Chip temperature characteristic test method, system and equipment

Abstract

The invention relates to the technical field of chip testing, and provides a chip temperature characteristic testing method, system and equipment, and the method comprises the steps: obtaining transient thermal response data and steady-state thermal distribution data of a chip, carrying out the feature extraction of the transient thermal response data, obtaining a feature thermal mode set, and carrying out the feature extraction of the feature thermal mode set; performing heat transfer analysis on the steady-state heat distribution data by utilizing the characteristic heat mode set to obtain a long-term temperature drift curve; and evaluating the long-term temperature drift curve to obtain a thermal reliability score and a thermal degradation prediction parameter, and generating life prediction data of the chip based on the thermal reliability score and the thermal degradation prediction parameter. By combining the analysis of transient thermal response data and steady-state thermal distribution data, the accuracy and reliability of temperature characteristic test are improved, and the problem that the traditional method is difficult to accurately capture the transient change characteristic of the chip temperature and identify the long-term temperature drift trend under a dynamic working condition or a complex load condition, and the test accuracy is insufficient is solved.

Description

Technical Field

[0001] The present application relates to the technical field of chip testing, and in particular to a chip temperature characteristic testing method, system and equipment. Background Art

[0002] With the continuous development and widespread application of chip technology, the integration and computing power of chips are gradually improving. However, with the increase in the complexity of the internal circuits of chips and the increase in power consumption, the heat generated by chips during operation has increased significantly, posing severe challenges to the stability and reliability of chips. In order to ensure the safety and performance of chips in different environments and working conditions, the testing and evaluation of chip temperature characteristics has become a crucial part of the chip design and production process.

[0003] Among the related technical means, the temperature distribution of the chip is analyzed by thermal imaging technology and thermocouple measurement. Among them, thermal imaging technology intuitively displays the temperature distribution on the chip surface through infrared imaging equipment, and can quickly locate the hot spot area; while thermocouple measurement technology provides accurate temperature values ​​inside or on the surface of the chip. These technologies have achieved the monitoring of chip temperature characteristics to a certain extent, and provide an important basis for chip thermal management design.

[0004] Regarding the above technical solutions, although the above technical means can preliminarily realize the visualization of chip surface temperature distribution and temperature detection of key parts, under dynamic working conditions or complex load conditions, traditional methods are difficult to accurately capture the transient change characteristics of chip temperature and identify long-term temperature drift trends, and there is a problem of insufficient test accuracy. Summary of the invention

[0005] In order to improve the problem of insufficient testing accuracy caused by traditional methods that are difficult to accurately capture the transient change characteristics of chip temperature and identify long-term temperature drift trends under dynamic working conditions or complex load conditions, the present application provides a chip temperature characteristic testing method, system and equipment.

[0006] The present invention provides a chip temperature characteristic testing method, comprising: obtaining transient thermal response data and steady-state thermal distribution data of a chip under different workload conditions, performing feature extraction on the transient thermal response data to obtain a characteristic thermal mode set; performing spatial heat transfer analysis on the steady-state thermal distribution data using the characteristic thermal mode set to obtain temperature drift deviation data and a temperature gradient evolution sequence; performing thermal influence factor tracing analysis on the characteristic thermal mode set using the temperature gradient evolution sequence to obtain a hotspot contribution matrix, performing adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix to generate a temperature offset correction factor; performing global thermal stability analysis on the temperature change data of the chip using the temperature offset correction factor to obtain a dynamic temperature distribution curve, performing multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain a number of local temperature rise trend sets, performing time series pattern comparison on all the local temperature rise trend sets to obtain a long-term temperature drift curve; evaluating the long-term temperature drift curve to obtain a thermal reliability score and a thermal degradation prediction parameter, and generating chip life prediction data based on the thermal reliability score and the thermal degradation prediction parameter.

[0007] As a preferred solution, the step of acquiring transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, performing feature extraction on the transient thermal response data, and obtaining a characteristic thermal pattern set includes: acquiring temperature sampling data of the chip under multiple preset workload modes, performing time series decomposition on the temperature sampling data, and obtaining transient thermal response data and steady-state thermal distribution data; performing feature extraction on the transient thermal response data in the time dimension and the space dimension, and obtaining a local thermal evolution feature set and a global heat transfer feature set; constructing a heat conduction association matrix based on the local thermal evolution feature set, and performing feature clustering analysis on the global heat transfer feature set using the heat conduction association matrix, and obtaining a characteristic thermal pattern set.

[0008] As a preferred scheme, the step of extracting features of the transient thermal response data in the time dimension and the space dimension to obtain a local thermal evolution feature set and a global heat transfer feature set includes: performing time series decomposition on the transient thermal response data to obtain thermal fluctuation features in the time dimension and local temperature change features in the space dimension; extracting the heat change rate and the temperature change frequency based on the thermal fluctuation features, and generating a transient thermal fluctuation data set based on the heat change rate and the temperature change frequency; performing spatial hotspot analysis on the temperature change data of the chip using the local temperature change features to obtain hotspot area distribution data and heat source intensity parameters; performing spatial clustering analysis on the hotspot area distribution data and the heat source intensity parameters to obtain a hotspot area heat source distribution pattern, and generating a local thermal evolution feature set based on the transient thermal fluctuation data set and the hotspot area heat source distribution pattern; performing global heat transfer calculation on the local thermal evolution feature set to obtain a global heat transfer feature set.

[0009] As a preferred solution, the step of using the characteristic thermal mode set to perform spatial heat transfer analysis on the steady-state thermal distribution data to obtain temperature drift deviation data and temperature gradient evolution sequence includes: using the characteristic thermal mode set to perform spatial heat transfer analysis on the steady-state thermal distribution data to obtain a thermal coupling matrix and regional thermal contribution data, optimizing the steady-state heat transfer path of the steady-state thermal distribution data based on the thermal coupling matrix to obtain a heat transfer weight matrix and local thermal conductivity distribution data; using the local thermal conductivity distribution data to calculate the thermal conduction characteristics of the characteristic thermal mode set to obtain A thermal influence factor model and a time-evolving thermal feature set are obtained, and time series modeling is performed on the time-evolving thermal feature set to obtain a transient thermal characteristic mapping function and a temperature transition parameter set; based on the transient thermal characteristic mapping function, a local thermal field evolution analysis is performed on the thermal influence factor model to obtain local temperature change data and a thermal inertia correction factor, and the temperature transition parameter set is dynamically compensated using the thermal inertia correction factor to obtain temperature drift deviation data and corrected thermal gradient data; based on the corrected thermal gradient data, a trend prediction calculation is performed on the local temperature change data to obtain a temperature gradient evolution sequence.

[0010] As a preferred scheme, the step of using the temperature gradient evolution sequence to perform a thermal influence factor tracing analysis on the characteristic thermal mode set to obtain a hotspot contribution matrix, and performing adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix to generate a temperature offset correction factor includes: using the temperature gradient evolution sequence to perform a thermal influence path analysis on the characteristic thermal mode set to obtain a thermal influence factor weight matrix, and performing thermal influence contribution calculation on the characteristic thermal mode set based on the thermal influence factor weight matrix to obtain a hotspot contribution matrix and characteristic thermal mode offset parameters; performing deviation calculation on the temperature gradient evolution sequence according to the hotspot contribution matrix A matching calculation is performed to obtain a hotspot influence correction factor, and a nonlinear error correction calculation is performed on the characteristic thermal mode offset parameter based on the hotspot influence correction factor to obtain a hotspot compensation parameter and a temperature offset correction matrix; the temperature offset correction matrix is ​​applied to perform a multi-dimensional compensation calculation on the temperature drift deviation data to obtain a dynamic temperature correction factor and a steady-state temperature correction factor, and a hotspot thermal influence optimization is performed on the hotspot compensation parameter based on the dynamic temperature correction factor to obtain a hotspot temperature compensation factor and a balanced thermal influence factor; a steady-state temperature deviation optimization calculation is performed on the steady-state temperature correction factor according to the hotspot temperature compensation factor and the balanced thermal influence factor to obtain a temperature offset correction factor.

[0011] As a preferred scheme, the steps of using the temperature offset correction factor to perform a global thermal stability analysis on the temperature change data of the chip to obtain a dynamic temperature distribution curve, performing a multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain a number of local temperature rise trend sets, and performing time series pattern comparison on all the local temperature rise trend sets to obtain a long-term temperature drift curve include: using the temperature offset correction factor to perform a global thermal stability analysis on the temperature change data of the chip to obtain a dynamic temperature distribution curve and local temperature rise data, performing multi-scale time series decomposition on the dynamic temperature distribution curve to obtain a short-term temperature change pattern and a long-term temperature change trend; performing pattern matching analysis on the local temperature rise data based on the long-term temperature change trend to obtain a local temperature rise trend set, performing pattern classification calculation on the local temperature rise trend set to obtain a drift trend set; and using the drift trend set to perform pattern adjustment on the short-term temperature change pattern to obtain a long-term temperature drift curve.

[0012] As a preferred solution, the step of evaluating the long-term temperature drift curve to obtain a thermal reliability score and thermal degradation prediction parameters, and generating chip life prediction data based on the thermal reliability score and the thermal degradation prediction parameters includes: obtaining temperature stability change data and trend offset factor of the long-term temperature drift curve, performing steady-state temperature distribution analysis on the temperature stability change data to obtain temperature reliability characteristic parameters and abnormal thermal drift factors; performing temperature stability evaluation calculation on the trend offset factor using the temperature reliability characteristic parameters to obtain a steady-state thermal reliability index and temperature anomaly prediction parameters, and performing working stability analysis on the long-term temperature drift curve based on the steady-state thermal reliability index to obtain a thermal reliability score and a long-term temperature stability curve ; The steady-state thermal change trend of the temperature anomaly prediction parameter is evaluated through the thermal reliability score to obtain working stability evaluation data and a thermal degradation parameter set, and the life decay characteristics of the long-term temperature stability curve are extracted based on the thermal degradation parameter set to obtain thermal degradation prediction parameters and life influencing factors; The thermal aging trend of the working stability evaluation data is calculated according to the thermal degradation prediction parameters to obtain an initial life degradation trend and a thermal safety margin parameter, and the life prediction optimization calculation is performed on the initial life degradation trend based on the thermal safety margin parameter to obtain a chip life degradation trend and a temperature safety threshold; The chip life degradation trend and the temperature safety threshold are used to perform a life evaluation calculation on the long-term temperature drift curve of the chip to obtain chip life prediction data.

[0013] The present application also provides a chip temperature characteristic testing system, including: an acquisition module, used to acquire transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, perform feature extraction on the transient thermal response data, and obtain a characteristic thermal mode set; an analysis module, used to perform spatial heat transfer analysis on the steady-state thermal distribution data using the characteristic thermal mode set, and obtain temperature drift deviation data and a temperature gradient evolution sequence; a calculation module, used to perform thermal influence factor tracing analysis on the characteristic thermal mode set using the temperature gradient evolution sequence, and obtain a hotspot contribution matrix, and calculate the temperature drift deviation data according to the hotspot contribution matrix. The invention relates to a method for performing adaptive temperature compensation calculation on the chip to generate a temperature offset correction factor; a comparison module is used to perform a global thermal stability analysis on the temperature change data of the chip using the temperature offset correction factor to obtain a dynamic temperature distribution curve, perform a multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain a number of local temperature rise trend sets, perform a time series pattern comparison on all the local temperature rise trend sets to obtain a long-term temperature drift curve; a generation module is used to evaluate the long-term temperature drift curve to obtain a thermal reliability score and thermal degradation prediction parameters, and generate chip life prediction data based on the thermal reliability score and the thermal degradation prediction parameters.

[0014] The present application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the chip temperature characteristic testing method described in any one of the above is implemented.

[0015] Compared with the prior art, the present application has the following beneficial effects: accurate identification and high test accuracy. By acquiring the transient thermal response data and steady-state thermal distribution data of the chip, combined with the generation and use of a characteristic thermal pattern set, a comprehensive temperature analysis from dynamic to steady-state, from local to global is achieved. Through in-depth mining of temperature drift deviation data and temperature gradient evolution sequence, an accurate assessment of the heat transfer path and hot spot contribution inside the chip is completed. The temperature offset correction factor generated by adaptive temperature compensation calculation and the analysis of the long-term temperature drift curve greatly improve the accuracy and reliability of the chip thermal stability prediction, and improve the traditional method under dynamic working conditions or complex load conditions. It is difficult to accurately capture the transient change characteristics of the chip temperature and identify the long-term temperature drift trend, and there is a problem of insufficient test accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0017] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0018] Figure 1 It is a schematic diagram of the flow of a chip temperature characteristic testing method provided by an embodiment of the present invention; Figure 2 is a schematic block diagram of the structure of a chip temperature characteristic testing system provided by an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0019] Description of reference numerals: 10. Chip temperature characteristic test system; 11. Acquisition module; 12. Analysis module; 13. Calculation module; 14. Comparison module; 15. Generation module; 20. Electronic device; 21. Memory; 22. Processor. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0022] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0024] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0025] Embodiment 1: like Figure 1 As shown, the present application provides a chip temperature characteristic testing method, including steps S100 to S500.

[0026] Step S100: Obtain transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, perform feature extraction on the transient thermal response data, and obtain a set of characteristic thermal patterns.

[0027] In this step, the chip operating conditions are adjusted to collect transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions. Specifically, the chip is placed under multiple load modes (such as light load, medium load, and full load), and thermocouples or high-precision infrared thermal imaging equipment are used to record the transient temperature changes and steady-state temperature distribution of the chip. In addition, the collected transient thermal response data is analyzed with the help of digital signal processing algorithms, feature data is extracted, and a set of feature thermal pattern sets is generated for use in subsequent steps.

[0028] For example, in light load mode, the transient thermal response data of the chip was recorded, and it was found that its heat rise rate was slow and the heat distribution was relatively uniform; while in full load mode, the transient thermal response data showed that the temperature rise of the chip hotspots was significantly accelerated, and the hotspot area was concentrated in the computing core area.

[0029] Step S200: Perform spatial heat transfer analysis on the steady-state heat distribution data using a set of characteristic thermal patterns to obtain temperature drift deviation data and a temperature gradient evolution sequence.

[0030] In this step, the characteristic thermal pattern set is used to perform spatial heat transfer analysis on the steady-state thermal distribution data to extract the heat transfer path and temperature difference characteristics inside the chip. Specifically, the characteristic thermal pattern set is matched point by point with the steady-state thermal distribution data, the temperature drift deviation data is calculated through the heat transfer matrix, and the temperature gradient evolution sequence is constructed in combination with the time stepping method to depict the flow path of heat inside the chip.

[0031] For example, by analyzing the steady-state thermal distribution data of a certain processor, it was found that the heat in the hot spot area inside the chip mainly diffuses to the surrounding area along the metal interconnection path, while the heat accumulation in the peripheral low-power area is less.

[0032] Step S300: Use the temperature gradient evolution sequence to perform thermal influence factor tracing analysis on the characteristic thermal mode set to obtain a hotspot contribution matrix, perform adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix, and generate a temperature offset correction factor.

[0033] In this step, the temperature gradient evolution sequence is used to trace the thermal influence factor of the characteristic thermal mode set, and the contribution of each hot spot to the overall temperature distribution of the chip is calculated. Specifically, the hot spot contribution matrix is ​​generated by establishing a hot spot contribution model, and iterative optimization is performed in combination with the temperature drift deviation data to calculate the temperature offset correction factor. This correction factor can be used to correct the temperature error of the chip and improve the accuracy of temperature measurement.

[0034] For example, in a certain graphics processor, the analysis results show that the hotspot contribution of the core computing unit is 70%, while the hotspot contribution of the memory controller is only 15%.

[0035] Step S400, using the temperature offset correction factor to perform a global thermal stability analysis on the temperature change data of the chip to obtain a dynamic temperature distribution curve, performing a multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain several local temperature rise trend sets, performing a time series pattern comparison on all local temperature rise trend sets to obtain a long-term temperature drift curve.

[0036] In this step, the temperature offset correction factor is used to perform a global thermal stability analysis on the temperature change data of the chip, and a dynamic temperature distribution curve is generated. Specifically, the dynamic temperature distribution curve is decomposed into multiple local temperature rise trend sets through multi-scale temperature balance analysis, and these trend sets are compared one by one in combination with time series pattern matching technology, and finally a long-term temperature drift curve is generated to characterize the long-term thermal stability characteristics of the chip.

[0037] For example, through multi-scale analysis of the dynamic temperature distribution curve of a low-power AI chip, it was found that the temperature rise trend of its computing core has a periodic fluctuation characteristic, while the temperature rise trend of the peripheral module shows a slow but continuous increase.

[0038] Step S500: Evaluate the long-term temperature drift curve to obtain a thermal reliability score and thermal degradation prediction parameters, and generate chip life prediction data based on the thermal reliability score and thermal degradation prediction parameters.

[0039] In this step, the thermal reliability score and thermal degradation prediction parameters of the chip are obtained by evaluating the long-term temperature drift curve. Specifically, the expected service life of the chip is calculated using the data of the long-term temperature drift curve in combination with the thermal reliability model and degradation prediction algorithm, providing guidance for the thermal design of the chip.

[0040] For example, after evaluating a certain communication chip, it was found that although its initial thermal stability score was high, due to the large long-term temperature drift of the core module, the thermal degradation prediction parameters indicated that its reliable service life could be shortened to 80% of the original design life.

[0041] In this implementation, transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions are obtained, and features are extracted from the transient thermal response data to generate a characteristic thermal pattern set. Subsequently, the characteristic thermal pattern set is used to perform spatial heat transfer analysis on the steady-state thermal distribution data, thereby obtaining temperature drift deviation data and temperature gradient evolution sequence. Next, the characteristic thermal pattern set is traced for thermal influence factor analysis through the temperature gradient evolution sequence to generate a hotspot contribution matrix, and the temperature drift deviation data is adaptively calculated based on the hotspot contribution matrix to obtain the temperature offset correction factor. Then, the temperature offset correction factor is used to perform global thermal stability analysis on the temperature change data of the chip to generate a dynamic temperature distribution curve, and a multi-scale temperature balance analysis is performed on the dynamic temperature distribution curve to generate a local temperature rise trend set. Finally, a time series pattern comparison is performed on all local temperature rise trend sets to obtain a long-term temperature drift curve, and the long-term temperature drift curve is evaluated to generate a thermal reliability score and thermal degradation prediction parameters, and finally the chip life prediction data is formed.

[0042] Through the characteristic thermal pattern set and adaptive temperature compensation calculation, the chip temperature distribution is accurately analyzed. This method can not only capture the transient thermal response data of the chip under dynamic conditions, but also track the hot spot contribution matrix through the temperature gradient evolution sequence, and fully reveal the internal heat transfer mechanism of the chip. In addition, through the time series pattern comparison and long-term temperature drift curve analysis, the scheme can effectively predict the thermal reliability and service life of the chip, thereby providing a scientific basis for chip thermal design optimization and greatly improving the stability and service life of the chip.

[0043] Embodiment 2: In step S100, temperature sampling data of the chip under a plurality of preset workload modes is acquired, and the temperature sampling data is decomposed into a time series to obtain transient thermal response data and steady-state thermal distribution data.

[0044] By setting different working modes such as low, medium and high load to operate the chip, high-precision thermocouples and infrared thermal imaging equipment are used to simultaneously collect temperature change data on the chip surface and inside, and record its thermal heating curve and steady-state distribution characteristics under each load condition. Specifically, by combining fast Fourier transform to decompose the collected data into time series, the temperature change characteristics of the transient thermal response stage and the temperature distribution characteristics of the steady-state stage are extracted respectively to ensure the accuracy and integrity of the two types of data.

[0045] For example, in low-load mode, temperature sampling data shows that the chip's transient thermal response is smooth and uniform, and the steady-state thermal distribution shows a small temperature difference; in high-load mode, the transient thermal response is violent and there are significant hot spots, and the steady-state thermal distribution characteristics reflect the local heat accumulation phenomenon of core components.

[0046] The features of transient thermal response data are extracted in time and space dimensions to obtain the local thermal evolution feature set and the global heat transfer feature set.

[0047] By performing multi-level feature extraction on transient thermal response data in the time dimension and space dimension, local thermal evolution feature set and global heat transfer feature set are generated. Specifically, the rate and periodic characteristics of temperature change are captured by wavelet decomposition in the time dimension; the change path and diffusion pattern of the hot spot area are extracted by thermal distribution gradient analysis in the space dimension, thereby describing the local thermal evolution and global heat transfer behavior inside the chip.

[0048] For example, in a high-performance processor, time dimension analysis found that its thermal fluctuation period was 0.1 seconds, while space dimension analysis showed that the thermal diffusion rate in the hot spot area reached 2 mm per second.

[0049] A heat conduction correlation matrix is ​​constructed based on the local thermal evolution feature set, and the heat conduction correlation matrix is ​​used to perform feature clustering analysis on the global heat transfer feature set to obtain a set of characteristic thermal modes.

[0050] By establishing a heat conduction correlation matrix based on the local thermal evolution feature set, the intensity and path of heat transfer between different regions of the chip are quantified. Specifically, the thermal coupling intensity is calculated through the correlation matrix, and the global heat transfer feature set is subjected to K-means clustering analysis, so that the complex heat transfer behavior can be classified into several characteristic thermal mode sets.

[0051] For example, the clustering results of a graphics processing chip show that the main characteristic thermal patterns can be divided into three categories: core module thermal aggregation pattern, peripheral module thermal diffusion pattern and uniform thermal distribution pattern.

[0052] Among them, the step of extracting features of the transient thermal response data in the time dimension and the space dimension to obtain a local thermal evolution feature set and a global heat transfer feature set includes: performing time series decomposition on the transient thermal response data to obtain thermal fluctuation features in the time dimension and local temperature change features in the space dimension.

[0053] By decomposing the transient thermal response data in time series, the thermal fluctuation characteristics and local temperature change characteristics of different time stages are separated and extracted. Specifically, in the time dimension, the temperature fluctuation behavior of different frequency components is captured by fast Fourier transform; in the spatial dimension, the temperature gradient change characteristics of local hot spots are fine-grained analyzed by combining spatial interpolation method.

[0054] For example, in a certain data processing chip, the time series decomposition results show that high-frequency thermal fluctuations mainly come from the computing core, while the spatial dimension analysis results show that the local temperature rise is concentrated near the memory unit.

[0055] The thermal change rate and temperature change frequency are extracted based on the thermal fluctuation characteristics, and a transient thermal fluctuation data set is generated based on the thermal change rate and temperature change frequency.

[0056] By analyzing the thermal fluctuation characteristics, the thermal change rate and temperature change frequency of the chip under transient conditions are extracted, and these key features are combined to generate a transient thermal fluctuation data set. Specifically, the thermal change rate is calculated using the gradient method, and the main frequency range of the temperature change is determined using spectrum analysis, thus forming a complete transient thermal fluctuation data set.

[0057] For example, the transient thermal fluctuation data set of a low-power chip shows that the thermal change rate of the core module reaches 15°C per second, while the temperature change frequency is mainly concentrated between 0.2Hz and 2Hz.

[0058] The local temperature change characteristics are used to perform spatial hot spot analysis on the temperature change data of the chip to obtain the hot spot area distribution data and heat source intensity parameters.

[0059] By spatially analyzing the temperature change data of the chip, the hotspot areas inside the chip are identified and the heat source intensity of the hotspot is quantified. Specifically, by mapping the collected temperature change data with the chip structure layout, the temperature distribution of each area is gradient analyzed to locate the hotspot area with the highest temperature; at the same time, the heat source intensity is calculated based on the time change rate of the regional temperature, thereby forming a complete set of hotspot area distribution data and heat source intensity parameters.

[0060] For example, in a high-performance graphics processing chip, spatial hotspot analysis found that the heat source intensity in the computing core area reached 20°C per second, while the heat source intensity in the peripheral circuit module was less than 5°C per second, and the hotspots were mainly concentrated in the central area of ​​the chip.

[0061] Spatial clustering analysis is performed on the hotspot area distribution data and heat source intensity parameters to obtain the heat source distribution pattern of the hotspot area. The local thermal evolution feature set is generated based on the transient thermal fluctuation dataset and the heat source distribution pattern of the hotspot area.

[0062] By clustering the hotspot distribution data and heat source intensity parameters, the distribution patterns and heat source characteristics of different hotspots in the chip are identified. Specifically, a density clustering algorithm (such as DBSCAN) is used to classify adjacent hotspot areas, and the heat source distribution pattern of the hotspot area is generated through the distribution characteristics of the heat source intensity; then, combined with the dynamic characteristics of the transient thermal fluctuation data set, the hotspot distribution pattern is mapped into a local thermal evolution feature set to describe the evolution behavior and impact range of the hotspot area.

[0063] For example, the analysis results of an AI inference chip show that its heat source distribution patterns include concentrated heat source distribution in the core module, strip-type heat source distribution in the peripheral module, and dispersed heat source distribution in the interface module. These patterns collectively constitute a local thermal evolution feature set.

[0064] The global heat transfer calculation is performed on the local thermal evolution feature set to obtain the global heat transfer feature set.

[0065] Through global calculation of the local thermal evolution feature set, the heat transfer path and influence range between different regions inside the chip are analyzed to generate a global heat transfer feature set. Specifically, the finite element analysis method is used to simulate the heat transfer process, build a thermal coupling matrix between regions, and combine time series analysis to predict the heat change trend in different regions, thus forming a complete global heat transfer feature set.

[0066] For example, in the global heat transfer calculation of a communication chip, it was found that the heat of the core module was mainly transferred to the peripheral area through the metal interconnection path, while the heat influence of the heat dissipation module on the core area was concentrated in the edge area, forming the main heat transfer path and mode.

[0067] In step S200, a spatial heat transfer analysis is performed on the steady-state thermal distribution data using a set of characteristic thermal patterns to obtain a thermal coupling matrix and regional thermal contribution data. Based on the thermal coupling matrix, a steady-state heat transfer path is optimized on the steady-state thermal distribution data to obtain a heat transfer weight matrix and local thermal conductivity distribution data.

[0068] By using a set of characteristic thermal patterns to conduct an in-depth spatial heat transfer analysis on the steady-state thermal distribution data of the chip, a thermal coupling matrix is ​​constructed to quantify the thermal coupling intensity between the various regions of the chip, and the thermal contribution of each region is calculated. Specifically, the heat conduction equation is used to analyze the temperature gradient in the steady-state thermal distribution, and the heat transfer path diagram between regions is generated in combination with the chip's geometric structure and material parameters to construct a thermal coupling matrix; on this basis, the main heat transfer path is optimized through a path optimization algorithm to generate a heat transfer weight matrix and local thermal conductivity distribution data to describe the priority of heat transfer and thermal conductivity differences between regions.

[0069] For example, in the steady-state thermal distribution analysis of a computing chip, it was found that the thermal coupling intensity between the core area and the heat dissipation module reached 85%. By optimizing the heat transfer path, the excess heat distribution in non-critical areas was significantly reduced, making the thermal conductivity distribution more balanced.

[0070] The local thermal conductivity distribution data is used to calculate the heat conduction characteristics of the characteristic thermal mode set to obtain the thermal influence factor model and the time-evolving thermal feature set. The time-evolving thermal feature set is modeled by time series to obtain the transient thermal characteristic mapping function and the temperature transition parameter set.

[0071] By combining the local thermal conductivity distribution data with the characteristic thermal mode set, the heat conduction characteristics are calculated, the influencing factors of each area in the chip on the overall thermal distribution are extracted, and a thermal influencing factor model is constructed. At the same time, the calculation results are used to extract and analyze the time-evolving thermal feature set, and a dynamically changing time series model is established to generate transient thermal characteristic mapping functions and temperature transition parameter sets. Specifically, by performing a fine-grained analysis of the local thermal conductivity in the heat conduction path, the dynamic impact of temperature changes on the surrounding areas is calculated, and these dynamic characteristics are converted into mapping functions and parameter sets to describe the transient behavior of heat transfer.

[0072] For example, the thermal conduction analysis results of a high-performance AI chip show that the thermal impact factor of the computing core is 0.75, while the thermal impact factor of the peripheral storage module is only 0.2. At the same time, the temperature transition parameter analysis shows that the temperature change frequency in the key area is concentrated between 0.5Hz and 1Hz.

[0073] Based on the transient thermal characteristic mapping function, the local thermal field evolution analysis of the thermal influence factor model is carried out to obtain the local temperature change data and thermal inertia correction factor. The thermal inertia correction factor is used to dynamically compensate the temperature transition parameter set to obtain the temperature drift deviation data and corrected thermal gradient data.

[0074] By analyzing the local thermal field evolution of the thermal impact factor model based on the transient thermal characteristic mapping function, the local temperature change characteristics of the chip under dynamic conditions are fully captured, and the thermal inertia correction factor is further calculated to correct the errors introduced in the heat transfer process. Specifically, the heat transfer path of the chip is dynamically analyzed through simulation modeling, local temperature change data is extracted, and the temperature transition parameters are compensated in combination with the thermal inertia correction algorithm, thereby generating temperature drift deviation data and corrected thermal gradient distribution data to characterize the true characteristics of the long-term temperature change of the chip.

[0075] For example, in the thermal inertia analysis of a certain memory chip, it was found that the thermal inertia of the local module was large, resulting in deviations in the original temperature transition parameters. The temperature drift data generated after correction showed that the overall thermal distribution of the chip was closer to the actual situation.

[0076] Based on the corrected thermal gradient data, the trend prediction calculation of the local temperature change data is performed to obtain the temperature gradient evolution sequence.

[0077] By using the corrected thermal gradient distribution data, the trend prediction analysis of the local temperature change data of the chip is carried out to generate a temperature gradient evolution sequence and predict the long-term trend of the temperature change inside the chip. Specifically, based on the time series prediction model, the temperature change rate and direction of the local area are analyzed in combination with the corrected thermal gradient data, and sequence data for characterizing the temperature change trend is generated to support subsequent thermal reliability evaluation and optimization design.

[0078] For example, the analysis results of a certain chip show that the temperature gradient evolution sequence in its core area presents a linear growth trend, while the temperature gradient evolution sequence in the peripheral module exhibits periodic fluctuation characteristics, providing an optimization basis for thermal design.

[0079] In step S300, the temperature gradient evolution sequence is used to perform heat impact path analysis on the characteristic thermal mode set to obtain a heat impact factor weight matrix, and the heat impact contribution of the characteristic thermal mode set is calculated based on the heat impact factor weight matrix to obtain a hotspot contribution matrix and characteristic thermal mode offset parameters.

[0080] By combining the temperature gradient evolution sequence and the characteristic thermal mode set, the heat impact path in the chip is finely analyzed to determine the influence of each hot spot in the overall heat distribution and generate a heat impact factor weight matrix. Specifically, the heat transfer model is used to calculate the heat impact intensity and path relationship of each characteristic thermal mode, and the contribution of each characteristic mode is quantified according to the analysis results to generate a hotspot contribution matrix. At the same time, based on the comparison between the hotspot contribution matrix and the characteristic thermal mode set, the offset parameter is calculated to identify the error range of the characteristic mode in the actual thermal impact.

[0081] For example, the thermal impact path analysis of a high-performance computing chip shows that the hotspot contribution of the central computing unit accounts for about 70%, while the contribution of the peripheral module is only 15%. The weight matrix reveals the significant impact of the core area on the overall heat distribution of the chip.

[0082] The temperature gradient evolution sequence is matched with the hotspot contribution matrix to obtain the hotspot influence correction factor. The characteristic thermal mode offset parameters are corrected by nonlinear error correction based on the hotspot influence correction factor to obtain the hotspot compensation parameters and the temperature offset correction matrix.

[0083] By performing deviation matching analysis on the hotspot contribution matrix and the temperature gradient evolution sequence, the difference between the actual observation value and the model prediction value is identified and calculated, thereby generating a hotspot impact correction factor. Specifically, the characteristic thermal mode offset parameters are corrected using a nonlinear regression method to eliminate the effect of errors, and the hotspot compensation parameters are generated from the corrected results. At the same time, a temperature offset correction matrix is ​​constructed for further optimization of the global heat distribution.

[0084] For example, in the analysis of a certain AI chip, it was found through correction calculation that the originally predicted hotspot offset was large. After using the hotspot impact correction factor, the deviation was reduced by 20%. The final generated temperature offset correction matrix significantly improved the accuracy of the temperature distribution.

[0085] The temperature offset correction matrix is ​​used to perform multi-dimensional compensation calculation on the temperature drift deviation data to obtain the dynamic temperature correction factor and the steady-state temperature correction factor. The hotspot thermal influence optimization of the hotspot compensation parameters is performed based on the dynamic temperature correction factor to obtain the hotspot temperature compensation factor and the balanced thermal influence factor.

[0086] By applying the temperature offset correction matrix to the temperature drift deviation data, multi-dimensional compensation calculations are performed to generate dynamic temperature correction factors and steady-state temperature correction factors to correct transient and steady-state heat distribution errors respectively. Specifically, the hotspot compensation parameters are optimized through the dynamic temperature correction factor to improve the accuracy of the heat transfer characteristics of the hotspot area, and the balanced heat impact factor is generated to describe the balanced contribution of the hotspot area to the overall heat distribution.

[0087] For example, in a certain graphics processing chip, after optimization using the dynamic temperature correction factor, the hotspot temperature of the core module dropped by 15%, while the introduction of the balanced heat influence factor made the temperature rise control of the peripheral modules more stable.

[0088] The steady-state temperature correction factor is optimized and calculated based on the hot spot temperature compensation factor and the equilibrium heat influence factor to obtain the temperature offset correction factor.

[0089] By using the hotspot temperature compensation factor and the balanced heat impact factor, the steady-state temperature correction factor is optimized and calculated to generate the final temperature offset correction factor for precise control of the overall temperature change of the chip. Specifically, through the joint analysis of dynamic and steady-state data, the deviation in the steady-state temperature is corrected to improve the consistency of the overall temperature distribution.

[0090] For example, in the optimization calculation of a certain communication chip, the temperature offset correction factor finally generated significantly reduced the thermal drift phenomenon of the chip during long-term operation and improved the reliability of its thermal design.

[0091] In step S400, the temperature change data of the chip is subjected to a global thermal stability analysis using a temperature offset correction factor to obtain a dynamic temperature distribution curve and local temperature rise data. The dynamic temperature distribution curve is subjected to a multi-scale time series decomposition to obtain a short-term temperature change pattern and a long-term temperature change trend.

[0092] By combining the temperature offset correction factor to perform a global thermal stability analysis on the temperature change data of the chip, the dynamic characteristics of the heat distribution inside the chip are quantified. Specifically, a heat transfer simulation model is used to track and analyze the dynamic temperature change process of the chip under different working conditions, extract the overall temperature change curve, and separate the local temperature rise data from it. Next, through multi-scale time series decomposition, the dynamic temperature distribution curve is decomposed into short-term temperature change patterns and long-term temperature change trends, which respectively characterize the thermal response characteristics of the chip in a short period of time and the thermal stability changes under long-term operation.

[0093] For example, in a chip analysis, the short-term temperature change pattern reflects the rapid temperature rise caused by transient loads, while the long-term temperature change trend shows that the temperature of the computing core area presents a slow linear growth characteristic over a long period of time.

[0094] Based on the long-term temperature change trend, pattern matching analysis is performed on the local temperature rise data to obtain a local temperature rise trend set. Pattern classification calculation is performed on the local temperature rise trend set to obtain a drift trend set.

[0095] By performing pattern matching analysis on the long-term temperature change trend and the local temperature rise data, the temperature rise rules of different regions are extracted and a local temperature rise trend set is generated. Specifically, a matching algorithm is used to compare the temperature change curve of each local region with the known trend template, and the most relevant pattern set is extracted according to the similarity. Subsequently, the local temperature rise trend set is further classified and analyzed through pattern classification calculation to generate a drift trend set to describe the thermal drift characteristics and evolution paths of different regions inside the chip.

[0096] For example, in the pattern analysis of a certain AI chip, it was found that the temperature rise trend of the core module showed an exponential growth pattern, while the peripheral circuit area showed periodic fluctuation characteristics. These trends were classified into different drift trend sets.

[0097] The drift trend set is used to adjust the short-term temperature change pattern to obtain the long-term temperature drift curve.

[0098] By combining the drift trend set to correct and adjust the short-term temperature change pattern, a long-term temperature drift curve is generated to characterize the thermal stability of the chip. Specifically, a pattern synthesis algorithm is used to embed the features of the drift trend set into the short-term temperature change pattern to correct its local deviation and improve the accuracy of the overall trend description, thereby forming a long-term temperature drift curve that comprehensively describes the thermal stability evolution process of the chip.

[0099] For example, in the analysis of a high-power chip, by adjusting the drift trend of the short-term temperature rise data of the core module, the generated long-term temperature drift curve accurately reflects the temperature rise and heat diffusion behavior of the module under continuous high-load operation, providing a reference for chip heat dissipation design.

[0100] In step S500, the temperature stability change data and trend shift factor of the long-term temperature drift curve are obtained, and the steady-state temperature distribution analysis is performed on the temperature stability change data to obtain the temperature reliability characteristic parameter and the abnormal thermal drift factor.

[0101] By analyzing the temperature stability change data of the long-term temperature drift curve and combining it with the trend offset factor, the thermal stability change characteristics of the chip under long-term operation can be identified. Specifically, by calculating the mean and standard deviation of the steady-state temperature distribution, the stability characteristics of the temperature change are extracted, and combined with the anomaly detection algorithm, potential abnormal thermal drift behavior is identified and an abnormal thermal drift factor is generated.

[0102] For example, in the analysis of an AI inference chip, the temperature stability of the core module was 85°C on average and 5°C on standard deviation, but abnormal thermal drift was detected at certain time points, with an abnormal factor of 10%.

[0103] The temperature reliability characteristic parameters are used to evaluate the temperature stability of the trend shift factor, and the steady-state thermal reliability index and temperature anomaly prediction parameters are obtained. Based on the steady-state thermal reliability index, the long-term temperature drift curve is analyzed for working stability, and the thermal reliability score and long-term temperature stability curve are obtained.

[0104] By using temperature reliability characteristic parameters and trend offset factors, the thermal stability of the chip is evaluated and calculated to quantify its steady-state thermal reliability level. Specifically, the steady-state thermal reliability index is calculated based on the statistical model, and the possible temperature abnormal behavior in the future is predicted through trend analysis to generate temperature abnormality prediction parameters. Then, these indicators are applied to the long-term temperature drift curve to analyze the working stability of the chip under different working conditions. Finally, a comprehensive thermal reliability score and long-term temperature stability curve are obtained to describe the dynamic changes of the chip's thermal reliability.

[0105] For example, the steady-state thermal reliability index of a low-power chip is 0.9, the predicted temperature anomaly probability is 5%, the thermal reliability score generated by working stability analysis is 85 points, and a trend curve is obtained that the temperature gradually tends to be stable under long-term operation.

[0106] The steady-state thermal change trend of the temperature anomaly prediction parameters is evaluated through thermal reliability scoring to obtain the working stability evaluation data and thermal degradation parameter set. The life decay characteristics of the long-term temperature stability curve are extracted based on the thermal degradation parameter set to obtain the thermal degradation prediction parameters and life influencing factors.

[0107] Combined with the thermal reliability score and temperature anomaly prediction parameters, the steady-state thermal change trend of the chip is evaluated, key stability data is extracted, and a thermal degradation parameter set is generated based on these data. Specifically, the life decay characteristics in the long-term temperature stability curve are analyzed through the thermal degradation model, the thermal degradation prediction parameters are extracted, and the life impact factor of the chip is calculated to evaluate the change trend of its thermal reliability during the life cycle.

[0108] For example, an evaluation of a high-performance computing chip showed that its initial working stability score was 90 points, but as the operating time increased, the thermal degradation prediction parameters showed that the reliability score decreased at a rate of 5% per year.

[0109] According to the thermal degradation prediction parameters, the thermal aging trend of the working stability evaluation data is calculated to obtain the initial life degradation trend and thermal safety margin parameters. Based on the thermal safety margin parameters, the life prediction optimization calculation of the initial life degradation trend is performed to obtain the chip life degradation trend and temperature safety threshold.

[0110] Through further analysis of thermal degradation prediction parameters, combined with working stability assessment data, the thermal aging trend of the chip is calculated, the main characteristics of the initial life decay stage are identified, and the initial life degradation trend and thermal safety margin parameters are generated. Specifically, the thermal aging model is combined with reliability data to predict the degradation process of the chip from the initial state to the failure state, and the life prediction model is optimized to finally determine the chip's life degradation trend and temperature safety threshold, which are used to evaluate the chip's usable period and thermal management requirements.

[0111] For example, thermal aging analysis of a graphics chip shows that the life degradation is slow in the initial stage, but the thermal safety margin decreases rapidly under high load conditions. The optimization model predicts that its safe operating temperature threshold is 95°C and its expected life is 5 years.

[0112] The chip life degradation trend and temperature safety threshold are used to perform life assessment calculations on the long-term temperature drift curve of the chip to obtain chip life prediction data.

[0113] By applying the chip's life degradation trend and temperature safety threshold to the long-term temperature drift curve, life assessment calculations are performed to generate chip life prediction data. Specifically, the reliability prediction algorithm is used to comprehensively calculate the chip's operating performance and thermal stability, predict the chip's service life and performance degradation rate, and provide important guidance for chip design and optimization.

[0114] For example, the life prediction results of a low-power AI chip show that its expected service life under standard operating conditions is 7 years, while under high-load conditions its life will be shortened to 5 years, providing reliable life reference data for design.

[0115] In this implementation, by obtaining the temperature sampling data of the chip under multiple preset workload modes and decomposing it in time series, transient thermal response data and steady-state thermal distribution data are generated, and features are extracted in time and space dimensions to construct local thermal evolution feature sets and global heat transfer feature sets, laying the foundation for analyzing the heat transfer law inside the chip. This scheme constructs a heat conduction association matrix based on the local thermal evolution feature set, generates a characteristic thermal mode set through cluster analysis, and uses this set to perform spatial heat transfer analysis on the steady-state thermal distribution data, extracts the thermal coupling matrix and regional thermal contribution data, and optimizes the steady-state heat transfer path to obtain the heat transfer weight matrix and local thermal conductivity distribution data. In addition, by further calculating the thermal impact factor model and the time evolution thermal feature set, combined with the transient thermal characteristic mapping function and the temperature transition parameter set, the local thermal field evolution is accurately analyzed, the corrected temperature drift deviation data and thermal gradient data are generated, and the trend is predicted to obtain the temperature gradient evolution sequence. This method also generates a long-term temperature drift curve by matching and analyzing the local temperature rise trend set, classifying the drift trend set and adjusting the short-term temperature change mode to comprehensively characterize the thermal stability evolution of the chip. Finally, based on the long-term temperature drift curve, the temperature stability change data and thermal degradation characteristics are extracted, the life prediction model is optimized, and the thermal degradation prediction parameters, chip life degradation trend and temperature safety threshold are obtained. The chip thermal characteristics analysis from transient to steady state, from local to global is realized, and the reliability of thermal design and the accuracy of chip life prediction are improved through optimization calculation, which is suitable for thermal management and reliability evaluation of high-performance chips.

[0116] Embodiment 3: like Figure 2 As shown, the present application also provides a chip temperature characteristic testing system 10 , including an acquisition module 11 , an analysis module 12 , a calculation module 13 , a comparison module 14 and a generation module 15 .

[0117] The acquisition module 11 is mainly used to acquire the transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, perform feature extraction on the transient thermal response data, and obtain a set of characteristic thermal patterns.

[0118] The analysis module 12 is mainly used to perform spatial heat transfer analysis on the steady-state heat distribution data using a set of characteristic thermal modes, and obtain temperature drift deviation data and a temperature gradient evolution sequence.

[0119] The calculation module 13 is mainly used to perform thermal influence factor tracing analysis on the characteristic thermal mode set using the temperature gradient evolution sequence, obtain the hotspot contribution matrix, perform adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix, and generate a temperature offset correction factor.

[0120] The comparison module 14 is mainly used to perform a global thermal stability analysis on the temperature change data of the chip using the temperature offset correction factor to obtain a dynamic temperature distribution curve, perform a multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain several local temperature rise trend sets, perform a time series pattern comparison on all local temperature rise trend sets to obtain a long-term temperature drift curve.

[0121] The generation module 15 is mainly used to evaluate the long-term temperature drift curve, obtain the thermal reliability score and thermal degradation prediction parameters, and generate chip life prediction data based on the thermal reliability score and thermal degradation prediction parameters.

[0122] In this embodiment, through the coordinated work of the acquisition module 11, the analysis module 12, the calculation module 13, the comparison module 14 and the generation module 15, a comprehensive test and accurate analysis of the chip temperature characteristics are achieved. The acquisition module 11 collects the transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, and generates a characteristic thermal pattern set through feature extraction, laying a data foundation for subsequent analysis; the analysis module 12 uses the characteristic thermal pattern set to perform spatial heat transfer analysis on the steady-state thermal distribution data, accurately obtains the temperature drift deviation data and the temperature gradient evolution sequence, and reveals the heat transfer law inside the chip; the calculation module 13 performs a traceability analysis of the thermal influence factor through the temperature gradient evolution sequence, generates a hotspot contribution matrix, and generates a temperature offset correction factor in combination with an adaptive temperature compensation algorithm to correct the heat transfer error; the comparison module 14 performs a global thermal stability analysis on the chip temperature change data based on the temperature offset correction factor, obtains the dynamic temperature distribution curve, and generates a local temperature rise trend set and a long-term temperature drift curve through multi-scale equilibrium analysis to comprehensively characterize the thermal stability changes of the chip; the generation module 15 uses the long-term temperature drift curve to predict the thermal reliability and thermal degradation of the chip, generates a thermal reliability score and life prediction data, and provides a scientific basis for chip design and optimization. Through modular design and algorithm optimization, this system realizes full-process automated analysis from data acquisition to thermal reliability prediction, effectively improving the accuracy and efficiency of chip thermal characteristics testing.

[0123] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the aforementioned embodiment 1, and will not be repeated here.

[0124] Embodiment 4: like Figure 3 As shown, the present application also provides an electronic device 20, including a memory 21 and a processor 22, the memory 21 stores a computer program that can be run on the processor 22, and the processor 22 implements the chip temperature characteristic testing method of embodiment 1 when executing the computer program.

[0125] In this embodiment, by storing the relevant computer program of the chip temperature characteristic test method in the memory 21 of the electronic device 20 and executing the program on the processor 22, the full process function of the chip temperature characteristic test method described in Example 1 is successfully realized. Specifically, the processor 22 obtains the transient thermal response data and steady-state thermal distribution data of the chip under various workload conditions by executing the computer program stored in the memory 21, performs feature extraction to generate a characteristic thermal mode set, and combines the steady-state and transient data to perform spatial heat transfer analysis, extract key temperature gradient evolution sequence and temperature drift deviation data. Subsequently, the processor 22 uses the above data to perform thermal influence factor tracing analysis and adaptive temperature compensation calculation, generates a hotspot contribution matrix, a temperature offset correction factor and a dynamic temperature distribution curve, and obtains a local temperature rise trend set and a long-term temperature drift curve through a multi-scale analysis method. Finally, the device performs thermal reliability and thermal degradation evaluation on the chip according to the long-term temperature drift curve, generates life prediction data, and provides comprehensive support for the thermal design optimization and reliability verification of the chip. The design of the electronic device 20 enables the chip temperature characteristic testing method to be deployed and applied in an automated and modular manner, thereby significantly improving the convenience and practicality of the testing process.

[0126] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0127] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A chip temperature characteristic testing method, characterized in that: include: Acquire transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, perform feature extraction on the transient thermal response data, and obtain a set of characteristic thermal patterns; Using the characteristic thermal mode set to perform spatial heat transfer analysis on the steady-state thermal distribution data, to obtain temperature drift deviation data and a temperature gradient evolution sequence; Using the temperature gradient evolution sequence to perform a thermal influence factor traceability analysis on the characteristic thermal mode set, obtaining a hotspot contribution matrix, and performing an adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix to generate a temperature offset correction factor; Using the temperature offset correction factor, a global thermal stability analysis is performed on the temperature change data of the chip to obtain a dynamic temperature distribution curve, a multi-scale temperature balance analysis is performed on the dynamic temperature distribution curve to obtain a number of local temperature rise trend sets, and a time series pattern comparison is performed on all the local temperature rise trend sets to obtain a long-term temperature drift curve; The long-term temperature drift curve is evaluated to obtain a thermal reliability score and a thermal degradation prediction parameter, and life prediction data of the chip is generated based on the thermal reliability score and the thermal degradation prediction parameter.

2. The chip temperature characteristic testing method according to claim 1, characterized in that: The step of acquiring transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, and performing feature extraction on the transient thermal response data to obtain a characteristic thermal pattern set comprises: Acquire temperature sampling data of the chip under multiple preset workload modes, perform time series decomposition on the temperature sampling data, and obtain transient thermal response data and steady-state thermal distribution data; Extracting features of the transient thermal response data in time and space dimensions to obtain a local thermal evolution feature set and a global heat transfer feature set; A heat conduction association matrix is ​​constructed based on the local thermal evolution feature set, and a feature clustering analysis is performed on the global heat transfer feature set using the heat conduction association matrix to obtain a feature thermal mode set.

3. The chip temperature characteristic testing method according to claim 2, characterized in that: The step of extracting features of the transient thermal response data in time dimension and space dimension to obtain a local thermal evolution feature set and a global heat transfer feature set comprises: Decomposing the transient thermal response data in time series to obtain thermal fluctuation characteristics in the time dimension and local temperature change characteristics in the space dimension; Extracting a thermal change rate and a temperature change frequency based on the thermal fluctuation characteristics, and generating a transient thermal fluctuation data set based on the thermal change rate and the temperature change frequency; Using the local temperature change characteristics, spatial hot spot analysis is performed on the temperature change data of the chip to obtain hot spot area distribution data and heat source intensity parameters; Performing spatial cluster analysis on the hotspot area distribution data and heat source intensity parameters to obtain a heat source distribution pattern in the hotspot area, and generating a local thermal evolution feature set based on the transient thermal fluctuation data set and the heat source distribution pattern in the hotspot area; A global heat transfer calculation is performed on the local thermal evolution feature set to obtain a global heat transfer feature set.

4. The chip temperature characteristic testing method according to claim 1, characterized in that: The step of using the characteristic thermal mode set to perform spatial heat transfer analysis on the steady-state thermal distribution data to obtain temperature drift deviation data and a temperature gradient evolution sequence includes: Performing spatial heat transfer analysis on the steady-state heat distribution data using the characteristic thermal mode set to obtain a thermal coupling matrix and regional thermal contribution data, and optimizing a steady-state heat transfer path on the steady-state heat distribution data based on the thermal coupling matrix to obtain a heat transfer weight matrix and local thermal conductivity distribution data; Calculating the heat conduction characteristics of the characteristic thermal mode set using the local thermal conductivity distribution data to obtain a thermal impact factor model and a time-evolving thermal feature set, performing time series modeling on the time-evolving thermal feature set to obtain a transient thermal characteristic mapping function and a temperature transition parameter set; Based on the transient thermal characteristic mapping function, the thermal influence factor model is subjected to local thermal field evolution analysis to obtain local temperature change data and thermal inertia correction factor, and the temperature transition parameter set is dynamically compensated by using the thermal inertia correction factor to obtain temperature drift deviation data and corrected thermal gradient data; A trend prediction calculation is performed on the local temperature change data based on the corrected thermal gradient data to obtain a temperature gradient evolution sequence.

5. The chip temperature characteristic testing method according to claim 1, characterized in that: The step of performing a heat influence factor tracing analysis on the characteristic thermal mode set by using the temperature gradient evolution sequence to obtain a hotspot contribution matrix, and performing an adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix to generate a temperature offset correction factor comprises: Using the temperature gradient evolution sequence to perform a heat impact path analysis on the characteristic thermal mode set, obtaining a heat impact factor weight matrix, and calculating the heat impact contribution of the characteristic thermal mode set based on the heat impact factor weight matrix to obtain a hotspot contribution matrix and a characteristic thermal mode offset parameter; Performing deviation matching calculation on the temperature gradient evolution sequence according to the hotspot contribution matrix to obtain a hotspot influence correction factor, and performing nonlinear error correction calculation on the characteristic thermal mode offset parameter based on the hotspot influence correction factor to obtain a hotspot compensation parameter and a temperature offset correction matrix; Applying the temperature offset correction matrix to perform multi-dimensional compensation calculation on the temperature drift deviation data to obtain a dynamic temperature correction factor and a steady-state temperature correction factor, and optimizing the hotspot thermal impact of the hotspot compensation parameters based on the dynamic temperature correction factor to obtain a hotspot temperature compensation factor and a balanced thermal impact factor; The steady-state temperature correction factor is subjected to a steady-state temperature deviation optimization calculation according to the hot spot temperature compensation factor and the balanced heat influence factor to obtain a temperature offset correction factor.

6. The chip temperature characteristic testing method according to claim 1, characterized in that: The steps of using the temperature offset correction factor to perform global thermal stability analysis on the temperature change data of the chip to obtain a dynamic temperature distribution curve, performing multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain a plurality of local temperature rise trend sets, performing time series pattern comparison on all the local temperature rise trend sets to obtain a long-term temperature drift curve include: Using the temperature offset correction factor, a global thermal stability analysis is performed on the temperature change data of the chip to obtain a dynamic temperature distribution curve and local temperature rise data, and a multi-scale time series decomposition is performed on the dynamic temperature distribution curve to obtain a short-term temperature change pattern and a long-term temperature change trend; Performing pattern matching analysis on the local temperature rise data based on the long-term temperature change trend to obtain a local temperature rise trend set, and performing pattern classification calculation on the local temperature rise trend set to obtain a drift trend set; The short-term temperature change pattern is adjusted by using the drift trend set to obtain a long-term temperature drift curve.

7. The chip temperature characteristic testing method according to claim 1, characterized in that: The step of evaluating the long-term temperature drift curve to obtain a thermal reliability score and a thermal degradation prediction parameter, and generating chip life prediction data based on the thermal reliability score and the thermal degradation prediction parameter comprises: Acquiring temperature stability change data and trend shift factor of the long-term temperature drift curve, performing steady-state temperature distribution analysis on the temperature stability change data, and obtaining temperature reliability characteristic parameters and abnormal thermal drift factors; Using the temperature reliability characteristic parameter to perform temperature stability evaluation calculation on the trend shift factor, to obtain a steady-state thermal reliability index and a temperature anomaly prediction parameter, and based on the steady-state thermal reliability index, to perform working stability analysis on the long-term temperature drift curve, to obtain a thermal reliability score and a long-term temperature stability curve; Performing a steady-state thermal change trend evaluation on the temperature anomaly prediction parameter through the thermal reliability score to obtain working stability evaluation data and a thermal degradation parameter set, and performing life decay feature extraction on the long-term temperature stability curve based on the thermal degradation parameter set to obtain thermal degradation prediction parameters and life influencing factors; Performing thermal aging trend calculation on the working stability evaluation data according to the thermal degradation prediction parameters to obtain an initial life degradation trend and a thermal safety margin parameter, and performing life prediction optimization calculation on the initial life degradation trend based on the thermal safety margin parameter to obtain a chip life degradation trend and a temperature safety threshold; The chip life degradation trend and the temperature safety threshold are used to perform life evaluation calculation on the long-term temperature drift curve of the chip to obtain chip life prediction data.

8. A chip temperature characteristic test system, characterized in that: include: An acquisition module is used to acquire transient thermal response data and steady-state thermal distribution data of the chip under different workload conditions, and perform feature extraction on the transient thermal response data to obtain a feature thermal mode set; An analysis module, used to perform spatial heat transfer analysis on the steady-state heat distribution data using the characteristic thermal mode set to obtain temperature drift deviation data and a temperature gradient evolution sequence; A calculation module, used to perform a heat influence factor traceability analysis on the characteristic thermal mode set using the temperature gradient evolution sequence to obtain a hotspot contribution matrix, perform an adaptive temperature compensation calculation on the temperature drift deviation data according to the hotspot contribution matrix, and generate a temperature offset correction factor; A comparison module, used to perform a global thermal stability analysis on the temperature change data of the chip using the temperature offset correction factor to obtain a dynamic temperature distribution curve, perform a multi-scale temperature balance analysis on the dynamic temperature distribution curve to obtain a plurality of local temperature rise trend sets, perform a time series pattern comparison on all the local temperature rise trend sets to obtain a long-term temperature drift curve; A generation module is used to evaluate the long-term temperature drift curve to obtain a thermal reliability score and thermal degradation prediction parameters, and generate chip life prediction data based on the thermal reliability score and the thermal degradation prediction parameters.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the chip temperature characteristic testing method according to any one of claims 1 to 7 when executing the computer program.

Images