Originally Posted by maarten12100
We have to blame Intel for unfair competition and they still do it.
Not much reason to do so anyway I put my trust in an arch that can be improved upon way more than P3.
Maybe one day we'll see a universal compiler not manipulating non Intel hardware and being widly adopted
I believe even with SSE2 AMD might be equal or faster with fully hardware HSA (not OpenCL), as there would be thousands of GPU cores reading and processing CPU's pointers. No doubt It will beat even AVX2 ( >512 bit instructions ).
The only thing AMD might be defeated is the power consumption as few CPU cores consume much less power than many GPU cores.
I don't remember that link where I saw A10 6800k (w/ OpenCL) beating i7 3770k (w/ and w/o OpenCL) in Handbrake Video encoding .
Though it requires special conditions where apps must support multithreading with OpenCL. So its not that big deal as most general software don't support these.
But If they can do it with just a high level API, imagine how fast hardware HSA would be !