Microsoft’s Surface Pro is something of an odd duck. On the one hand, it offers the performance and x86 compatibility that Surface RT lacks. It’s also far faster than Clover Trail-based tablets, which use Intel’s low-power Atom processor. This performance, however, comes at a cost. The Surface Pro’s battery life is substantially worse than Surface RT or Atom-based tablets, it’s significantly heavier (910 grams, compared to 680 grams for the RT) and it’s noticeably thicker than low-power designs.
While Surface Pro isn’t out in the UK yet, it should be available in the very near future. When combined with our recent Surface Pro review, this deep dive into how the Pro version spends its increased TDP, and what the trade-offs and benefits are, will help you decide whether to splash your cash on the device when it does emerge into the UK market.
For this article, I tested Surface Pro against the Samsung Ativ and Surface RT. All three devices were measured using a Watts Up power meter, and all batteries were kept 100 per cent charged at all times. Screen brightness, aside from where mentioned, was kept at maximum. All three devices were set to use Microsoft’s “Balanced” power profile.
First up, there’s the display. Surface Pro has a 1920 x 1080, 10.6in panel, while RT and the Ativ feature 1366 x 768 displays at 10.6in and 11.6in respectively. All three systems were set to Airplane mode with a bright white background loaded. Power measurements were taken at both minimum and maximum brightness.
The Pro draws significantly more power than either of the low-resolution competitors. It’s a bit surprising to see the Samsung Ativ on par with the Surface RT – both models have a maximum brightness of 400 nits, but the Samsung’s screen is larger.
Surface Pro also has a maximum brightness of 400 nits, but its display consumes more at minimum brightness than the others do at maximum.
12 Watts, however, isn’t enough to explain the dramatic battery life difference. Let’s keep looking…
USB file copying
Surface Pro supports USB 3.0, while the Ativ and RT models are limited to USB 2.0. That added performance comes at a price. In the next test I copied an 8GB video file from a 16GB USB 3.0-capable thumb drive. Copy speeds were tested three times, with a reboot between each.
Surface RT’s copy performance is terrible. Throughput is limited to 12.5MB/sec, less than half of Clover Trail’s. RT complements its poor performance with high power draw; its MB/s throughput-to-power consumption ratio is 1.39 compared to the Samsung’s 3.8. Surface Pro is twice as fast as the Ativ, thanks to USB 3.0, but that speed comes at a cost – file copying draws 18 Watts of power, nearly 2.5 times as much as Clover Trail. Surface Pro’s throughput-to-watt ratio is correspondingly lower, at 3.08.
If you’ve been considering a Surface RT but don’t want to re-encode a library of MKV files, we’ve got good news for you. There’s an app available from the Windows Store (PressPlay) that converts MKV files into a format that the tablet can play. The bad news? Files have to be converted every time you view them, there’s no option to save in the converted format, and the playback stutters. Files that played flawlessly via Media Player Classic on Surface Pro and the Samsung Ativ stuttered when opened using PressPlay.
We decided to test the file conversion and power consumption using a 8.75GB MKV file as a baseline. The video output couldn’t play back smoothly on any platform, but benchmarks available across all three platforms are few and far between.
Surface RT performs remarkably poorly here. It draws more power than Clover Trail, and takes more than twice as long to perform the same task. Real compatibility with MKV files remains a pipe dream for RT users, though the appearance of PressPlay gives us hope that a proper solution will present itself.
For these tests, we used a 6.75GB MP4 movie encode (H.264 High@L4.0, 1920 x 800, 7.6Mbps). Before we dive into the results, we need to discuss the importance of playback software. The Ativ comes with PowerDVD’s Cyberlink suite – by default, MKV and MP4 file playback is handled by this utility. It’s not, however, handled very well. The Ativ 500T consumed about 5 Watts while decoding with PowerDVD, but the video stuttered on a regular basis. Installing the K-Lite codec pack and switching to Media Player Classic with hardware acceleration enabled fixed the stuttering problem, but it drew more power.
The Surface Pro draws more power than the other tablets, thanks in part to the high-density screen.
Cinebench is an excellent example of how the more powerful Core i5-3317U CPU can save total battery power by finishing tasks more quickly. Atom’s 4 to 7 Watt power consumption is dwarfed by Surface Pro’s 27 Watts in multi-threaded mode/20 Watts single-threaded, but it takes roughly two weeks to finish the single-threaded scene. System responsiveness during this test is basically nil.
As for device power consumption, the comparison breaks down like this.
Cinebench is also one of the few tests that really push the Core i5 3317U’s power consumption. The Clover Trail tablet can’t run the OpenGL test, so those results are omitted. The Surface Pro’s fan – silent for the majority of our tests – kicks into overdrive on this test and when running Torchlight. The system has no stability problems and the fan is tuned to keep the chip at or below 80 degrees C.
We tested Torchlight on both the Ativ and Surface Pro, but the Ativ wasn’t able to maintain an acceptable frame rate – even with updated video drivers. Surface Pro was tested at 1440 x 900 with details set to minimum levels and at 1920 x 1080 with details and antialiasing turned all the way up.
One of the interesting characteristics of Torchlight is the relationship between frame rate and power consumption. At 1440 x 900 with details set to minimum, Surface Pro maintains a frame rate of nearly 100 frames per second (fps) and draws just 24 Watts. When we increased detail levels to high and set the tablet to 1920 x 1080, fps fell by two thirds while power consumption sharply increased. We saw regular evidence of throttling at this point; power consumption would peak at 33 to 35 Watts, drop back to 24 Watts, then peak again.
Torchlight illustrates how much Surface Pro would benefit from software profiles to control power consumption. At 1440 x 900, I had frames to spare. Trimming CPU power consumption to something in the 16 Watts range, possibly by deactivating Turbo Mode, would’ve significantly extended battery life.
What have we learned?
Surface Pro’s power consumption varies rather dramatically, and in ways that could impact its end-user attractiveness. The 1920 x 1080 display is only part of the problem – shifting the device into Battery Saver mode will extend battery life somewhat, but finer-grained power controls would be useful. A master utility for defining TDP or temperature targets isn’t actually out of the question when you consider that this is, after all, a Microsoft product.
A temporary setback
Relative to the RT, the Surface Pro’s thicker, heavier design makes a lot more sense when you consider the device’s peak power draw. Microsoft deserves credit here – the tablet’s weight and noise (with the fan going) are annoying, but not nearly as annoying as a tablet that allows a 95 degrees C operating temperature or one that becomes unstable.
Taken in aggregate, our performance and power consumption figures show a considerable gap between Clover Trail and the Core i5. If Clover Trail is too little, the Core i5 is often too big. Even after accounting for the high-density screen, Ivy Bridge draws significantly more power.
Multiple competing solutions to this Goldilocks problem will hit the market this year. On the Windows RT side, there are new SoCs from Qualcomm and Nvidia’s Tegra 4, both of which offer substantially improved performance over Tegra 3. Intel’s 22nm Atom is a quad-core design with an out-of-order execution engine, though it’s not scheduled to ship until the end of the year. Haswell should also improve power consumption at the high end of the market, though it’s not clear how large the reduction will be.
Then, of course, there’s AMD. We’ve previously discussed how the upcoming 28nm Kabini/Temash parts could fit neatly into the gap between Clover Trail and Ivy Bridge. After getting some hands-on time with Surface Pro, it’s absolutely clear that we need a solution that can span the gap between them. AMD has previously stated that Kabini is a native quad-core with a 9 Watt TDP. The company claims that its new chip will be 10 to 15 per cent more efficient than Brazos in terms of instructions per clock cycle (IPC) and clock 10 per cent higher (presumably at the same TDP).
AMD’s enthusiast-oriented OverDrive software could end up a star player on Windows 8 tablets. OverDrive is a utility that lets users tweak CPU speeds and define Turbo Mode functionality from within the operating system. In theory, it could be used to define TDP targets for gaming or other “real work” activities.
In this case, it makes a lot of sense to take a wait-and-see approach on the Surface product line, especially if you aren’t thrilled with either model’s trade-offs. New SKUs this year should improve performance and battery life at every level.