Question.

>could they not surpass Intel and nVidia in performance with some sort of proprietary synergy between the 2?

APUs
And Intel already beat them in that(albiet more expensive)

>>56603440
It's called HSA and it's coming along nicely
Not proprietary though
>>56603455
>And Intel already beat them in that(albiet more expensive)
They don't, Intel doesn't even support HSA, their OpenCL drivers are somehow more awful than AMD's and iGP wise AMD rapes them

>>56603440
They already do

>>56603455
intel's apu gpu performance is horse shit compared to amd's.

what holds back amd's apu is the cpu performance in cpu related tasks.

in games the gpu in amd's apu's are powerful enough that in most games amd's apu's come out ahead of most intel's.

the only gpu's intel has that offers competitive performance are their edram iris pro ones (found in broadwell) but even they lack behind amd's newer offerings now.

zen based apu's should win in both cpu and gpu departments.

one thing amd has going for itself with zen is pairing a amd apu with a amd discrete gpu and running them in crossfire. a little extra free performance.

File: igpu.png (753KB, 844x2210px) Image search: [Google]

753KB, 844x2210px

>>56603455
Intel's integrated graphics solutions are still utter trash. Not that APUs are particularly great, but at least they're a third of the price.

Unfortunately they can't do that. The majority of people use intel. Even if they target the low end to get that majority, there's no guarantee those poorfags would just abandon their already working intel/pcie system for a slightly better amd/proprietary connection.

>>56603581
>>56603455
>>56603455
Is amd apu better at dual crypto mining?

They would have to make a gpu so badass that they have to tone it down to match nvidia and lock that extra power unless you use an amd cpu with a special bridge. Just doesn't seem very likely.

>>56603728
>make a gpu so badass that they have to tone it down to match nvidia and lock that extra power unless you use an amd cpu with a special b
I guess that's one way of looking at it. I was thinking more along the lines as having a more efficient cross talk to boost the performance, rather than purposely throttle down performance with certain hardware configurations.