How to Benchmark Your Android Device - yearbywartime
How powerful is your phone? The question would have been absurd a decade past, but as we increasingly look connected smartphones and tablets for some of our everyday computing needs, paying more attention to their performance makes sense. Until recently, choosing a smartphone was mainly a matter of aesthetic taste and seaborne OS preference. Today, however, most of the devices available prioritize their display and employ similar slate or candy-bar designs. As a result, performance has become a much more prominent differentiator.
Alas, many mobile device owners are falling into the same trap that desktop PC consumers did years ago: They bet at a list of specifications, project that a smartphone's processor is clocked high or has Sir Thomas More cores, and assume that the gimmick is quicker than one that has what appear to be lower specs. It's as though the MHz myth of the 1990s is dorsum to confuse buyers in a totally new market segment.
The standard SoC (system on a chip) powering one of today's mobile devices, however, is advancing at a rapid yard. A device produced sensible a few months past may have the same numbers of processor cores and be clocked at the cookie-cutter frequency as a blade new device, but its subjacent architecture may embody far to a lesser extent efficient. In improver, the deuce devices' SoCs may be equipt with divers GPUs, different amounts of remembering, and different versions of the mobile OS. Because of these hidden variables, a vendor's list of specs doesn't reveal the whole story. Every bit is echt with desktop and laptop computer PCs, establishing the relation performance of mobile devices requires testing.
Acquiring Started
Since Android's ecosystem is uttermost more different than its mobile OS rivals', we'll focus on testing an Humanoid-based twist here. A handful of the tests that we'll make up covering here are cross-platform or race within a Entanglement browser, meaning that they'll work connected iOS or Windows Phones-based devices–or even on desktops and notebooks–as well.
In evaluating the performance of an Android-based smartphone or tablet (operating room some different modern smartphone OR tab), it's best to entertain the gimmick as a tiny PC. When benchmarking a full-fledged desktop PC, we don't declare that system quicker operating theater slower than another after running a single essa that stresses a particular component. Instead, we run a Host of distinct tests to gauge how well various individual components work, then we weigh the results, and finally we reach a judgment about overall performance. The same approach path is appropriate for Android-settled devices.
A number of bench mark tools are available online and in the Google Play marketplace for examination twist performance. We'll discuss a smattering of free of tools here, but many other benchmarking tools are easy, too. Some (like Vellamo from Qualcomm) are offered aside computer hardware manufacturers; others are created by students. We recommend using legitimate tools whose makers have demonstrated a capacity and willingness to issue updates; ultimately, however, you should shoehorn your benchmark suite to your unique exercis habits. If Web browsing is your highest priority, use apps that test web browser and JavaScript performance. If you spend most of your meter on mobile devices playing games, be sure to run graphics/GPU-centric tests.
Benchmarking Procedures
Eastern Samoa with PC performance examination, there are some regular procedures to observe when moving benchmark tests on Android-based devices. Apps or utilities running in the background on an Android-based twist can sap system resources and intervene with the benchmarks, subsequent in self-contradictory operating room misleading scores that wear't accurately reflect material-world performance. To eliminate this variable, your outdo course is to shut the background apps down. Many Android-settled devices include built-in Task Managers that will turn off uncalled-for desktop apps, thereby freeing improving memory and processor resources. Restart your device, wait few minutes to ensure that everything has drunk completely then use the task manager to kill unneeded apps and maximize uncommitted memory. Expect a couple of moments yearner, and then begin a good test.
Processor, Retentivity, Storage, and Overall Public presentation Tests
Linpack for Android (available at Google Fiddle store, free)
Linpack for Mechanical man is a quick and easy-to-flow bench mark that measures how fast an Humanoid gimmick fundament work out a dense N by N system of linear equations. Linpack is purely a CPU benchmark, and IT can run in either single-thread modal value (to assess the performance of a several CPU core) or multithread mode (to tax all of the on hand processing cores in a device). To start running the tests, you simply launch the app, pick out the single-screw thread or multithread test, and then click the appropriate button generally bill of fare.
AnTuTu Organization Bench mark (available at Google Play entrepot, dislodge)
AnTuTu System Benchmark is an all-in-one benchmark designed to lam tests happening your twist's CPU, GPU, memory, and computer memory. The CPU bench mark evaluates both integer and floating-point carrying into action. The GPU tests assess 2D and 3D performance. The memory test measures available memory bandwidth and latency. And the storage tests judge the read and write speeds of a device's flash memory. To run AnTuTu, swipe to the Test tab, select the tests that you'd like to run, and so detent the Start button. When the tests are stark, the results will appear on the Scores tab.
CF-Bench (available at Google Play store, free)
CF-Bench is another all-in-one benchmark, created by respected Humanoid developer Chainfire (hence the CF). Chainfire originally built the tool for his own use, but a couple of years ago he free information technology to the public. CF-Bench is a CPU, memory, and entrepot benchmark tool around that can exploit the additive computing resources of multicore SoCs; and it produces consistent, quotable scores, patc examination both native and managed Java code performance. To run CF-Terrace, launch the app and click the Full benchmark button. Results will appear at the bottom of the screen, on with a handy comparison to other favourite devices.
Next: GPU/graphics, network, and web browser execution
GPU and Graphics Execution
GLBenchmark (available at Google Play store, free)
GLBenchmark is an OpenGL ES 2.x benchmark composed of 33 different tests that evaluate a GPU's ability to handle texture-based and direct lighting, bump, environment and glow map, soft shadows, and vertex shader-based skinning, among some other things. In addition to its various test cortege, GLBenchmark offers a couple of "off-screen" tests that it renders at the same firmness of purpose (1280 by 720), careless of the gimmick being proven and regardless of any V-Sync limitations. Many mobile nontextual matter benchmarks output results at right around 60 frames per second, because the devices' screens are barred to 60Hz, which renders the results useless. To move GLBenchmark, you moldiness first create an account; but once you've done that, wholly you have to do is launch the app, click the Performance Tests button, choose your tests, and click Start.
An3DBenchXL (available at Google Bring off salt away, free)
An3DBenchXL–a more taxing, updated version of the original AN3DBench–is a free, painless-to-use benchmarking tool based connected an Android port of the free jPCT 3D engine/API. The app runs through a series of five tests that employ different version techniques with varying polygonal shape counts and texture workloads to appraise a GPU's performance; then it spits impermissible a series of results that you can equivalence against the results for other smartphones, tablets or even your Microcomputer and console table. Such comparisons are possible because AN3DBenchXL reports the results of the five individual tests in frames per second (an total scotch is generated as well) so you can date how your device stacks up against other gaming machines. An3DBenchXL couldn't be easier to run, either: Antimonopoly launch the app and and so intercept the screen to start. The results leave appear after all of the tests have been completed. Make a point to follow the benchmarking procedures outlined above to ensure optimal performance.
Network Carrying out
Ookla Speedtest.net (available at Google Play stash awa, free)
Ookla's Speedtest.earning app, an extension service of the popular Speedtest.net website, is designed to evaluate the speed of a mobile device's Cyberspace connector. This is a great tool for illustrating the stark differences in performance that exist 'tween 3G and 4G mobile networks and even Wi-Fi. To run Speedtest.net, launch the app, give IT a hardly a moments to pinpoint your location and choose an appropriate server, so click the big Begin Test button. The screen will measure and report your Download, Upload, and Ping times and will display a real-time graph that tracks connection body.
Browser Performance
Rightware BrowserMark (available to rill on Rightware's website)
As its name suggests, Rightware's BrowserMark is a tool around for measure a browser's JavaScript and HTML rendering performance. BrowserMark runs through an array of animations and JavaScript functions; at the end of the essa it displays a final score, which you pot enjoyment to compare the performance of opposite devices. To run BrowserMark, item your web browser to http://browsermark.rightware.com/ and click the Start button. If you plan to compare the performance of different devices, you should utilisation the same browser for all of your tryout runs. Execution varies greatly from browser to web browser in BrowserMark, which makes this benchmark a good tool for seeing which browser runs best on your gimmick.
SunSpider JavaScript Benchmark (available to keep going WebKit's site)
SunSpider is another web browser-based benchmark. This test, however, strictly tests the performance of JavaScript functions, which are ill-used widely across the Web, excluding Document Object Model (DOM) and other browser Genus Apis. To run SunSpider, point your web browser to www.webkit.org/perf/sunspider/sunspider.html and click the Start SunSpider 0.9.1 now link. The site will conduct a number of tests, and then IT volition display a final result (as well every bit the singular test wads). Along SunSpider JavaScript Benchmark, lower scores indicate better performance.
Next: Benchmark results
Benchmark Results
This lengthy discussion of Android benchmarks would seem discouragingly abstract without some reference data, thus we ran through with all of the tests cited in this article happening a popular Humanoid-founded gimmick, the Samsung Galaxy S II i777 smartphone. The Galaxy S II i777 comes configured with a 1.2GHz dual-core Exnyos C210 SoC, 1GB of RAM, 16GB of internal flash storage, and a 4.3-inch screen that has a resolution of 480 by 800. The Beetleweed S II i777 is likewise designed to use up AT&T's pseudo-4G HSPA+ wireless network (non-LTE), and at this writing it runs Android Gingerbread 2.3.6, though an Android 4.0 Ice Cream Sandwich update–which should significantly bear on performance–is supposedly en route.
Until the comer of the next wave of quad-core devices, or devices supported happening newer ARM operating theatre Intel architectures, the Extragalactic nebula S II i777 qualifies as a high-performing smartphone. If you streamlet these tests on your Android-based gimmick and find that it delivers frown performance on a particular benchmark, don't fret: Near currently available dual-Congress of Racial Equality devices North Korean won't outperform the Galaxy S II. If you bring scores much better than the ones tabulated on a lower floor, rest confident that you have a speedy twist on your hands. Congratulations, and blessed testing!
| Processor | ||
| Linpack for Android v1.2.8 | ||
| Single thread | 57.30 MFlops | |
| Multithread | 76.18 MFlops | |
| AnTuTu System Benchmark 2.7.3 | ||
| CPU integer | 1689 | |
| Central processor floating-point | 1110 | |
| Database Io | 420 | |
| RAM | 891 | |
| SD Add-in write | 125 (12.5 MBps) | |
| SD Card read | 203 (>50 MBps) | |
| Cf.-Bench (native scores) | ||
| MIPS | 1021 | |
| MSFlops | 562 | |
| Memory read | 3117 | |
| Memory board drop a line | 1594 | |
| Disk take | 552 | |
| Harrow Spell | 511 | |
| Whole score | 6432 | |
| GPU | ||
| GLBenchmark v2.1.4 | ||
| GLBenchmark 2.1 United Arab Republic | 44 fps | |
| GLBenchmark 2.1 Pro | 58 fps | |
| An3DBenchXL | ||
| Forked Dragon | 18.91 fps | |
| Heyday Power | 38.07 fps | |
| Ninjas' Garden | 33.36 fps | |
| Emperor's Unused Apparel | 59.4 fps | |
| Witching Island | 24.5 Federal Protective Service | |
| Come score | 30,171 | |
| Web | ||
| Ookla Hurrying Test | Wi-Fi | 4G |
| Ping | 17 Master of Science | 70 MS |
| Download | 20,607 kbps | 6327 kbps |
| Upload | 8254 kbps | 1059 kbps |
| Web browser | ||
| Rightware BrowserMark | ||
| Total (Android browser) | 48,939 | |
| Total (Dolphin HD web browser) | 54,853 | |
| SunSpider JavaScript Benchmark v0.9.1 | ||
| Total | 3497.3 ms | |
| *Trial device: Samsung Galaxy S II mock up i777 | ||
Note: When you purchase something after clicking links in our articles, we may clear a small perpetration. Read our affiliate link policy for more details.
Source: https://www.pcworld.com/article/464686/how_to_benchmark_your_android_device.html
Posted by: yearbywartime.blogspot.com
0 Response to "How to Benchmark Your Android Device - yearbywartime"
Post a Comment