source configured to provide the adjustable supply voltage to the circuit on the chip; and
if the determination is made to perform temperature mitigation, then the temperature control module is configured to:
instruct the clock generator to reduce the clock frequency; and
instruct the power source to reduce the supply voltage.
12. The method of claim 5 , wherein reducing at least one of the adjustable clock frequency of the clock signal provided to the circuit on the chip and the adjustable supply voltage provided to the circuit on the chip comprises reducing both the clock frequency and the supply voltage.
13. A temperature management system, comprising:
a temperature sensor configured to generate temperature readings on a chip; and
a temperature manager configured to receive the temperature readings from the temperature sensor and dynamic power estimates for a circuit on the chip, to determine whether there is an increase in a dynamic power of the circuit based on the dynamic power estimates, to detect a start of thermal runaway on the chip based on the temperature readings and the determination of whether there is an increase in the dynamic power of the circuit, and, if the temperature manager detects the start of thermal runaway, to reduce at least one of an adjustable clock frequency of a clock signal provided to a circuit on the chip and an adjustable supply voltage provided to the circuit on the chip.
14. The system of claim 13 , wherein the temperature manager is configured to detect the start of thermal runaway by determining a second derivative of temperature with respect to time based on the temperature readings, comparing the determined second derivative to a threshold, and detecting the start of thermal runaway based on the comparison.
15. The system of claim 13 , wherein:
the system comprises a clock generator configured to output the clock signal with the adjustable clock frequency to the circuit on the chip; and
the temperature manager is configured to instruct the clock generator to reduce the clock frequency if the determination is made to perform temperature mitigation.
16. The system of claim 13 , wherein:
the system comprises a power source configured to provide the adjustable supply voltage to the circuit on the chip; and
the temperature manager is configured to instruct the power source to reduce the supply voltage if the determination is made to perform temperature mitigation.
17. The system of claim 13 , wherein if the determination is made to perform temperature mitigation, then the temperature manager is configured to reduce both the clock signal and the supply voltage.
18. The system of claim 13 , wherein:
the system comprises a clock generator configured to output the clock signal with the adjustable clock frequency to the circuit on the chip;
the system comprises a power source configured to provide the adjustable supply voltage to the circuit on the chip; and
if the determination is made to perform temperature mitigation, then the temperature manager is configured to:
instruct the clock generator to reduce the clock frequency; and
instruct the power source to reduce the supply voltage.
19. The system of claim 13 , wherein the temperature manager is configured to determine an increase in the dynamic power of the circuit if a current one of the dynamic power estimates is greater than a previous one of the dynamic power estimates.
20. The system of claim 14 , wherein the temperature manager is configured to detect the start of thermal runaway if the second derivative of temperature is greater than the threshold and the temperature manager does not determine an increase in the dynamic power of the circuit.