According to the Basemark ES 2.0 Taiji GPU benchmark, the most powerful mobile device in the world (when it comes to the number of megapixels rendered per second) is an unknown Pantech device, codenamed QCT MSM8960 EF46 L. The impressive results obtained by this device, combined with the glaring absence of a GPU model number, have lead many to believe that we are actually looking at the first benchmark results of Qualcomm’s upcoming Snapdragon S4 Pro processor. Presumably, the S4 Pro chip swaps the Adreno 225 GPU found in the “standard” S4 for a new and faster Adreno 320 chip.
Bearing a remarkable similarity (in terms of codename) to the QCT MSM8960 EF46S (also known as the Pantech Vega Sky 830), the mysterious Pantech device clearly features a dual-core S4 Krait CPU, but there is no way of knowing what GPU it uses. Some believe that the performance boost could have only been achieved with a new GPU, perhaps the Adreno 320 (planned to go into the Snapdragon S4 Pro). But there are others that think that the Pantech device was running on a lower resolution for the duration of the test (something that drastically alters the results). As clock speeds are also missing, it’s also possible that the CPU was overcklocked beyond the maximum 1.5GHz that S4 chips are usually clocked at. At this point, there is no way to know for sure.
Another interesting (although quite odd as well) detail, is the fact that the HTC One S is reportedly the second fastest mobile device in the world when it comes to rendering video, obtaining results that are far superior to the Tegra 3 Transformer Prime and HTC One X.
I should stop right now and inform you that Basemark’s benchmarks are notorious for favoring Adreno architectures, up to the point where the Samsung Galaxy S2 LTE (which uses the older Adreno 220 GPU) is reported as being faster than the GeFore ULP on the Tegra 3 HTC One X. The general perception is that the Tegra 3 GeForce ULP outperforms even the Ardeno 225 GPU in the S4, so I’ll go out on a limb and say that these results do not reflect real-life performance. So, the least you can do is to take in this info with a nice grain of salt.