Portal field news

Portal field news

in ,

📷 | "Zenfone 8 flip" where triple camera gets up and compact "Zenfone 8"…


"Zenfone 8 flip" where the triple camera gets up and compact "Zenfone 8" ...

If you write the contents roughly
You can see from the benchmark that there is almost no difference between the two machines in terms of performance.

The 6.67-inch smartphone "Zenfone 8 flip" released by ASUS this summer.Before ... → Continue reading

 My Navi News

Mynavi News is a comprehensive information news site with more than 100 categories updated daily. It consists of the latest news and analysis reports and columns from a wide range of genres such as business, digital, lifestyle, and entertainment to niche categories such as management, technology, railways, and cars.

Wikipedia related words

If there is no explanation, there is no corresponding item on Wikipedia.

Performance side


benchmark(British: benchmarking) Is originallySurveyingTo use inWater markIs a word that indicates金融,asset ManagementAnd stockinvestmentIt means the index used for comparison, such as the index stock in. Also widely社会Of thingsシステムThe wayspecificationIt means the level and standard as. AlsoBenchmarkingWhat is an excellent management method (such asbest practice) Refers to the process of analyzing companies with.

Benchmark in surveying

It is used as a standard or reference point for the position and height of buildings and structures. Positions that do not normally move are set as benchmark points, and the positions of buildings and structures are determined based on those points. Metal studs may be driven into the benchmark. In addition, where the benchmark points are always specified in the design document. The term "benchmark" originally means a standard used in surveying.

Benchmark on computer

ComputerIn the field of, refers to an index for measuring the performance of computer system hardware and software. One or moreProgramThe result of executing is called a benchmark score, and it is used as an index showing the relative performance of a certain object. In addition, a special program (benchmarking program) itself for measuring a benchmark score may be called a benchmark. As a well-known benchmarking programHPCMeasure performanceLINPACKAnd measure the performance of database processingTPC,DirectXBy3DCGThe performance of3DMarkand so on.

In addition, for PCs with high required specifications to ensure sufficient operation speed by making full use of hardware 3D functionGame software,online gameSo, mainly for the purpose of checking the operation and advertising, the manufacturer uses benchmark software that uses game materials andTrial versionIs also being produced and distributed.

Benchmarks have different component configurations andア ー キ テ ク チ ャIt provides a means to compare the performance from various viewpoints among the systems that have the above and cannot directly compare the performance according to the specifications.There are also many websites that can aggregate benchmark results for graphing and comparison.Retouch softwareな どVideo card OfBut by the full GPU acceleration techSoftware that has functions that depend on it can be used as a guide for understanding the functions available on your own hardware before installation and for changing to the appropriate equipment.Adobe ThePhotoshopRegarding the minimum requirement of GPU corresponding to, we are guiding you to refer to the site that aggregates the benchmark results[1].

In Japanese, "measuring a benchmark score" is commonly referred to as "taking a benchmark". This is because the practice of recording some value is called "taking a record" or "taking a data". ing.

Benchmark test on computer

As computer architectures have evolved, it has become difficult to compare the performance of various computers simply by looking at their specifications. Therefore, test programs that can operate on different systems have been developed. This is to compare the performance of different architectures by comparing the results of such tests. For example,Intel OfPentium 4Shows the same level of computing performance as a processorAMD OfAthlon XPThe processor has a lower operating frequency (than Pentium 4). In other words, when running the same benchmark test, AMD's low-frequency processor and Intel's high-frequency processor show similar results.

Benchmarks are designed to reproduce a particular load situation. Programs originally developed for benchmarking purposely create such loads. In some cases, a normal application is used as a benchmark, in which case the load of that application is reproduced. The actual performance is better represented by using an application, but if you want to measure the performance of specific components in the system (disk performance, network performance, etc.), a dedicated benchmark program is often used.

MicroprocessorIn designing, the benchmark serves as a measurement tool and an important criterion for determining the part to be circuitized. For example, for an application with a benchmarkalgorithmIf the key part of the is extracted, the benchmark can be considered to be the part that has a great influence on the performance of the application. By running such a small program on the simulator, we can get a clue to improve the performance.

Until around 2000, engineersSPECWas used at the time of design, but SPEC is a very large program and was not suitable for running on a simulator. In 1998, Markus Levy and 12 semiconductor companiesEmbedded Microprocessor Benchmark Consortium, Embedded Processor Benchmark Association)Embedded Microprocessor Benchmark Consortium, Standardized the benchmark for the embedded market. As a result, it became possible to perform measurements with a program smaller than SPEC, focusing on algorithms that have performance problems. Benchmark results measured by semiconductor companies can be published after verification and certification by EEMBC. EEMBC is trying to standardize the power consumption benchmark, as embedded processors will continue to consume less power.


Computer companies have so far presented unrealistic benchmark performance far from actual performance. For example, some compilers of the 1980s replaced certain operations that are commonly found in floating-point benchmarks with semantically equivalent but faster operations. However, speeding up such a small hand is not effective other than the benchmark. Users should be careful when using the benchmark results published by the manufacturer, except for benchmarks that are directly related to the application load they know or some certified benchmark. Benchmarks are a replacement for real-world applications and should only be used when the application is unavailable for some reason. If performance is paramount, use the actual system load as a benchmark. If that is not possible, use a benchmark that is as close as possible to the actual load. If the benchmark is not certified, prudence is important. For example, even if the performance is measured with the same program, it is not impossible for the measurement result to be reversed depending on the input and load conditions.

In the 1980s, when this kind of situation became a problem, CPUア ー キ テ ク チ ャIs conventional16 bitsから32 bitsAt the time of a major shift to, companies were publishing their own benchmarks that they would find convenient. Comparisons in environments with different architectures and OSs were not straightforward, and trial and error continued, involving application vendors and users. By the mid-1990sRISC,VLIWWith the advent of architecture and the simplified structure of hardwarecompilerIt has become clear that technology has a significant impact on performance. Benchmarks are now often used in compiler development. It is used not only to improve the results of benchmark programs, but also to improve the performance of general applications.

Manufacturers still publish only benchmark results in the areas they are good at, but sometimes they call such methods "bench marketing." However, an ideal benchmark should be a substitute for a real application, where the application is not available, or when porting to a particular processor or computer system is difficult. If performance is important, then the application suite in the target environment can be the only benchmark.

Benchmark in programming work

Perform a certain functionProgramIf there are multiple methods when writing, benchmark it to see which method is the most efficient. The benchmark used at this time does not use the ready-made benchmark software introduced in the following sections, but creates a process that can record survey items as shown below, embeds it in the program, executes it, and measures it. ..

In addition,compilerEven if the same source is used, the execution efficiency changes depending on the setting of, so when clarifying this point, benchmark it to see which is the best. In particular,CPUア ー キ テ ク チ ャIf you are writing a program that runs on top of this, this work is well worth doing.

Main survey items:

  • The time it takes to complete a particular process
  • The number of times it was processed within a specific time
  • memoryChanges in consumption
  • diskFrequency of access
  • pagingNumber of times
  • temperatureOf

Benchmark classification

  1. The actual program
    • word processing software
    • User application software
  2. Kernel (basic benchmark)
    • Contains basic code
    • Generally, use the code extracted from the actual program
    • Typical example: Livermore loop
    • linpack benchmark (including basic linear algebra subroutines written in FORTRAN)
    • Results are often expressed in MFLOPS
  3. Simple benchmark
    • Users can write programs to evaluate the basic parts of the computer
  4. Synthetic Benchmark
    • Steps to create a synthetic benchmark
      • Run many applications to get operational statistics
      • Determine the ratio of each operation
      • Create a program based on that ratio
    • Example of synthetic benchmark:
      • whetstone
      • Dhrystone

Prominent benchmark software

  • Dhrystone -Integer arithmetic performance
  • whetstone -Floating point performance
  • SPEC -Overall performance, multiprocessor performance
  • TPC-C-Transaction performance
  • LINPACK -Floating point performance

Benchmark software for personal computers

  • HDBENCH-CPU, RAM, HDD, graphics
  • 3DMark -3D overall performance
  • PCMark -CPU, graphics, RAM, HDD total performance
  • Windows Experience Index
  • PassMark-Comprehensive performance of CPU, graphics, RAM and HDD.AdobeRefers to GPU requirements[1].
  • -CPU, graphics, RAM, HDD total performance
  • CrystalDiskMark -HDD
  • CINEBENCH --CPU (GPU could be measured before R15)
  • Blender Benchmark --CPU, GPU (general purpose arithmetic andRay tracingAccelerator)
  • V-Ray Benchmark --CPU, GPU (general purpose arithmetic andRay tracingAccelerator)
  • -CPU, graphics, RAM, HDD total performance
  • -CPU, graphics, RAM, HDD total performance
  • -Benchmark software that can measure according to the type of parts distributed by AMD.
  • Super π - Yasumasa KanedaDeveloped by professorPiCalculation program. Using this to measure the CPU speed is commonly called "π-baking".

Automotive benchmark

JapaneseAutomotive industryIn most cases, it is used in the sense of a goal to be aimed at rather than an indicator, which is slightly different from the usage in other fields. Therefore, a vehicle with a high degree of perfection is designated as a "benchmark", and when a vehicle exceeding that is announced, that vehicle is newly set as a benchmark, and the standard is frequently changed. However, especially cars that have a high reputation over the years are likely to be benchmarks even if there are better cars,C segmentThen.Volkswagen GolfBut,sports carThen.PorscheBut in handlingBMWBut,Light carThen.Suzuki Wagon RIs often used as a benchmark.

As such, there are few clear definitions for benchmarks that are customarily used in the automotive industry,Car criticIt is an ambiguous thing that is decided by the evaluation of. Disassemble and analyze automobile products of other companies and clarify the mechanism and cost structurereverse engineeringIs being done worldwide. The information that can be obtained by this benchmarking is not a trade secret, so it is basically used by rival companies.Illegal activityis not[2].

Note that there are the following benchmarks in the original meaning of automobiles, but these are not often called benchmarks.


[How to use footnotes]
  1. ^ a b "Frequently Asked Questions about Photoshop Graphic Processor (GPU) Cards". helpx.adobe.com. 2020/12/11Browse.
  2. ^ Our rival company purchased our product, disassembled it, investigated the mechanism (reverse engineering), and developed a product with equivalent performance. Is the act of such a rival company illegal? Are there any measures to protect confidentiality? | Clare Law Office

Related item


Back to Top