System Performance

One of the highlights of the Pixel C is the fact that it is powered by NVIDIA's Tegra X1 SoC. This technically isn't the first time that X1 has shown up in an Android device since it powers the SHIELD Android TV, but it's the first time it has been used in a mobile device. This implementation of Tegra X1 has four Cortex A57 cores with a peak frequency of 1.91GHz, along with four Cortex A53 cores. There are a few things that need to be mentioned about how Tegra X1 differs from other SoCs with eight cores. The first is the fact that in theory, Tegra X1 uses cluster migration rather than Heterogeneous Multi-processing. The reason I said in theory is because the implementation in the Pixel C is really only using the A57 cluster and scaling the frequency as low as 51MHz and as high as 1.91GHz.

In any case, all of our benchmarks should trigger the A57 cores in a mobile device that pairs A53 and A57 cores, so the discussion of the A53 cores being disabled is more relevant from a power consumption standpoint than a performance one. To analyze the general performance of the Pixel C I've run it through our standard JavaScript tests, along with PCMark and BaseMark OS II.

Kraken 1.1 (Chrome/Safari/IE)

Google Octane v2  (Chrome/Safari/IE)

WebXPRT 2015 (Chrome/Safari/IE)

Unfortunately, the Pixel C doesn't make a very strong showing in our JavaScript benchmarks. I can't help but feel that this is the result of a rushed Android build, as there are regressions compared to the Nexus 9 in Octane and Kraken, and in all cases the score just isn't that impressive. JavaScript performance is definitely something Google will have to work on in the future, as the gap between the performance of iOS and Android devices is growing fairly large.

Basemark OS II 2.0 - System

Basemark OS II 2.0 - Memory

Basemark OS II 2.0 - Graphics

Basemark OS II 2.0 - Web

Basemark OS II 2.0 - Overall

The Pixel C only achieves decent results in BaseMark OS II. The NAND memory score definitely drags down the overall rating, with the web score being lower than expected, and the system score being shocking low. The stand out score is in the graphics test, with the Pixel C only coming second to Apple's iPad Pro, and beating out the next best Android device, which is also NVIDIA-powered, by a huge margin. Unfortunately, high as it may be, Tegra X1's graphics performance just isn't enough to carry the Pixel C here.

PCMark - Web Browsing

PCMark - Video Playback

PCMark - Writing

PCMark - Photo Editing

PCMark - Work Performance Overall

PCMark paints a different picture than BaseMark OS II, and highlights how performance can be highly variable based on workload, software, and whether an app is run as a native binary or through ART. In PCMark the Pixel C performs exceptionally well, with the exception of the writing test which seems to have been impacted by some change made to ART on Google's devices when moving to Marshmallow. Putting that aside, we see quite high and often chart topping scores in the remaining tests, resulting in a final score that only sits below the Huawei Mate 8.

Characterizing the Pixel C's performance is difficult. PCMark paints a very good picture, and I'm tempted to lean toward it as it is representative of the tasks that users will perform on a daily basis. There's a bit more discussion to be had about the Pixel C's general performance, and that'll come a bit later in the software section of the review. What I can say is that Tegra X1 certainly is fast as far as raw speed goes, but there's more to be said about the Pixel C's performance when you consider the role of both hardware and software.

Intro and Design System Performance Cont'd and NAND Performance
POST A COMMENT

122 Comments

View All Comments

  • tipoo - Monday, January 25, 2016 - link

    Ars theorized that this was never meant to run Android...Reading this review, it seems to add some plausibility to that.

    http://arstechnica.com/gadgets/2015/12/the-pixel-c...
    Reply
  • tipoo - Monday, January 25, 2016 - link

    Not that ChromeOS would really help on the app front, so Google is in a bit of a pickle. Reply
  • testbug00 - Tuesday, January 26, 2016 - link

    ChromeOS would help a lot on the multitasking and functionality front. Think of it as a high end chromebook that has the ability to be used as a tablet. Reply
  • marcryan - Monday, January 25, 2016 - link

    I just finished returning my Pixel C. I didn't have issues in connecting the device and keyboard nor did I have problems with Internet connectivity. But the major show stopper for me was the touch screen latency. It was insane how many times I had to touch the device to get it to respond. I even tried a stylus on the screen in a drawing program and the pixel was unable to track a straight line without breaks along the way where the screen did not register the input.

    Beyond that there was a supreme amount of waiting time for an app to respond after tapping on it on the home screen. I did a side by side with my Nexus 6P and the nexus (not without its own flaws) performed noticeably faster which is shocking considering the benchmark scores on the pixel C processor.

    There's something fatally flawed in the device which is unfortunate because I really wanted to like it, it's a nice piece of hardware.
    Reply
  • evefavretto - Monday, January 25, 2016 - link

    Another interesting characteristic of the A-series, and by extension, Pixel C' screen is the fact that, once divided in half, the two remaining parts keep the same proportion: 1:√2.
    Probably a design choice for a multitasking feature that was scrapped from Chrome OS and now will see the light of the day in Android.
    Reply
  • Dribble - Monday, January 25, 2016 - link

    Thank you for the NAND performance charts - I have been bitten twice by cheap and nasty NAND killing my tablet's performance after a year or two. I now won't touch one unless I know it's NAND is decent - and pixel C really is pretty borderline considering how high end it's meant to be.

    Agree google has software problems, although I would say it goes beyond missing features. It's clearly far to hard to keep something working as you upgrade. Almost every device I have owned has worked best on whatever version of android it came with. Every upgrade generally introduces problems. I get the feeling they developed the pixel c with android 5 and I bet it worked fine there, then they did an upgrade too 6 and now have all manor of niggles to sort out. Hence I am actually more drawn to devices that don't have major android version upgrades. I'll take an out of date but working version of android over the latest and greatest but slightly broken every time.
    Reply
  • Dobson123 - Monday, January 25, 2016 - link

    There is a mistake in the chart on the first page: The Tegra K1 has 192 shader units, not 128.

    And do you know why they have disabled the A53 cores?
    Reply
  • Kepe - Monday, January 25, 2016 - link

    Judging by the overall unfinishedness of the Pixel C, I'd guess migrating from the small cores to the big ones and vice versa causes some kind of a performance issue Google couldn't solve, as the Tegra X1 doesn't have heterogenous multi-processing.. So there'd be even more lag and performance issues than there is now.
    But perhaps it doesn't really even matter, since the big cores go as low as 51 MHz. It probably wouldn't save much power (if at all) if the little cores were used when there's a low workload. Battery performance seems to be really good even without the little cores.
    Reply
  • Brandon Chester - Monday, January 25, 2016 - link

    Indeed you are right. I've corrected the error.

    As for the cluster migration, just keep in mind that the SHIELD TV doesn't do it either. Granted, that is plugged into the wall so power isn't a big deal, but it's important to note that we haven't see a single implementation where the A53s are used so I wouldn't be quick to blame it on the Pixel C specifically.
    Reply
  • tipoo - Monday, January 25, 2016 - link

    For the SHIELD does it not do switching, or does it turn the A53s off completely? As something plugged into a wall and with overkill cooling for an ARM SoC, it may as well use all 8 all the time, no? Reply

Log in

Don't have an account? Sign up now