Like educational experts who complain when instructors “teach to the test,” to create positive report numbers rather than smarter kids, Android-watchers are angry with Samsung for doing much the same thing with its hardware this week.
A post on the Beyond3D forums prompted esteemed hardware blog AnandTech to investigate claims that Exynos Octa-based Samsung devices – including some models of the Galaxy S4 – ramp up their GPU and CPU substantially when they detect that one of a number of well-known benchmarks is being run, in order to provide more eye-catching statistics.
[MORE ANDROID: Samsung edges Apple in U.S. smartphone satisfaction survey]
[Editor’s note: The biggest Android news of the week should be Google’s Moto X smartphone, which we’ve written about a number of times, but that is being formally announced later today (Thursday, Aug. 1).]
And it turns out that, in fact, such is the case. AnandTech even found a string called “BenchmarkBooster” in a system file, apparently designed to max out performance on common test suites. Uh oh.
The corporate response was a pretty uninspired example of the ostrich-in-the-sand school of public relations - “maximum GPU frequences for the Galaxy S4 have been varied to provide optimal user experience for our customers, and were not intended to improve certain benchmark results.” It's particularly unconvincing when you consider that many of those certain benchmarks are actually referred to by name in Samsung's code, as uncovered by AnandTech.
Essentially, Samsung's cooking the books on its hardware. The overclocking prompted by “BenchmarkBooster” isn't triggered by most other apps – like games and other stuff that people would actually want high performance for – so the full horsepower of the Octa-powered S4 and its ilk are reserved only for making Samsung look good, not for actually providing a better user experience. (The only user-facing apps that do get the S4's full capacity Samsung's own pre-installed software, like S Browser.)
As AnandTech points out, this is further evidence of why users shouldn't rely too heavily on benchmarks to gauge performance – but it's also evidence of Samsung going out of its way to muddy the waters. Like it or not, benchmarks are often a useful, semi-objective measurement of performance, and gaming the system makes everybody's life – including those of Samsung's consumers – just that much more complicated.
If you were struggling to put your finger on why Google's new Chromecast media streamer seemed more like an Android device than the name might suggest, here's another Androidian fact about it – it's already been rooted, thanks to the people of GTV Hacker. (These helpful hackers also state explicitly that the device's software is Android-based.)
It's a promising development for the media dongle that's already made such a big splash, but XDA Developers reports that there isn't a lot you can actually do with a rooted Chromecast yet . Still, the tinkering can begin.
A survey from the American Customer Satisfaction Index shows the Galaxy S3 and Galaxy Note 2 both outperforming the iPhone 5, 4S and 4 among U.S. consumers by two to three points on a 100-point scale.
Interestingly, however, consumers in Samsung's home territory of South Korea actually preferred the iPhone 5 to the Galaxy S3, the researchers noted. Familiarity breeding a little contempt?
Finally, we go back to AnandTech for the revelation that Android 4.3 includes support for TRIM, a feature that cleans up blocks of data on solid-state storage after they've been deleted.
Brian Klug explains that, since solid-state storage never actually deletes anything through software, “deleted” files are simply unindexed, but still present.
“Let's say you copy a 3GB movie to your internal storage, watch the movie and later delete it. You'd have 3GB free to re-use, but until you re-write those blocks the eMMC controller would treat all 3GB as valid data. There's a data structure used by the eMMC controller that tracks mapping logical locations to physical locations in NAND. I won't go into great detail here but the more complex that mapping becomes, and the more locations that have to be tracked, the slower internal NAND management works,” he wrote.
Clearing the disk I/O bottleneck caused by un-TRIMmed data should extend the performance life of many old devices. Here's hoping that saves at least a few people some money down the road.
Email Jon Gold at firstname.lastname@example.org and follow him on Twitter at @NWWJonGold.