But how many of those single threaded apps are rigged? After al if I shoot your horse when it starts out the gate my horse will "win the race" even if its a nag on its last legs, doesn't mean my nag was better or yours worse, just means the race was rigged.
Look up "Intel cripples compiler" and next time you run into a single threaded program that jumps way up on an Intel CPU? WRITE THE COMPANY, you'd be surprised how many of them have not heard about Intel rigging their compiler and when its pointed out will be happy to offer an AMD64 compiled with GCC. Once you compare the same program compiled with GCC you'll find its less than 20% on the high end and often single digits on the low and midrange.
So do some checking, see for yourself. And the benchmarks BTW are 100% rigged, they receive large checks in the form of Intel ad dollars and won't use anything but ICC and therefor are totally worthless for judging squat anymore. The only place where benches work anymore are on the GPU side but just like "Windows isn't done until Lotus won't run" I'm sure Intel is hard at work trying to figure out a way for ICC code to cripple any non Intel GPU and when the day comes they find it? You'll find the Intel HD6k suddenly "stomps" the latest from nvidia and AMD...its rigged, its a scam, look up the tests for yourself and see.
I do agree on bulldozer though, which is why I stuck with Phenom II until the better Piledriver units came out. Bulldozer was a server chip that the former CEO tried to shove down the throats of the populace because he fired most of the engineers on the consumer side to get a stock bounce. Luckily he's gone now and the new guy hired back one of the leads on Athlon64 so the new chips should be a lot better but as long as ICC is pulling a quack.exe you can't trust the numbers, you have to find out what compiler was used for the code.