Tegra 3's Graphics: Untapped Potential?
Graphics gets the priority here because gaming is one of Nvidia's foci. The company is best known for its work on GPUs, and we've seen some pretty phenomenal demos used to promote Tegra 3. Naturally, rich visuals were expected to be one of Asus' top selling points. And we'll warn you in advance: they are.
Although comparing benchmark results between Android-based tablets is easy enough, comparing dissimilar platforms is more challenging. Benchmarks are only able to yield accurate measurements when they're equally optimized for the environments under which they run.
As you can see in the following table, a few of the Transformer Prime's scores come close to the iPad 2 and 3. But for the most part, though, the numbers suggest that Nvidia trails its competition by a significant margin.
| GLBenchmark 2.1.2*native resolution | iPad 2 | iPad 3 | Kindle Fire | LePan II | Transformer Prime |
|---|---|---|---|---|---|
| GPU | PowerVR SGX543MP2 | PowerVR SGX543MP4 | PowerVRSGX540 | Adreno 220 | Tegra 3 (ULP GeForce) |
| Egypt Standard* | 6661 frames (59 FPS) | 6709 frames (59 FPS) | 2847 frames (25 FPS) | 3485 frames (31 FPS) | 5388 frames (48 FPS) |
| Egypt Fixed Time Step* | 47.598 s (59 FPS) | 47.261 s (60 FPS) | 116.138 s | 112.659 s (25 FPS) | 69.055 s (41 FPS) |
| Egypt Off Screen (720p) | 10 146 frames (90 FPS) | 15 663 frames (139 FPS) | 2516 frames (22 FPS) | 3603 frames (32 FPS) | 6496 (58 FPS) |
| Pro Standard* | 2962 frames (59 FPS) | 2975 frames (60 FPS) | 1981 frames (39 FPS) | 2140 frames (43 FPS) | 2726 frames (54 FPS) |
| Pro Fixed Time Step* | 20.868 s (60 FPS) | 20.857 s (60 FPS) | 34.658 s | 31.982 s (39 FPS) | 24.189 s (52 FPS) |
| Pro Off Screen (720p) | 7352 frames (147 FPS) | 12546 frames (251 FPS) | 2277 frames (46 FPS) | 2332 frames (47 FPS) | 3717 frames (74 FPS) |
Because Apple only uses its own SoC in the iPad, we're forced to compare A5X under iOS to Tegra 3 in an Android-based environment. This means we have to trust that GLBenchmark's developers wrote code well-optimized for both platforms. Otherwise, the synthetic turns back unrepresentative results. Assuming GLBenchmark does run equally well on both architectures, Tegra 3 is quite a ways behind the A5’s SGX543MP2 and the A5X’s SGX543MP4. That's an oversimplification, though.
Nvidia’s ULP GeForce is based on technology derived from the company’s experience with desktop and professional GPUs, wherein power savings often take a backseat to performance. Although Nvidia successfully adapted its design to the embedded space, GeForce is still fundamentally different from PowerVR, Mali, and Adreno.
In order to push a performance-oriented piece of hardware, though, you need a demanding workload. Today's smartphone- and tablet-oriented games generally cannot be characterized as demanding. They're intended to run on a broad ecosystem of hardware. So, when it comes time to benchmark Tegra 3 against its competition, Nvidia argues that today's tests can't really tap into the platform's capabilities.
Since GLBenchmark doesn't give us any way to adjust texture quality, we can't rule out the possibility that a more taxing test might swing the scales in favor of Nvidia's design. Lacking better means to quantify performance, we're left to lean on real-world experiences in order to create more telling comparisons.
When we actually play through titles optimized for Tegra 3, the results are are impressive. Frame rate didn't seem to change after switching from the basic and advanced graphics modes (indicating that there is no graphics hardware bottleneck). However, the change in graphics quality is indisputable. Riptide, pictured above, is clearly written to show off Nvidia's hardware.
In the video, we have Riptide running side-by-side on two Transformer Prime tablets. On the left, the graphics are in basic mode, while the right tablet has all of the bells and whistles enabled. As you can see, additional effects do not impact the frame rate at all, and both systems run smoothly.
