El Reg reports
The NRDC [Natural Resources Defense Council] reckons TV makers are configuring sets to perform well on government tests, while in the living room they become energy hogs.
Its specific claims are:
- The TVs perform well on the US Department of Energy-mandated energy use test--but that's based on a clip that doesn't match real-world video content. ([To El Reg,] that seems like a slip-up by the DoE);
- TVs from Samsung, LG, and Vizio are designed to disable energy-saving features if the user changes their screen settings, but there's little or no warning about this. This, the NRDC says, can as much as double the power consumption; and
- UHD TVs turn into energy hogs when they're playing high dynamic range (HDR) content, but HDR isn't included in the DoE's test (again, surely that means the DoE needs to update its tests?).
The NRDC says European testing seemed to match another observation it made: that during the DOE test loop, some TVs seemed to exhibit "inexplicable and sustained drops in energy use". It suggests that software is specifically detecting the test loop and adjusting the TV's performance to suit.
One assumes that "a clip" refers to the standard video loop used in the tests.
(Score: 2) by choose another one on Saturday September 24 2016, @09:13AM
No, this looks like exactly the same thing as VW (and possibly other car makers - in real-world testing VW diesels actually are some of the cleanest, we know how VW cheated, we just don't know how the other guys passed the tests).
The "test loop" in question is a test piece of video that is looped, and it is known content, fixed as part of the test. The TVs appear to drop energy consumption when playing this particular test video loop, at the moment we don't know why, but it looks exactly like the VW situation before we knew how they were cheating - the cars ran suspiciously cleaner when running the (known, fixed) test cycle, the TVs run suspiciously greener when playing the (known, fixed) test content. I don't see any way this can be a design feature other than designing to defeat the test.
(Score: 0) by Anonymous Coward on Saturday September 24 2016, @02:48PM
It depends upon what the test loop actually is. If it is running a pattern that is easy to render than what would typically be run, then it simply might be the test. If all the manufacturers seem to run with less power when running the test loop, then they could all be cheating, or it might be the test loop itself. That's why it is different. If only VW diesels did better on the tests as compared to "real life", that is suggestive of cheating, but if all diesels tested showed better performance during testing, then that looks like a test scenario issue.
(Score: 2) by tonyPick on Monday September 26 2016, @06:37AM
Detecting a specific video sequence is a lot harder than picking out the vehicle emissions testing that VW did...
And From the NRDC Article:
So - since the test scene has short scenes and frequent cuts, this clip trips this feature more and does better on the power consumption tests than content which does not. This sounds (to me) like a problem with the test itself being non-representative: the MDD feature is something that's always there, and looks like a valid power saving feature...
(Score: 2) by choose another one on Tuesday September 27 2016, @09:19AM
That explanation is from the article itself, which also says that there is a further _unexplained_ power usage drop when playing the test clip.
Detecting a specific test video clip will be well within the capabilities of todays TVs, and this sort of thing is not new - video drivers have been found detecting specific benchmark runs for years.
(Score: 2) by tonyPick on Wednesday September 28 2016, @06:26AM
Really? I didn't get that from TFA - I saw The analysis showed higher energy use with the real-world clip than the one used in DOE testing. which I though meant a clip content + MDD accounted for it, and the rest looked to be obfuscation in the way the settings fell out, but I may have missed something there...
Well, it's certainly theoretically possible, but smart TV's are very resource constrained devices and clip identification isn't in the normal functional envelope, so unless the test clip is specifically flagged somehow at the TMDS or data layer or is somehow statistically identifiable it strikes me as unlikely they're doing direct detection of the test clip itself, which some posters here have mentioned.
(and even the dodgy video drivers I've seen reports on were using API call sequences and/or binary names, rather than direct video detection, but it's been awhile since I followed that sort of thing).