BenchmarkXPRT Blog banner

Author Archives: Justin Greene

Sleep studies

Last week, we discussed the increasing complexity of power options in Android 6.0. Features such as Doze and App Standby have changed the way that the operating system manages app activity, and the wide array of UI skins used by many vendors ensures that the steps needed for pre-test configuration differ considerably from device to device.

Managing Android’s Doze feature is critical to getting a good BatteryXPRT score. To show how involved this process can be, we thought it might be helpful to present the steps for one device. Below my sig are the configuration steps we used for the Huawei Mate 8, which we recently featured in the XPRT Weekly Tech Spotlight. For other phones we’ve tested, the steps have been quite different. We’re working on distilling our experience for our tips and tricks document, and the updated version of the document will be available soon. If you have any useful tips, please let us know.

Justin

Whitelist BatteryXPRT (there are two ways to do this)
1) Access Battery manager from Settings/Advanced settings or from the Phone Manager app on the home screen.
2) Select Protected apps.
3) Use the toggles beside BatteryXPRT and BatteryXPRT Tests to allow them to keep running after the screen turns off.

Configure sleep settings
1) Open Settings from the home screen.
2) Select Display.
3) Select Sleep.
4) Select Never. This may reset to a default setting on its own. In our case, it reset to 10 minutes.

Configure screen lock settings
1) Open Settings from the home screen.
2) Select Advanced settings.
3) Select Security.
4) Scroll to the bottom of the list and use the toggle to turn off Screen lock. This keeps the device screen from locking after standby periods during the test.

Focusing the spotlight

As you may have heard, the Samsung Galaxy S7 is the XPRT Weekly Tech Spotlight this week.  As we were testing it, we noticed that our WebXPRT scores were about 8 percent lower than those reported by AnandTech.

The folks at AnandTech do a good job on their reviews, so we wanted to understand the discrepancy in scores. The S7 comes in a couple of models, so we started by verifying that our model was the same as theirs. It was.

The next step was to check their configuration against ours, and this is where we found the difference. Both phones were running the same version of Android, but the S7 AnandTech tested used Chrome 48 while the S7 we tested came preloaded with Chrome 49. In our testing, we’ve noticed that upgrading from Chrome 48 to Chrome 49 has a noticeable performance impact on certain devices. On the Samsung Galaxy S6, the scores went down about 10 percent. In all cases we’ve seen, the decrease is driven largely by the Stock Option Pricing workload.

This isn’t the first time we’ve written about browser versions affecting results. WebXPRT is a browsing benchmark, and the browser has a legitimate impact on performance. When you’re comparing results, it’s always important to look at all the factors involved.

Justin

Last week in the XPRTs

We published the XPRT Weekly Tech Spotlight on the Samsung Galaxy S7.
We added two new BatteryXPRT ’14 results.
We added one new MobileXPRT ’15 result.
We added four new WebXPRT ’15 results.

Detective work

A few weeks ago, we described an issue with TouchXPRT’s Create Slideshow test. On some systems, this test would take longer than expected to complete due to the output video not rendering correctly. Since then, we released TouchXPRT 2016 CP2, which includes a check that verifies the output and reports an error if there’s a problem.

It took a lot of detective work, but I’m happy to say that we’ve been able to resolve this issue. We had to make a couple of changes. First, we changed the pixel format in which the photo content is read from GUID_WICPixelFormat32bppPBGRA to GUID_WICPixelFormat32bppBGR. Second, we changed the media format of the input into the video converter from MFVideoFormat_ARGB32 to MFVideoFormat_RGB32.

After making these changes, the Create Slideshow test ran in the expected amount of time, and rendered the output video correctly.

In our testing so far, results have not changed noticeably. However, we’re continuing to test. We want to be sure that the updated build is stable and that results are comparable to earlier TouchXPRT CPs. We’ll then release it to the community.

Here’s looking forward to the release of TouchXPRT 2016 to the world at large!

The XPRTs at IDF15 Shenzhen

We recently traveled to Shenzhen, China to anchor Principled Technologies’ booth in the IDF15 Shenzhen technology showcase. Over 60 companies set up displays to interact with approximately 2,000 attendees, and it was a great opportunity for us to connect with innovators in one of the fastest-growing technology hubs in the world.

We spent most of our time talking with people about the benefits of the BenchmarkXPRT Development Community, demonstrating the XPRTs on our display systems, and describing the scope and abilities of each XPRT. Many of the people we talked with showed great interest in the XPRTs’ commitment to developing easy-to-use benchmarks that measure performance and battery life while doing everyday tasks.

Of course, we also spent a good deal of free time exploring Shenzhen and nearby Hong Kong. If you ever get the opportunity to visit either of these dynamic cities, you won’t be disappointed!

Technology showcase overviewThe IDF15 Shenzhen technology showcase floor. Our booth is just to the right of the gray-and-white balloon. Shenzhen towersThe booming Shenzhen skyline, including the soon-to-be-second-tallest building in the world.

If we missed you at IDF15, or you have questions or comments about the Benchmark XPRT Development Community, feel free to contact us.

Justin

Activity all around

It’s a busy time in XPRT land, and we wanted to share some news about upcoming events.

The first bit is an update about the CrXPRT 2014 Community Preview. We’ve been busy refining the benchmark, and feel like the current candidate build under test is likely to ship as our Community Preview. We’ll publish more details about the app and its workloads soon, but don’t be surprised to see a Community Preview announcement in the near future.

In Patching and future proofing, we discussed two potential problems while running MobileXPRT 2013 on some versions of Android 4.4 and the developer’s preview of Android L. The first problem involves the benchmark not showing scores when running on a Nexus 5 with the ART runtime enabled. The second problem was a failure of the Create Slideshow workload on Android L. In order to address the underlying cause of both problems, we’ll soon be issuing a patched MobileXPRT 2013 build. We’re still wrapping up our testing, but expect to release the patch at some point next week.

We’ll also be posting BatteryXPRT 2014 for Android APKs directly on the BatteryXPRT.com site. For users who have low bandwidth or trouble accessing the Google Play store, these APKs will make it easier to download the benchmark.

If you have any questions or concerns, please don’t hesitate to contact us at BenchmarkXPRTsupport@principledtechnologies.com.

Justin

Comment on this post in the forums

Patching and future proofing

We recently became aware of two potential issues with running MobileXPRT 2013 on versions of Android 4.4 (KitKat) and a pre-release version of Android L. The first involves a reported failure of MobileXPRT 2013 to display scores when running on a Nexus 5 running KitKat with the ART runtime enabled. ART is an experimental Android Runtime available on a number of mobile devices, and offers several features not found in Android’s current default runtime, Dalvik. We have not been able to reproduce this specific problem, and have successfully tested MobileXPRT on a Nexus 5 running Android 4.4.3 with ART enabled. If you encounter this problem, contact us at BenchmarkXPRTsupport@principledtechnologies.com.

The second issue is specific to the pre-release version of Android L. On that OS, we have seen a failure of the MobileXPRT Create Slideshow workload. We found that while previous versions of Android used specific addresses when loading libraries, this pre-release version of Android L loads libraries at different addresses as a security precaution. This appeared to be causing the workload failure. Recompiling with the –pie flag, so as to be position independent, seems to fix the problem.

As soon as thorough testing is complete, we’ll be releasing a patched build on MobileXPRT.com, and expect no problems running MobileXPRT on Android L.

If you have any questions about MobileXPRT or any of the other benchmarks, feel free to send us a message.

Justin

Comment on this post in the forums

Check out the other XPRTs:

Forgot your password?