Benchmarks
I haven't posted any browser benchmark since more than four years. For a good reason: if all contenders perform equally well, there's no need to benchmark them.
Of course, the recent excitement about Apple's Safari outpacing Chrome and Firefox still came to my attention. ;) As it turned out, however, Safari managed to do that only in benchmarks developed by Apple, but not in those provided by its competitors Google and Mozilla. This result seems to confirm qualified opinions as to which the available browser benchmarks are to be disregarded altogether.
Well, let's see what we have:
Apple: Jetstream and Speedometer Google: Octane Mozilla: Kraken
Now let's see what we've got:
| System | Jetstream | Speedometer | Octane | Kraken |
|---|---|---|---|---|
| 5: Office (i7 4790, Archlinux) | ||||
| Chromium 44.0.2403.155 | 225.72 ± 7.14 | 95.7 ± 2.0 | 41571 | 749.8 ms ± 0.6% |
| Firefox 40.0.2 | 215.21 ± 8.08 | 61.0 ± 2.1 | 35537 | 839.5 ms ± 2.4% |
| 4: Desktop (Xeon E3 1240 v2, Archlinux) | ||||
| Chromium 44.0.2403.155 | 180.72 ± 7.66 | 53.7 ± 0.35 | 32280 | 919.6 ms ± 1.7% |
| Firefox 40.0.2 | 164.03 ± 1.44 | 29132 | 1030.3 ms ± 4.1% | |
| 3: Notebook (Pentium P6200, Archlinux) | ||||
| Chromium 44.0.2403.155 | 82.44 ± 1.56 | 22.8 ± 0.34 | 13167 | 2138.6 ms ± 1.6% |
| Firefox 40.0.2 | 70.53 ± 5.91 | 11944 | 2300.4 ms ± 2.8% | |
| 2: Netbook (Atom N270, Debian Stretch) | ||||
| Chromium 44.0.2403.107 | 17.28 ± 1.37 | 5.15 ± 0.07 | 2924 | 12985.7 ms ± 3.7% |
| 1: Tablet (ARM Cortex A9, Android 5.1.1) | ||||
| Chrome 44.0.2403.133 | 15.46 ± 0.92 | 7.64 ± 0.19 | 2776 | 16248.3 ms ± 5.1% |
Chromium consistently performs better than Firefox across all benchmarks, even in Mozilla's own benchmark Kraken. The difference, however, is insignificant.
But that's not what I was actually interested in. What I really wanted to see was whether one can abuse these browser benchmarks for a kind of quick and dirty system benchmarking without the need to install anything. And as you see, all benchmarks scale fairly well:
If we restrict ourselves to x86 for the moment, Jetstream (red) and Octane (blue) scale essentially identically across the entire range of systems. As a matter of fact, they even seem way too close if we suppose that Jetstream and octane are independent benchmarks. g Kraken (yellow) scales very similarly, with the sole exception of the mini, for which it indicates only half of the performance than all other benchmarks. Finally, Speedometer (green) really seems to like the i4790. Perhaps it's making use of AVX2?
For the ARM architecture of the tablet, Jetstream and Octane are again almost identical, but Kraken suffers and Speedometer gains. No big deal, though: the notebook is still miles away. What's also interesting: a tablet from 2012 does not outperform a netbook from 2008, contrary to what the media want us to believe. But who believes them anyway anything anymore.
All in all, I view these benchmarks as a convenient means to test a system's performances in one of it's currently most important disciplines. In other words, it's relevant, meaningful, simple and quick. I use it when people in the office complain that their computer is too slow. Guess what's slow?