NVIDIA Tegra Architecture

I'd like to stand corrected but I doubt that even the A7's theoretical peak SoC consumption would be at "just" 8W. The graph you're linking to shows a max power a tad lower than 12W for the A7 so I wouldn't be at all surprised if the A7 TDP is even higher than 12W.

 

That 12W is including the display and everything. Also, the article mentions that this is done using a test that pretty much resembles a power virus. It says that power consumption for GFXBench 2.7’s T-Rex HD Offscreen results in roughly 6 W and a game like Infinity Blade 3 is roughly 5 W (All including screen and everything).

 

First i want to congratulate you on your anandtech employment and also to say that the handfull of articles i have read of yours have been informative and in the spirit of the site, keep it up .Secondly following on from my post in the other thread can i suggest that you guys look into using a fraps like app for your soc analysis ? There are two such apps for android fps meter and gamebench.

 

Hmm, there have been a number of hands on and early verdicts but no word on battery life?

 

Posted above

 

A second guy with a N9 review unit adds his thoughts to the discussion.

 

Sounds very underwhelming so far.

 

Not terribly surprising. Also interesting to note in that, more comments about how hot the device gets, likely due to the Tegra SOC.

A Maxwell enabled device can't come soon enough for Nvidia. That should hopefully make the Tegra SOC less power hungry and thus easier to cool and potentially more attractive to more device makers.

Regards,
SB

 

I'm not sure how much difference Maxwell would make here. The few reports we have thus far seem to indicate that it's the CPU that's running hot. And besides, if it was the GPU then we would have had similar reports from A15 based Tegra K1 devices. From what I know, which isn't much, the Tegra SHIELD Tablet seems to run reasonably okay.

 

So far we are aware only of a 4*A53+4*A57 big.LITTLE Erista SoC with a Maxwell GPU; IF the problem is the CPU in Flounder and if there won't be a Denver based Erista variant, the problem is practically solved.

 

Nexus 9 battery life testing from Phone Arena:

"Would it not have been for the Sony Xperia Z3 Tablet Compact's excellent endurance, Google's Nexus 9 slate would have ranked on top of our battery benchmark for tablets with its excellent score of 9 hours and 24 minutes. As it stands, however, the Nexus 9 is still 23 minutes shy of matching the Compact, and that's with it sporting a considerably larger battery (6,700 mAh versus 4,500 mAh). One could speculate that this has to do with the power efficiency of the dual-core, 64-bit Nvidia Tegra K1 Denver chipset on board the Nexus 9, but that could obviously be just part of the reason.

Putting those numbers into words, you should expect two days of usage with the Nexus 9, so long as you don't go down on it crazy hard, of course. We know because of the script we use to test battery life – it replicates the average user's habits and arrives at a final figure that is indicative of the hands-on time you can squeeze out of it on average.

The Nexus 9's score puts it in front of competitors like the Apple iPad Air 2 (7 hours 27 minutes), the Samsung Galaxy Tab S 10.5 (7 hours 2 minutes), and even the enormous Samsung Galaxy NotePRO 12.2 (8 hours 58 minutes). With a charging time of 254 minutes, the Nexus 9 is also one of the fastest tablet devices to recharge from zero to full on our list – behind only the Apple iPad mini 2, which has a battery with similar capacity (6,471 mAh). "

http://www.phonearena.com/news/The-...ngest-lasting-tablet-weve-ever-tested_id62408

Note that the Xperia Z3 tablet has a lower screen resolution, smaller screen, and lower performance SoC in comparison to Nexus 9, so the platform power consumption is clearly lower in comparison.

 

Full Phone Arena Nexus 9 review here: http://www.phonearena.com/reviews/Google-Nexus-9-Review_id3842

They had very good things to say regarding screen quality and color accuracy, performance, battery life, and charging time.

The only oddity is the Sunspider score. Perhaps this is one benchmark where the dynamic code optimization process is not effective due to lack of ability to reuse or reorder code. In the previews I have seen so far, reviewers claim that web browsing on Nexus 9 is very fast and subjectively just as fast as iPad Air 2, so clearly the real world web browser performance appears to be excellent running Android "L" OS and Chrome browser on this device.

 

New Chrome versions aren't very fast in SunSpider for some reason. I got ~1000 ms on my Galaxy Note 4 with Exynos 5433's 1.9 Ghz CortexA57, that's very odd since it's possible to reach 380 ms in SunSpider even on 1.7 Ghz Cortex A15 - http://images.anandtech.com/graphs/graph7439/59147.png

 

I would wait for proper reviews on release firmware in a few days. The embargo also only lifts later in the day.

 

Which version of Sunspider did you use the old 0.9.1 or the newer 1.x, the newer version produces 'slower' scores?

 

It looks like GFXBench driver overhead scores for Nexus 9 with Android "L" OS have dramatically improved vs. other Android devices that are using older OS's, and the driver overhead score is for the first time virtually the same as iOS devices such as iPad:

http://gfxbench.com/subtest_results_of_device.jsp?D=Google+Nexus+9&id=553&benchmark=gfx30e

 

I've used all versions, 0.9.1 run time is 987.3 ms, 1.0.2 run time is 950.5 ms at 90% battery

 

Kraken score for Chrome is 6555.1 ms
Default browser scores are much better though, SunSpider 1.0.2 score - 409.4 ms, Kraken score is 4178.2 ms

 

Samsung seems to optimize their browsers a lot. What is the browser that Samsung uses based on? Because I think that recent versions of android don't include a browser other than Chrome. And WebView of course. Are they using WebView? That won't work anymore in Lollipop though, as WebView will become part of the Google Play Services package.

BTW, Samsung's browser optimizations sometimes make me wonder what kinds of sacrifices they make to get the much higher benchmark scores compared to Google Chrome.

 

None. They just target compile and optimize for that specific CPU arch and GPU that the device ships on. And it's just their own browser based on AOSP.

BTW, please install this and see if it makes a difference: https://play.google.com/store/apps/details?id=com.chrome.deviceextras.samsung&hl=en