JSBench doesn’t take advantage of more than two cores and seems insensitive to CPU clock rate, its performance primarily dependent on CPU architecture. Apple’s A7 posts the best time followed by the three SoCs with -A15 cores. Krait doesn’t handle this benchmark well, posting the two worst times.
The one anomaly is Tegra K1, completing the benchmark 20% faster than both the Tegra 4 and Exynos 5 Octa. It’s tempting to ascribe this result to K1’s frequency advantage, which is about the same percent difference. However, none of the other SoCs sharing the same CPU architecture see any benefit from frequency scaling or additional memory bandwidth. All three SoCs also use the same 2MB L2 cache. Tegra K1 does use the newer revision -A15 cores (r3p3), but looking through the ARM revision history only reveals “engineering errata fixes” and some additional clock gating features introduced in r3p0. There's nothing that would explain a 20% performance improvement.
We see nearly the exact same trend in Peacekeeper. The iPad Air holds a 24% advantage over the Shield Tablet, which is 23% faster than the other two SoCs with -A15 cores.
Principled Technologies' WebXPRT 2013 is an HTML5-based benchmark that simulates common productivity tasks traditionally handled by locally-installed applications, including photo editing, financial charting and offline note-taking.
Tegra K1 isn’t able to break the A7s dominance in our browser benchmarks, only achieving 81% of its high score. The Exynos 5 Octa matches Tegra K1s performance, but Tegra 4 trails behind. Curiously, Qualcomm’s highest-clocked Snapdragon, the 805, comes in last, trailing the lower-clocked Snapdragon 800 in the Samsung Galaxy Note 10.1 LTE.