If the GeekBench score of 1600 for the iPhone 5 is to be believed, it’s somehow as fast as Android phones with twice as many cores an similar (if not higher) clock speeds. If you take a look into the results, you can see the Quad Core 1.4 GHz S III gets higher Integer and Floating Point scores (the Arithmetic tests) sort of. The aggregate 2857 vs the 2051 for floating point performance isn’t as wide as it should be for heavily parallelized tests. At a higher clock and twice as many cores, the S III should be at LEAST 2x the iPhone 5’s scores, not ALMOST 2X. It’s hard to believe that Geekbench would only test two threads but I won’t write off the possibility.
Where it gets really weird is the Memory and Stream performance comparisons. All the S II benchmarks have a bizarrely low sequential read score compared to their sequential write score. These results are essentially worthless unless both devices were “fresh off a reboot”.
If things continue to pan out this way, Apple has done some serious optimization that the Java based Android can never hope to do, even if it had custom chips.
Perhaps more relevant news though, is that the Mirrored Drive Doors PowerMac G4 (dual 1.4GHz) has an aggregate score of 1234 and is the highest scoring G4 based machine. The 2005 2.5GHz Quad G5 is still a long ways off at 3319 but it’s worth noting that that machine wasn’t passed by Intel MacBooks for a while depending on model and neither of my Core2Duo Minis beat it. Considering that iPhones are doing a good job of obeying Moore’s law and doubling their score with each iteration, the last G5 should be scared, very scared.
Since there’s no need to push pixel densities higher than 300ppi (or less, depending on viewing distance), we’re getting closer to certain specs no longer needing upgrading. That’s what I find most exciting.