+Andrei C Zamfir
Somewhat, yes - it depends on what you expect.
I have run Geekbench on my fully stock clean factory reset SGS3 (international) about 20 times now, and it scores around 1800 on average. Not once has my score been nearly as low as PCMag claims. It has beat the iPhone5 score every single run - mostly on the INT and FP tests. Others are reporting the same thing, so PCMag's SGS3 results are fishy to say the least - this could well be related to the specific SGS3 model they use.
Now, before anybody goes shouting about JB vs ICS, if Geekbench is a proper benchmark, that shouldn't actually matter significantly, at least not on the INT and FP tests (that show a large difference between my own and PCMag numbers). Assuming most of the processing power of the device is actually taken by those tests (as should be the case) the INT and FP tests could only be different if they were implemented in Java instead of native code, in which case Geekbench on iPhone vs Geekbench on Android is comparing apples with oranges - the numbers simply cannot be compared reliably, as you're not testing the same thing. So it's either that (and Geekbench is useless cross-platform) OR ICS/JB doesn't make a difference.
Not to mention, Geekbench runs such an array of tests in such a short time (mere seconds total) it's never going to be very reliable - the shorter the test, the bigger the error - especially with background activity "noise". That's why real
benchmarks take a while to run.
Browsermark, haven't tried. Sunspider, I get 1300 on Sunspider - cleanly halfway between iPhone5's and PCMag's SGS3 score.
Both Browsermark and Sunspider don't really test a phone's speed though, they test how fast your browser implementation is combined with your hardware.
That doesn't mean Browsermark and Sunspider results aren't valid, but they only indicate max "browser performance". It says absolutely zilch about the performance of apps running natively on the device.
Contrary to the other results, GLBenchmark is a pretty reliable benchmark. The iPhone5's GPU appears to be a clear win over the SGS3 model used by PCMag.