Results: Shield And Tegra 4 Platform
A device like Shield interests me for two reasons. First, there’s the product itself—the quality of its Tegra-optimized games, its ability to sustain a smooth experience while streaming PC content, how hot it gets, how long its battery lasts, and so on. Then there’s the fact that this is the first Tegra 4-based device in our lab.
So, we have two types of benchmark results to present today: those that measure Shield’s performance relative to other tablets and phones, and those that compare Tegra 4 to competing SoCs. Each step of the way I’ll do my best to explain what the results mean and why they’re significant. Let’s start with the platform-oriented tests, then.
MobileXPRT
MobileXPRT, from the folks at Principled Technologies, consists of two test categories: performance and user experience.
The performance suite has five tests, applying photo effects, creating photo collages, creating slideshows, encrypting content, and detecting faces. Results are measured in seconds, and the overall score is generated by taking a geometric mean of the ratio between a calibrated machine (Motorola’s Droid Razr M) and each platform for the five tests.
The outcome isn’t even close. Shield significantly outperforms the Tegra 3 in Google’s previous-gen Nexus 7 and the newer model’s Qualcomm APQ8064. Of course, the two tablets never stood a chance. Terga 4 enjoys a big, thick form factor and active cooling. If this SoC makes it into a tablet, it won’t get those same advantages.
MobileXPRT’s UX suite attempts to quantify smoothness by measuring the frame-per-second performance of scrolling through a list view, a grid of icons/text, flicking through a photo gallery, browsing a webpage, and zooming in and out of a photo. Again, we’re looking at the geomean of five ratios.
Nvidia’s Shield turns in the best score. But as you might expect, there isn’t a ton of difference between the perceived smoothness of mobile devices armed with modern SoCs.
Geekbench
Geekbench runs a series of tests on various parts of a processor. The integer test, for example, includes 12 different tests split between single- and multi-threaded metrics. Based on the result of each benchmark, a score is assigned. Those individual scores are used to calculate a final score in each category, and then an overall Geekbench score.
Tegra 4 rises to the top across the board, again benefiting from Shield’s big chassis and active cooler to run at up to 1.9 GHz. The next-fastest contender, Google’s 2013 Nexus 7, employs Qualcomm’s S4 Pro down-clocked to 1.5 GHz.
AnTuTu
Similarly, AnTuTu 3.3.2 tests a number of different SoC sub-systems, including CPU integer and floating-point performance, memory, 2D and 3D graphics performance, and storage.
Tegra 4’s dual-channel LPDDR3-1800 memory configuration yields a particularly high RAM score (the aggregate 64-bit pathway also excelled in Geekbench, though not to the same extreme). Storage performance is fairly comparable across the board, which we’d expect.
Bear in mind that AnTuTu’s GPU test runs at the device’s native resolution. So, while Qualcomm’s APQ8064 is being benchmarked at 1920x1200, Tegra 4 is rendering to 1280x720. That reflects device performance, as opposed to each SoC’s potential.
PassMark Performance Test Mobile
For some reason, PassMark is crashing on the Qualcomm-powered Nexus 7. Fortunately, there’s an iOS-based version we can use to get an iPhone 5 in the chart. Two numbers stand out most prominently: the Shield’s Disk test and the iPhone 5’s CPU test.
Nvidia had no explanation for why the storage test is scoring so high. Digging into our test notes, we’re seeing reads from flash approaching 1 GB/s, which definitely is not right. We’re also suspect of Apple’s 1.3 GHz dual-core A6 outperforming Tegra 4’s 1.9 GHz quad-core –A15-based SoC.
