Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
  • G06F1/26—Power supply means, e.g. regulation thereof
  • G06F1/32—Means for saving power
  • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
  • G06F1/26—Power supply means, e.g. regulation thereof
  • G06F1/32—Means for saving power
  • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
  • G06F1/3234—Power saving characterised by the action undertaken
  • G06F1/324—Power saving characterised by the action undertaken by lowering clock frequency
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
  • G06F1/26—Power supply means, e.g. regulation thereof
  • G06F1/32—Means for saving power
  • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
  • G06F1/3234—Power saving characterised by the action undertaken
  • G06F1/3296—Power saving characterised by the action undertaken by lowering the supply or operating voltage
• • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
  • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
  • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
  • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

• Embodiments of the present writing relate to the manufacture and operation of integrated circuits. More particularly, embodiments of the present writing . relate to compuLer implemented processes for operating an integrated circuit at an optimal voltage for that integrated circuit- Broadly, this writing discloses a method and system of adaptive power control. " Characteristics of a specific integrated circuit are used to adaptively c o ntrol power of the integrated cir c uit.
• a significant problem faced by mobile, e.g., battery powered computers is the length of time such computers are capable of operating between charges.
• An additional concern is heat dissipation of the electronic-..
• An computer processors become more capable, they tend to run at higher clock rates and to dissipate more power.
• more capable peripherals e.g., higher capacity hard drives and wireless high speed networks, also compete with the main computer processor for an ever-increasing portion of the available battery power.
• the size and weight of portable computers is constantly being reduced to make them more portable. Since batteries generally are a very significant component of the weight of portable computers and other portable devices, the tendency has been to maintain battery size and thus battery capacity at a minimum.
• the compact nature of portable devices also intensifies heat dissipation issues.
• processors and computer systems include circuitry and software for disabling various power-draining functions of portable computers when tho3c functions are unused for some extensive period.
• various techniques have been devised for turning off the display screen when it has not been used for some selected period- Similar processes measure the length of time between use of hard drives and disable rotation after some period. Another of these processes is adapted to put a central processor into a quiescent condition after a period of inactivity.
• CMOS complimentary metal oxide semiconductor
• the frequency can be decreased by approximately a factor of two.
• the voltage can also be decreased by a factor of two. Therefore, dynamic power consumption can be reduced by a factor of eight.
• Various methods of implementing this dynamic frequency-voltage scaling have been described in the conventional art.
• Halepete et al. disclose a processor's ability to dynamically adapt its voltage and clock frequency to correspond to the demands placed on the processor by software. Because dynamic power varies linearly with clock speed and by the square of voltage, adjusting both can produce cubic or higher order reductions in dynamic power consumption, whereas prior processors could . adjust power only linearly (by only adjusting the frequency).
• Figure 1 illustrates an exemplary operating frequency versus supply voltage curve 110 for the operation of a microprocessor, according to the conventional art.
• Cn ⁇ v 1 1 indicates the required supply voltage Ndd to achieve a desired frequency of operation. For example, in order to operate at frequency 120, a supply voltage 160 is required.
• Curve 110 may be established as a standard for a population of microprocessors, e.g., by collecting frequency versus voltage information for a number of microprocessor samples. By using a variety of well known statistical analysis methods, curve 110 may be established to achieve an optimal trade-off of performance and process yield. For example, curve 110 may be established such that 90% of all production exhibits better frequency- voltage performance than curve 110. Using curve 110, a set of frequency- voltage operating points may be established. For example, frequency 150 is paired with voltage 190, frequency 140 with voltage 180, frequency 130 with voltage 170 and frequency 120 with voltage 160. Such a set of frequency-voltage pairings (or points) may also be expressed in the form of a table, as illustrated in Table 1, below:
• Such processors arc configured to operate at a number of different frequency and voltage combinations, or points.
• Special power management software monitors the processor and dynamically switches between these operating points as runtime conditions change in order to advantageously minimize dynamic power consumption by the processor.
• a set of frequency-voltage operating points is determined, for example, during qualification testing of a specific processor model for a specific manufacturing process, and used in the operation of every device of that processor model.
• Such a set of frequency-voltage operating points is determined based upon a worst case operation of a population of processor devices, determined, e.g., prior to general availability of the processor devices.
• microprocessor Numerous characteristics related to power consumption of a processor integrated circuit (“microprocessor”) are highly variable across a manufacturing process. For example, maximum operating frequency, threshold voltage and capacitance may each vary by 30% or more, from batch to batch and even within the same wafer. Leakage current is exponential in the threshold voltage and may vary by 500% from nominal. As an unfortunate consequence, such a set of frequency-voltage operating points is based upon the worst case operation of a population of processor devices, determined, e.g., during a qualification process, generally prior to general availability of the processor devices.
• processors typically a majority of a manufactured population, are able to operate at more desirable frequency-voltage operating points than the single standard set based upon worst case performance. For example, if the standard set specifies 600 MHz operation at i.2 volts, many individual processor devices may be able to operate at 600 MHz at only 1.1 volts. Such better performing processors, however, are detrimentally set to operate at the standard frequency-voltage operating point, wasting power. Further, such parts will typically have a lower threshold voltage, and if forced to operate at a higher supply voltage will demonstrate an increased leakage current. Both effects consume excess power and cause an undesirably shorter battery life than is optimal for such a particular processor device.
• maximum operating frequency is generally proportional to the quantity (1 - Nt/Ndd), that is, one minus the threshold voltage divided by the supply voltage (for small process geometries).
• supply voltage (Vdd) typically also is decreased in order to avoid deleterious effects such as oxide breakdown. Consequently, threshold voltage should also be decreased in order to maintain or increase a desirable maximum operating frequency.
• gate oxides are made thinner so that a gate can maintain control of the channel. A thinner gate oxide leads to an increased gate capacitance. Since "off or leakage current of a CMOS device is generally proportional to gate capacitance, the trend to make gate oxidc ⁇ thinner tends to increase leakage current. Unfortunately, increasing leakage current deleteriously increases static power consumption. As an unfortunate result, the on-going decrease in semiconductor process size also leads to an ever-increasing share of total power consumption deriving from static power dissipation.
• static power may vary by as much as 500% above or below nominal levels.
• performance in terms of power consumption is a critical attribute for a mobile processor, processor 0
• a particular processor device If a particular processor device is unable to meet the maximum power limit at the standard frequency-voltage operating point, then it is rejected or placed into a higher power, and therefore less desirable category or "bin" at manufacturing test.
• Such a part represents a yield loss and a loss of potential revenue to the manufacturer. However, such a part may oftentimes be able to achieve both the required clock rate and power consumption specifications at a lower voltage.
• Figure 1 illustrates an exemplary operating frequency versus supply voltage curve for the operation of a microprocessor, according to the conventional art.
• Figure 2 illustrates exemplary frequency- voltage characteristics of specific integrated circuits, according to an embodiment of the present invention.
• FIG. 3 illustrates an integrated circuit module, according to an embodiment of the present invention.
• Figure 4 illustrates a device according to an embodiment of the present invention.
• FIGS 5A, 5B and 5C illustrate configurations of computing elements, according to embodiments of the present invention.
• Figure 6 illustrates a method of manufacturing a microprocessor, according to an embodiment of the present invention.
• Figure 7 illustrates a method of manufacturing a microprocessor, according to an embodiment of the present invention.
• Figure 8 illustrates a method of manufacturing a microprocessor, according to an embodiment of the present invention.
• Figure 9 illustrates a method of operating an integrated circuit, according to an embodiment of the present invention.
• Figure 10 illustrates a method of operating a microprocessor, according to an embodiment of the present invention.
• Figure 11 illustrates a method of operating a microprocessor, according to an embodiment of the present invention.
• Figure 12 illustrates a method of characterizing a microprocessor, according to an embodiment of the present invention.
• Figure 13 illustrates a method of operating a microprocessor, according to an embodiment of the present invention.
• these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated in a computer system. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.
• Embodiments of the present invention are described in the context of design and operation of highly integrated semiconductors. More particularly, embodiments of the present invention relate to adaptive power management of microprocessors. It is appreciated, however, that elements of the present invention may bo utilized in other areas of semiconductor operation.
• a semiconductor manufacturing process is generally considered L ⁇ be highly self-consistent, that is, the process is very good at ro uci "ex-act" copies of an integrated circuit design. This is especially the case for semiconductor products operated in the digital domain. Functionally, the semiconductor arts have succeeded in producing essentially perfect copies that function similarly.
• numerouE analog characteristics of a semiconductor are highly variable.
• threshold voltage, capacitance, gate delay, current consumption, minimum operating voltage and maximum operating frequency may vary 30% or more from chip to chip.
• Leakage, or "off current may be even more variable. For example, it is not uncommon for leakage current to vary 500% above or below a nominal level. More particularly, parameters related to power consumption of an integrated circuit are highly variable.
• Figure 2 illustrates exemplary frequency-voltage characteristics of specific integrated circuits, according to an embodiment of the present invention.
• Curve 210 illustrates measured frequency-voltage characteristics corresponding to a specific integrated circuit out of a manufactured population. Also shown for reference is standard frequency-voltage curve 110 . It is to be appreciated that the frequency-voltage characteristic displayed in curve 210 is different from standard curve 110, reflecting variations in results of a manufacturing process.
• Curve 210 represents a specific integrated circuit with frequeu Y - voltage characteristics that are better than standard curve 110.
• the integrated circuit represented by curve 210 is able to achieve frequency 150 at supply voltage 260, which is less than supply voltage 190, as specified by the standard frequency-voltage characteristic (Table 1).
• Table 1 the standard frequency-voltage characteristic
• such an integrated circuit could provide the same level of processing performance (represented by frequency) at a lower voltage (e.g., voltage 260) and consequently lower power compared to operating at a standard voltage (e.g., voltage 190).
• Curve 210 represents an integrated circuit with a frequency- voltage characteristic that is more desirable than the standard curve (110).
• a majority of a manufacturing population will have superior frequency- ullage characteristics, as a standard curve is determined, at least in part, L ⁇ maximiz the yield of the manufacturing process.
• a deleterious effect of such a standard frequency-voltage characteristic is that most of the integrated circuits are capable of better power performance than they are rated. Since performance in terms of power consumption ie a critical attribute for a mobile processor, processor manufacturers typically test and certify that a part will meet a particular clock rate at a maximum power level.
• “Fast” parts for example an integrated circuit corresponding to curve 210, are able to achieve higher operating frequencies at lower voltages compared to “slower 3* parts. Fast parts typically also have increased current consumption at a given voltage compared to slower parts. Accordingly, while the integrated circuit corresponding to curve 210 is easily able to operate at frequency 150 at supply voltage 190, because it generally will consume more current, it may exceed the power limit at the test voltage. Deleteriously, under the conventional art, an integrated circuit corresponding to curve 210, though able to achieve a desired operating frequency, e.g., frequency 150, at a lower voltage than specified by the standard frequency-voltage characteristic, would be rejected for consuming too much power.
• a desired operating frequency e.g., frequency 150
• Testing of integrated circuits at typical operating frequencies e.g., a few hundred megahertz into the gigahertz range, such as the testing required to generate curve 210, is generally performed after an integrated circuit has been packaged.
• Test fixtures required to test a packaged part are generally less expensive than test fixtures required to test at a bare die or wafer level.
• the packaging including, for example a lead frame, introdnr.es numerous additional electrical and thermal characteristics that may affect the performance, particularly the frequency performance, of a packaged device.
• a packaged semiconductor is found to have undesirable power characteristics after packaging, the expense of packaging is lost.
• certain types of non-volatile elements of a semiconductor e.g., data storage fuses, may typically be set only prior to packaging.
• a data storage fuse may be a semiconductor structure that is directly accessed (e.g., via laser) or via pads that are generally not bonded out. For tho ⁇ c o-nd other reasons it is frequently desirable to determine power performance of an integrated circuit prior to packaging-
• Power consumed b ⁇ *- CMOS circuitry is comprised of two components, a static portion and a dynamic portion.
• C the active switching capacitance
• N the supply voltage
• f the frequency of operation
• static power is expressed as idle current times voltage.
• capacitance, C, and idle current are relatively unaffected by packaging and may be accurately measured at a bare die and/or wafer level.
• dynamic power may be estimated prior to packaging based upon an assumed operating frequency, according to an embodiment of the present invention. For example, based upon a measured value for capacitance, dynamic powei may be estimated for a desired frequency of operation.
• An operating voltage may be specified, for example, per a standard frequency-voltage operating point. Further, since idle current may be measured on an unpackaged device, total power (at a desired operating frequency) may be directly estimated.
• such an estimation of power consumption could be used to test reject parts based upon excessive power consumption, prior to packaging.
• a rejection based upon such criteria may deleteriously reject integrated circuit devices, e.g., the integrated circuit corresponding to curve 210 of Figure 2.
• ⁇ f Lhe present invention it is more desirable to determine a maximum voltage required to achieve a maximum power limit, assuming that a desired operating frequency is achievable. For example, in the power estimation above, specify power (as a maximum limit) and solve for voltage, The resulting voltage is the maximum voltage at which the integrated circuit device may be operated at a desired frequency to achieve an acceptable power level. It in Lo be appreciated that the integrated circuit has not been tested at the desired frequency- However, to meet the power limit, the integrated circuit should operate at the desired frequency at a voltage equal to or less than the determined resulting voltage.
• a representation of such resulting voltage may be encoded, e.g., programmed, into a nonvolatile clement, e.g., a data storage fuse (e.g., a non-rc ⁇ cttablc semiconductor feature), electrically erasable programmable read only memory (EEPROM) or the like, of the integrated circuit.
• a nonvolatile clement e.g., a data storage fuse (e.g., a non-rc ⁇ cttablc semiconductor feature), electrically erasable programmable read only memory (EEPROM) or the like
• the maximum voltage allowed may be encoded into the integrated circuit.
• a mapping into varying sets of frequency-voltage characteristics may be encoded.
• curve 220 represents a dividing line between portions of a manufacturing population, according to an embodiment of the present invention.
• Some integrated circuits may have a voltage frequency characteristic to the "left" of curve 220, e.g., the integrated circuit corresponding to curve 210, and some may have a voltage frequency characteristic between curve 220 and curve 110.
• Curve 220 may become a second set of frequency- voltage operating points.
• curves 110 and 220 With two sets of frcqucncy-volt ⁇ ge operating points, represented by curves 110 and 220.
• integrated circuits could he operated at more optimal power levels based upon specific characteristics of a given, specific integrated circuit. For example, the integrated circuit corresponding to curve 210 could be operated at the frequency-voltage operating points of curve 220, saving power relative to operating the same integrated circuit at the standard frequency-voltage operating points of curve 110.
• An encoded non-volatile circuit element may determine which set of frequency-voltage operating points should be used during operation.
• Table 2 illustrates an exemplary pair of sets of frequency-voltage operating points, according to an embodiment of the present invention.
• Table 2 illustrates two sets of frequency- oltage operating points, a standard set in the two left columns and a set for "fast" parts in the two right columns.
• the "fast" set of frequency-voltage operating points corresponds with curve 220
• an integrated circuit corresponding to curve 210 ay be advantageously operated at frequency-voltage points in the "fast” set of Table 2.
• a "fast" integrated circuit When operated according to such "fast" operating points, a "fast” integrated circuit will consume less power than when operated at a standard set of operating points.
• Embodiments of the present invention are well suited to other methods of representing frequency- oltage characteristics ⁇ f integrated circuits.
• a frequency-voltage characteristic could be represented as coefficients of a polynomial describing, or approximately describing, a frequency-voltage characteristic.
• such an encoded non-volatile circuit element may be used to determine a supply voltage for dynamic power testing of a packaged device.
• fast parts such as the integrated circuit corresponding to curve 210
• a fast part may have failed a maximum power limit test operating at a standard frequency-voltage operating point, e.g., curve 110
• a higher percentage of devices will pass such a test operating at a lower voltage corresponding to curve 220.
• embodiments of the present invention advantageously increase the manufacturing yield.
• Speed or operating frequency for most semiconductors varies with temperature. For example, at a given voltage an integrated circuit may run faster at a lower temperature than at a higher temperature. As a corollary, an integrated circuit may typically be operated at a desired frequency with a lower supply voltage at a lower temperature when a higher supply voltage is required to operate at the same desired frequency at a higher temperature.
• chip temperature may be used to select an optimal voltage of operation for a desired operating frequency.
• Table 3, below, illustrates exemplary sets of frequency- oltage operating points incorporating chip temperature, according to an embodiment of the present invention.
• Table 3 illustrates four sets of frequency -voltage operating points, two standard sets in the second and third columns and two sets for "fast" parts in the two right columns. There are two standard sets and two fast sets, based upon chip temperature, e.g., "hot” and “cool.” If chip temperature is known
• an integrated circuit may be advantageously configured to be less than or equal to 50° C, an integrated circuit may be advantageously configured
• a first limitation relates to the amount of information that may actually be encoded into an integrated circuit.
• Microprocessors typically have extremely limited amounts of or no non-volatile storage. The reasons for a paucity of such on-chip non-volatile storage include circuit size, e.g., a bit of non-volatile storage consumes valuable circuit area, more optimally used for microprocessor circuitry.
• Another reason relates to semiconductor processes, e.g., many types of integrated circuit non-volatile storage require additional semiconductor masks and process steps not required for standard microprocessor circuitry. Such masks and process steps are expensive and add complexity to a manufacturing process, thereby increasing costs and reducing yield. For these reasons and others, non- volatile storage requiring additional process steps is typically not included in ⁇ microprocessor.
• a second limitation involves the capacity and capabilities of die/wafer level testing. Die and wafer level testers are expensive, more so than packaged testers, and as previously described typically can not fully characterize the frequency-voltage behavior of an integrated circuit because of the electrical effects of packaging. Consequently, it is sometimes more ' desirable to characterize the frequency-voltage behavior of an integrated circuit after packaging.
• embodiments of the present invention are well suited to encoding a mapping into varying sets of frequency-voltage chai cteris ics L ⁇ l ⁇ an integrated circuit after packaging-
• on- chip non-volatile storage may be aeftes ⁇ ible via package pins, and a tester could write mapping information into such storage.
• on-chip non-volatile storage may not be encoded after packaging. At other times there may not be sufficient on-chip nonvolatile storage available for encoding the desired amount of information.
• alternative means of associating one or more frequency-voltage characteristics with a specific integrated circuit should be employed.
• a frequency- voltage characteristic, or a mapping of a frequency-voltage characteristic, of a specific integrated circuit may be encoded in a non-volatile memory device that is associated with the specific integrated circuit.
• Example devices include without limitation electrically erasable programmable read only memory (EEPROM). battery-backed up random access memory (RAM), mask ROM, ferro-magnetic RAM, data storage fuses within semiconductor packaging, data storage fuses that are a part of an integrated circuit, wiring connections (e.g., encoding by coupling pins to Vcc and/or ground), and the like.
• One desirable method of associating a non-volatile memory device with specific integrated circuit is to include a memory device in the packaging of the specific integrated circuit, for example in a multi-chip module.
• Embodiments of the present invention are well suited to other methods of associating a frequency-voltage characteristic, or a mapping of a frequeney- voltage characteristic, of a specific integrated circuit with the specific integrated circuit.
• One method maintains a database in computer readable media of a plurality of frequency-voltage characteristics for a plurality of integrated circuits.
• a specific frequency-voltage characteristic of a specific integrated circuit may then be referenced by an identifying attribute of the integrated circuit, e.g., a unique serial number.
• the specific frequency-voltage characteristic may then be transferred by a wide variety of means, e.g., internet download, and referenced at many points during the manufacture and/or use of products using the integrated circuit.
• the integrated circuit was a microprocessor
• the manufacturer of a computer comprising tlie microprocessor could assemble the computer, electronically access the. microprocessor serial number, access the specific frequency-voltage characteristic, e.g., in a data file, and configure a unique ROM for that computer.
• Such a ROM would then reflect the specific frequency-voltage characteristic of the microprocessor at the heart of that particular computer.
• Encoding information after a packaging operation typically makes a greater amount of storage available that pre-packaging encoding.
• a memory semiconductor typically comprises millions of bits of information as opposed to a few bits which might be included in a microprocessor.
• numerous types of well known computer systems that may be coupled to a microprocessor, e.g., testing machines, may have, or be further coupled to essentially unlimited storage.
• an integrated circuit may be fully characterized after packaging, it is typically possible to encode greater amounts of information specific to a particular integrated circuit after packaging. According to an embodiment of the present invention, information of specific integrated circuit frequency-voltage characteristics may be encoded after packaging the integrated circuit.
• Such post packaging encoding may store a variety of types of information.
• an integrated circuit may be characterized as belonging to one of a plurality of groups of integrated circuits which have similar frequency-voltage characteristics.
• Information encoded may identify which of such groups best describes a specific integrated circuit.
• a frequency-voltage characteristic of a specific integrated circuit may be encoded.
• attributes of curve 210 Figure 2 may be encoded.
• a single set of frequency-voltage characteristics may be utilized by power management software operating on a specific integrated circuit.
• power management software may be simplified and operate more efficiently with reduced indirection in referencing a single set of operating points.
• Embodiments of the present invention are well suited to a wide variety of well known methods of representing curves, including data reduction techniques, such as a frequency-voltage curve of an integrated circuit.
• tables containing coordinate points are used to represent frequency-voltage characteristics .
• a frequency-voltage characteristic may be represented as coefficients of a polynomial describing, or approximately describing, a frequency-voltage characteristic.
• Table 4 illustrates a frequency-voltage characteristic of a specific microprocessor, according to an embodiment of the present invention.
• FIG. 3 illustrates an integrated circuit module 300, according to an embodiment of the present invention.
• Module 300 comprises a microprocessor 310 and a non-volatile storage 320. Microprocessor 310 and storage 320 are coupled by coupling 330.
• Module 300 generally conforms to a wide variety of types of integrated circuit packaging, e.g., a package designed for direct or indirect mounting on a printed circuit board. Example packages include, without limitation ball grid arrays, pin grid arrays, thin quad flat packs, leadless chip carriers and the like.
• Module 300 may be a multi-chip module, capable of containing a plurality of integrated circuits.
• the processor 310 and storage device 320 are separate integrated circuits.
• Embodiments of the present invention are also well suited to modules comprising a microprocessor which is directly attached to a circuit board, e.g., via direct chip attach (DCA).
• Integrated circuits which are DCA- mounted typically have an encapsulant or "glop top" applied after mounting, although this is not required by embodiments of the present invention. No requirement of physical proximity of components of a module is intended by the foregoing discussion.
• Non-volatile storage 320 is well suited to a wide variety of types of nonvolatile storage.
• storage 320 may be an integrated circuit device, including without limitation electrically erasable programmable read only memory (EBPROM), battery-backed up random access memory (RAM), mask ROM, and ferro-magnetic RAM.
• EBPROM electrically erasable programmable read only memory
• RAM battery-backed up random access memory
• RAM mask ROM
• Ferro-magnetic RAM ferro-magnetic RAM.
• Storage 320 is also well suited to a wide variety of other types of data storage, including data storage fuses within semiconductor packaging, data storage fuses that are a part of an integrated circuit, wiring connections (e.g., encoding by coupling pins to Vcc and/or ground), and the like.
• Figure 4 illustrates a device 400, according to an embodiment of the •present invention.
• Device.400 may form a computer, or portions thereof, e.g., a computer motherboard.
• Device 400 comprises a microprocessor 410 and a memory space 450.
• Microprocessor 410 is capable of operating at a plurality of voltages and further capable of operating at a plurality of frequencies.
• Memory space 450 is coupled to microprocessor 410.
• Memory space 450 is well suited to a wide variety of well known types of memory, including without limitation internal RAM or ROM within icroprocessor 410, RAM or ROM outside of microprocessor 410 and other forms of computer readable media.
• Voltage supply 420 is coupled to microprocessor 410.
• Voltage supply 450 furnishes a selectable voltage 430 to microprocessor 410.
• Microprocessor 410 is coupled to voltage supply 420 via coupling 440, in order to cause voltage supply 420 to select an optimal voltage.
• Memory space 460 is coupled to microprocessor 410, for example via the same coupling means as memory space 450.
• Memory space 460 is well suited to similar types of memory as previously described for memory 450, and is further well suited to being within the same physical device as memory space 450.
• memory space 460 is not required to be in the same physical device as memory space 450, and some o i ⁇ ii -cations of computer 400 may preferentially separate the two memories by type and/or device.
• device 400 may comprise logic 470- Logic 470 is coupled to memory 460 and microprocessor 410.
• Logic 470 accesses a desired frequency of operation, which may be determined by microprocessor 410, and uses memory 460 to determine a corresponding voltage. Logic 470 causes microprocessor 410 to operate at the desired frequency of operation and at the corresponding voltage. Logic 470 is well suited to software embodiments executing, for example, on microprocessor 410. Logic 470 may also execute on a second processor (not shown), or comprise a functional set of gates and/or latches. Logic 470 is well suited to being internal or external to microprocessor 410,
• Figures 5A, 5B and 5C illustrate configurations of computing elements, according to embodiments of the present invention.
• Figure 5A illustrates configuration 570 of computing elements, according to an embodiment of the present invention.
• Configuration 570 is well suited to a wide variety of placing, mounting and interconnection arrangements-
• configuration 570 may be a portion of a computer "mother board.' 7
• configuration 570 may be a multi-chip module.
• Data structure 505 comprises voltage-frequency relationship information specific to processor 501.
• information of data structure 505 is based upon testing of processor 501, aa opposed to being information generally about a population of processors of similar design.
• Data structure 505 may reside in memory 506, which is external to processor 501.
• Memory 506 may be, for example, a ROM integrated circuit.
• Embodiments of the present invention are well suite to other types of memory as well, e.g., RAM, rotating magnetic storage, etc.
• Control logic 507 uses information from data structure 505 to control the voltage and or frequency of operation for processor 501. Such control may be to minimize power consumption of processor 501. Control logic 507 is well suited to software executing on processor 501.
• Figure 5B illustrates configuration 580 of computing elements- according to an embodiment of the present invention.
• Configuration 580 is well suited to a wide variety of placing, mounting and interconnection arrangements.
• configuration 580 may be a portion of a computer "mother board.”
• configuration 580 may be a multi- chip module.
• Data structure 530 comprises voltage-frequency relationship information for a class of processors of the type of processor 503.
• data structures 540 and 550 comprises voltage-frequency relationship information respectively for other classes of processors of the type of processor 503.
• information of data structures 530 — 550 represents, segments of a manufacturing population of processors of the same type as processor 503.
• information 530, 540 and 550 will differ.
• Embodiments of the present invention are well suited tn varying numbers of such data structures.
• Data structures 530, 540 and 550 may reside in memory 560, which is external to processor 503.
• Memory 560 may be, for example, a ROM integrated circuit.
• Embodiments of the present invention are well suite to other types of memory as well, e.g., RAM, rotating magnetic storage, etc.
• Class identifier 520 comprises information as to which voltage- frequency relationship (e.g., contained in data structure 530, 540 or 550) best corresponds to processor 503. Class identifier 520 is determined and recorded during production of processor 503. Such determination and recording of a class identifier may be performed prior to packaging of processor 503. Class identifier 520 is typically a few bits, and is well suited to many types of non- volatile storage, e.g., data storage fuses or electrically erasable programmable read only memory (EEPROM).
• EEPROM electrically erasable programmable read only memory
• Control logic 525 accesses information from class identifier 520 to determine which data structure, e.g., 530, 540 or 560, corresponds to processor 503 and should be used to control the voltage and/or frequency of operation for processor 503. S ⁇ h control may be to minimize power consumption of processor 503. Control logic 525 is well suited to software executing on processor 503.
• FIG. 5C illustrates configuration 590 of computing elements, according to an embodiment of the present invention.
• Configuration 590 is well suited to a wide variety of placing, mounting and interconnection arrangements.
• configuration 590 may be a portion of a computer "mother board.”
• configuration 590 may be a multi- chip module.
• Data structure 505 comprises voltage-frequency relationship information specific to processor 502. In contrast to the conventional art, information of data structure 505 is based upon testing of processor 502 * , as opposed to being information generally about a population of processors of similar design.
• Data structure 505 is internal to processor 502.
• Data structure 505 may be stored in a wide variety of well known memory types are may be included in .
• a processor e.g., non-volatile RAM, electrically erasable programmable read only memory (EEPROM), etc,
• Control logic 508 uses information from data structure 505 to control the voltage and/or frequency of operation for processor 502. Such control may be to minimize power consumption of processor 502. Control logic 508 is well suited to software executing on processor 502.
• Figure 6 illustrates a method 600 of manufacturing a microprocessor, according to an embodiment of the present invention.
• step 610 a voltage required by the microprocessor to meet a power limit at a specified frequency is determined.
• the voltage may be determined by operating the microprocessor at the specified frequency and measuring its power consumption, for example.
• step 620 information of the voltage is stored into computer readable media.
• Embodiments of the present invention are well suited to storing the information in a wide variety of types of computer readable media, including memory or storage systems of an integrated circuit tester or manufacturing data logging system-., integrated circuits within co ⁇ im ⁇ ii packaging of the microprocessor, separate integrated circuit memories and the like.
• the information comprises frequency voltage relations specific to the microprocessor.
• FIG. 7 illustrates a method 700 of manufacturing a microprocessor, according to an embodiment of the present invention.
• step 710 an optimal voltage required by the microprocessor to operate at each of a plurality of specified frequencies is determined, producing a plurality of voltage- frequency relations.
• the voltage frequency relations may be similar to Table 4, above.
• step 720 information of the plurality of relations is stored into computer readable media.
• Embodiments of the present invention are well E uited to storing the information in a wide variety of types of computer readable media, including memory or storage systems of an integrated circuit tester or manufacturing data logging systems, integrated circuits within common packaging of the microprocessor, separate integrated circuit memories and the like.
• FIG. 8 illustrates a method 800 of manufacturing a microprocessor, according to an embodiment of the present invention.
• step 810 a plurality of unpackaged microprocessors is accessed-
• semiconductor packaging e.g., a pin grid array package, rather than shipping or consumer packaging.
• the microprocessors may be in wafer form.
• a voltage required by one of the plurality of unpackaged microprocessors to meet a power limit at a specified frequency is determined.
• the voltage may be determined by measuring capacitance and/or idle current of the microprocessor.
• step SSO information of said voltage is encoded into the microprocessor.
• Embodiments of the present invention are well suited to a wide variety of well known types of storage that may be a part of a microprocessor and encoded in step 830. Examples include, without limitation, data storage fuses and electrically erasable read only memory and the like.
• Figure 9 illustrates a method 900 of operating an integrated circuit, according to an embodiment of the present invention.
• a frequency-voltage characteristic of an integrated circuit is measured and recorded.
• the integrated circuit may be a microprocessor, and the integrated circuit may be packaged or unpackaged.
• the frequency-voltage characteristic is associated with the integrated circuit for optimizing the operation of the integrated circuit.
• the frequency-voltage characteristic may be associated with the integrated circuit in a wide variety of ways, including via encoding information in. the integrated circuit or encoding information in a device that will be delivered to a customer with the integrated circuit.
• associating may take the form of maintaining a database in computer readable media of a plurality of frequency-voltage characteristics for a plurality of integrated circuits.
• a specific frequency-voltage characteristic of a specific integrated circuit may then be referenced by an identifying attribute of the integrated circuit, e.g., a unique serial number.
• the integrated circuit is operated at a voltage and frequency specified by the frequency-voltage characteristic.
• FIG. 10 illustrates a method 1000 of operating a microprocessor, according to an embodiment of the present invention.
• a desirable operating frequency is determined for the microprocessor.
• One typical means of determining a desirable frequency is based upon the processing needs of software executing on the microprocessor.
• step 1020 an optimal voltage for operating the microprocessor at the desirable operating frequency is selected.
• the optimal voltage selected is based upon characteristics that are specific for the microprocessor.
• the microprocessor is operated at the optimal voltage.
• the microprocessor may cause a selectable voltage supply, e.g., voltage supply 420 of Figure 4, to output the optimal voltage.
• the microprocessor may be operated at the desirable frequency lwar operating at the optimal voltage.
• Figure 11 illustrates a method 1100 of operating a microprocessor, ' according to an embodiment of the present invention.
• step 1120 a set frequency at which the microprocessor is to operate is accessed.
• a corresponding set voltage at which the microprocessor optimally operates at the set frequency is determined using the set frequency. The determining is based on a voltage -frequency relationship that is specific to the microprocessor.
• step 1140 the microprocessor is operated at the set frequency and the corresponding set voltage.
• the set frequency is determined by examining a contemporaneous set of operations performed by the microprocessor.
• Figure 12 illustrates a method 1200 of characterizing a microprocessor, according to an embodiment of the present invention.
• a frequency-voltage relationship specific to a microprocessor is measured and recorded.
• the frequency-voltage relationship may record frequency-voltage pairs for optimal operation of the microprocessor.
• Embodiments of the present invention are well suited to other well known methods of recording a relationship, including various well known data reduction methods.
• the frequency-voltage relationship is associated with the microprocessor so that said frequency-voltage relationship can be accessed and used during operation of the microprocessor for optimal use thereof.'
• Figure 13 illustrates a method 1300 of operating a microprocessor, according to an embodiment of the present invention.
• a desirable operating frequency is determined for the microprocessor.
• Embodiments of the present invention are well suited to a variety of optimizations to determine a desirable frequency.
• One typical means of determining desirable frequency is based upon the processing needs of software executing on the microprocessor.
• step 1320 information specific to the microprocessor is accessed.
• tlie iuformation may be stored in memory, e.g., memory 450 of Figure 4.
• the information is used to determine an optimal voltage for operating the microprocessor at the desirable frequency.
• the optimal voltage is based upon characteristics of the microprocessor.
• the information may be accessed from within the microprocessor.
• the microprocessor is operated at the optimal voltage.
• the microprocessor may cause a selectable voltage supply, e.g., voltage supply 420 of Figure 4, to output the optimal voltage.
• the microprocessor may be operated at the desirable frequency while operating at the optimal voltage. Steps 131 through 1330 may be repeated, if desired. For example, if a processing load changes, a new desirable operating frequency may be determined, a corresponding voltage may be accessed, and the microprocessor may be operated at the new frequency and voltage.
• a system and method of adaptive power control of integrated circuits are disclosed.
• a desirable operating frequency is determined for the integrated circuit.
• the integrated circuit may b p a microprocessor.
• the desirable operating frequency may be determined based upon a determined processing load desired of the microprocessor.
• An optimal voltage for operating the microprocessor at the desired operating frequency is selected. The selection is based upon characteristics that are specific to the microprocessor.
• the micr ⁇ pr ⁇ ue-.s ⁇ r is operated at the optimal voltage.
• Embodiments of the present invention provide a means to adaptively control power consumption of an integrated circuit based upon specific characteristics of the integrated circuit. Further embodiments of the present invention provide for the above mentioned solution to be achieved with existing semiconductor processes and equipment without revamping well established tools and techniques.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Semiconductor Integrated Circuits (AREA)
• Testing Or Measuring Of Semiconductors Or The Like (AREA)
• Power Sources (AREA)
• Microcomputers (AREA)
• Feedback Control In General (AREA)

Priority Applications (3)

Applications Claiming Priority (2)

Publications (3)

Family

ID=32655204

Family Applications (1)

Country Status (5)

Cited By (4)

Families Citing this family (25)

Family Cites Families (215)

• 2002
  • 2002-12-31 US US10/334,918 patent/US7941675B2/en not_active Expired - Lifetime
• 2003
  • 2003-12-29 AU AU2003300016A patent/AU2003300016A1/en not_active Abandoned
  • 2003-12-29 JP JP2004565786A patent/JP2006515448A/ja active Pending
  • 2003-12-29 WO PCT/US2003/041489 patent/WO2004061635A2/en not_active Ceased
  • 2003-12-29 CN CN2003801096001A patent/CN1745358B/zh not_active Expired - Lifetime
• 2011
  • 2011-05-09 US US13/103,952 patent/US20110219245A1/en not_active Abandoned

Also Published As

Similar Documents

Legal Events