So, is there a reason why the people with Intel in their rigs focused on the x264 bench I posted (which I mentioned was older, but it's about a year old, not TWO) and completely ignored the POV-RAY benchmark?
Haswells gains in x264 and Dolphin emulator come strictly from AVX2. Excavator will be getting AVX2 so in the future that point will be moot.
However if you feel like making chip features a part of the argument. the K series Intel chips have VT-D and IOMMU disabled. Virtualization is a very helpful tool if you're actually doing work and you need access to a different OS without rebooting.
An AMD CPU has the potential to use the hardware directly inside a virtual machine. Meaning you can get 3d acceleration in a guest OS. Everything on an Intel would have to be done in software only on the CPU.
So yes, Haswell is better at x264 thanks to AVX2. It's not much better at rendering, specially when you factor in the price. It's not much better at games that are properly threaded.
I guess we can sit here and circlejerk over how much faster Haswell is in x264 because it's one of two programs I know which use AVX2.
Like running Gentoo on my A4 5000 and FX 8350 and using every single instruction in every single program and part of the OS, it doesn't always help:
In fact, sometimes enabling AVX2 can hurt performance. So you folks making the claim that because x264 gets an almost 20% increase in performance by enabling AVX2 and then implying that will affect every piece of software that uses AVX2 are not understanding how operating systems or compilers work.
I have seen the same thing with my FX 8350, although I do think it's more optimistic with FX because most software is compiled around Intel. But there are programs that are twice as fast, 60% fast, etc. And then there are some that are barely 10% faster.
You don't see me running around going "MY FX IS 60% FASTER AT LAME ENCODING AND TWICE AS FAST AT BLENDER RENDERING! SO WHEN FX USES ALL ITS INSTRUCTIONS IT'S TWICE AS FAST IN EVERYTHING!!!!!!!!!" because I know it's not true.
Haswell pulling away from FX means that the software has to use AVX2.
FX competing (not necessarily beating) FX means that the software uses multiple cores. 4770k and FX 8350 both have 8 threads. Do you see what I'm getting at?
I won't deny it. There are situations where Haswell has a huge advantage thanks to AVX2. However that doesn't mean that that huge advantage transfers over to all software. Even if AVX2 is enabled in other software, it might hurt performance.
I can play software tricks and make my FX 8350 at 5ghz 30% faster than a 4ghz 3930k in Blender, but I don't go running around claiming that 3930k is dead and FX is faster in everything.