Techniques for distributing power in electronic circuits and computer systems

• US 7,698,114 B2
• Filed: 11/02/2005
• Issued: 04/13/2010
• Est. Priority Date: 11/02/2005
• Status: Expired due to Fees

• Current Assignee: GlobalFoundries, Inc.
• Original Assignee: International Business Machines Corporation
• Inventors: Wakil, Jamil, Hamann, Hendrik F., Weger, Alan J., Lacey, James A.
• Primary Examiner(s): Shah; Kamini S
• Assistant Examiner(s): Saxena; Akash
• Application Number:
  US11/264,968
• Publication Number:
  US 20070098037A1
• Time in Patent Office: 1,623 Days
• Field of Search: 374/137, 374/130, 703/5, 703/9, 702/130, 702/136
• US Class Current: 703/5
• CPC Class Codes: G01K 2215/00 Details concerning sensor power supply
• CPC Class Codes: G01K 7/42 Circuits effecting compensation of thermal inertia; Circuits for predicting the stationary value of a temperature
• CPC Class Codes: G01K 7/425 Thermal management of integrated systems

Abstract

Techniques for enhancing thermal design of a system having a number of boundary values are provided. A method for such enhancement includes representing thermal response of the system to the boundary values, obtaining at least one constraining parameter, and determining spatial and/or temporal distribution of the boundary values. The thermal response is represented as a superposition of temperature fields associated with given boundary values. The spatial and/or temporal distribution of the boundary values is determined based on the thermal response represented in the representing step, so as to satisfy the constraining parameter. The boundary values can be, for example, power sources, and the at least one constraining parameter can be, for example, a spatial or temporal location of one of the power sources.

Claims

1. A computer-implemented method for enhancing thermal design of a physical system having a plurality of boundary values, said method comprising the steps of:

determining an initial spatial power distribution P_t=0 at an initial time t=0;

incrementing said time t by a time step t_step less than a characteristic thermal time constant of said physical system to obtain an incremented time value;

representing thermal response of said physical system to said plurality of boundary values at said incremented time value as a superposition of temperature fields associated with given ones of said boundary values, based on a previous spatial power distribution at a time equal to said incremented time value less said time step t_step, said previous spatial power distribution comprising said initial spatial power distribution for a first increment;

obtaining at least one constraining parameter for said incremented time value;

calculating a subsequent spatial power distribution P_t at said incremented time value, wherein said subsequent spatial power distribution P_t both satisfies said at least one constraining parameter for said incremented time value and provides a minimum temperature distribution;

repeating said incrementing step for a plurality of repetitions to obtain a plurality of subsequent incremented time values;

wherein;

at each of said repetitions, said steps of representing thermal response, obtaining at least one constraining parameter, and calculating a subsequent spatial power distribution are repeated to obtain a transient analysis, said transient analysis comprising a plurality of said subsequent spatial power distributions, calculated at each of said plurality of subsequent incremented time values, in which at least some of said subsequent spatial power distributions are time-varying; and

at least one of said boundary values comprises a time-varying power source;

further comprising applying said time-varying power source to said physical system in accordance with said time-varying subsequent spatial power distributions.

2. The method of claim 1, wherein said time step t_step less than said characteristic thermal time constant of said physical system is no greater than one fifth of said characteristic thermal time constant of said physical system.

3. The method of claim 2, wherein said representing, obtaining and determining steps are performed in a substantially open-loop manner, substantially without using sensor feedback pertaining to power distribution and substantially without using sensor feedback pertaining to temperature distribution.

4. The method of claim 2, wherein at least another one of said boundary values comprises a fixed temperature region.

5. The method of claim 2, wherein:

said physical system is an integrated circuit chip; and

said time-varying power source represents an electrical device to be located on said chip.

6. The method of claim 2, wherein:

said physical system is a data center; and

said time-varying power source represents a data-processing component within said data center.

7. The method of claim 2, wherein said representing step is performed via predictive thermal modeling with a thermal model of said physical system.

8. The method of claim 2, wherein:

at least some additional ones of said boundary values comprise power sources;

said representing step comprises generating M²×

M² matrices, A1 and A2 defining geometric and physical makeup of said physical system; and

said calculating step comprises calculating said subsequent spatial power distribution by solving;

T(n)=A1P(n)+A2T(n−

1)

where;

P(n) is a power vector of length M containing magnitudes of each individual one of said power sources at time t=n T_s; and

T(n), T(n−

1) are temperature vectors of length M containing magnitudes of each temperature at time t=n T_s and t=(n−

1) T_s.

9. The method of claim 8, wherein said representing step is performed via predictive thermal modeling with a thermal model of said system and further comprises:

determining an appropriate discretization granularity for said thermal model of said physical system; and

creating said thermal model according to said discretization granularity.

10. The method of claim 2, wherein:

at least some additional ones of said boundary values comprise power sources;

said representing step comprises generating M²×

M² matrices A1, A2, and A3, defining geometric and physical makeup of said physical system; and

said calculating step comprises calculating said subsequent spatial power distribution of said power sources by solving;

T(n)=A1P(n)+A2T(n−

1)+A3T(n−

2)

where;

P(n) is a power vector of length M containing magnitudes of each individual one of said power sources at time t=n T_s; and

T(n), T(n−

1), T(n−

2) are temperature vectors of length M containing magnitudes of each temperature at time t=n T_s, t=(n−

1) T_s, and t=(n−

2) T_s.

11. The method of claim 2, wherein:

at least some additional ones of said boundary values comprise power sources;

said representing step comprises generating M²×

M² matrices, A1, A2, A3, and A4, defining geometric and physical makeup of said physical system, A1, A2 being based on a time constant related to initial temperature rise based on heat capacity, A3, A4 being based on a time constant related to heat removal rate alone; and

said calculating step comprises calculating said subsequent spatial power distribution by solving;

T₁(n)=A1P(n)+A2T₁(n−

1)

T₂(n)=A3P(n)+A4T₂(n−

1)

T(n)=T₁(n)+T₂(n)

where;

P(n) is a power vector of length M containing magnitudes of each individual one of said power sources at time t=n T_s; and

T_i(n), T_i(n−

1), are temperature vectors of length M containing magnitudes of each temperature at time t=n T_s, t=(n−

1) T_s.

12. A computer program product comprising a tangible computer readable storage medium having computer useable program code for enhancing thermal design of a physical system having a plurality of boundary values, said computer program product including:

computer useable program code for determining an initial spatial power distribution P_t=0 at an initial time t=0;

computer useable program code for incrementing said time t by a time step t_step less than a characteristic thermal time constant of said physical system to obtain an incremented time value;

computer useable program code for representing thermal response of said physical system to said plurality of boundary values at said incremented time value as a superposition of temperature fields associated with given ones of said boundary values, based on a previous spatial power distribution at a time equal to said incremented time value less said time step t_step, said previous spatial power distribution comprising said initial spatial power distribution for a first increment;

computer useable program code for obtaining at least one constraining parameter for said incremented time value;

computer usable program code for calculating a subsequent spatial power distribution P_t at said incremented time value, wherein said subsequent spatial power distribution P_t both satisfies said at least one constraining parameter for said incremented time value and provides a minimum temperature distribution; and

computer useable program code for repeating said incrementing step for a plurality of repetitions to obtain a plurality of subsequent incremented time values;

wherein;

at each of said repetitions, said steps of representing thermal response, obtaining at least one constraining parameter, and calculating a subsequent spatial power distribution are repeated to obtain a transient analysis, said transient analysis comprising a plurality of said subsequent spatial power distributions, calculated at each of said plurality of subsequent incremented time values, in which at least some of said subsequent spatial power distributions are time-varying; and

at least one of said boundary values comprises a time-varying power source;

further comprising computer usable program code to facilitate applying said time-varying power source to said physical system in accordance with said time-varying subsequent spatial power distributions.

13. The computer program product of claim 12, wherein said computer useable program code for incrementing said time t by said time step t_step less than said characteristic thermal time constant of said physical system is configured to set said time step t_step to a value no greater than one fifth of said characteristic thermal time constant of said physical system.

14. An apparatus for enhancing thermal design of a physical system having a plurality of boundary values, comprising:

a memory; and

at least one processor coupled to said memory and operative to;

determine an initial spatial power distribution P_t=0 at an initial time t=0;

increment said time t by a time step t_step less than a characteristic thermal time constant of said physical system to obtain an incremented time value;

represent thermal response of said physical system to said plurality of boundary values at said incremented time value as a superposition of temperature fields associated with given ones of said boundary values, based on a previous spatial power distribution at a time equal to said incremented time value less said time step t_step, said previous spatial power distribution comprising said initial spatial power distribution for a first increment;

obtain at least one constraining parameter for said incremented time value;

calculate a subsequent spatial power distribution P_t at said incremented time value, wherein said subsequent spatial power distribution P_t both satisfies said at least one constraining parameter for said incremented time value and provides a minimum temperature distribution;

repeat said incrementing step for a plurality of repetitions to obtain a plurality of subsequent incremented time values;

wherein;

at each of said repetitions, said steps of representing thermal response, obtaining at least one constraining parameter, and calculating a subsequent spatial power distribution are repeated to obtain a transient analysis, said transient analysis comprising a plurality of said subsequent spatial power distributions, calculated at each of said plurality of subsequent incremented time values, in which at least some of said subsequent spatial power distributions are time-varying; and

at least one of said boundary values comprises a time-varying power source;

said at least one processor is further operative to facilitate applying said time-varying power source to said physical system in accordance with said time-varying subsequent spatial power distributions.

15. The apparatus of claim 14, wherein said at least one processor is operative to set said time step t_step to a value no greater than one fifth of said characteristic thermal time constant of said physical system.