blazorthon
Glorious
- Sep 24, 2010
- 14,022
- 21
- 43,265
[citation][nom]teh_chem[/nom]Your point doesn't still stand. It's not just that. Getting rid of the latency/interconnects is only half of it--Integrating a full graphical chipset (with the instruction set support) enables you to take advantage of acceleration with those. While the HD4K is a supreme advancement for Intel over their previous iGPUs, they still lack a lot of the instruction set support at the hardware level. Not saying that they'll be that way forever, but architecturally-speaking, the APU is more advanced while intel's CPU's undoubtedly reign king in (unnecessary) CPU power for the average user.[/citation]
Those're still not architectural differences that separate APUs from any other GPU/CPU hybrid chips. My point does still stand. Furthermore, other integrated GPUs (including Intel's) have support for some such instructions. That they don't support them all at this time is not an architectural difference, just a matter of supporting some code or not. It isn't even hardware problems, it's just Intel's drivers. Intel's GPUs are physically capable of running OpenCL and such and we know that because they have been tested with it. They simply aren't the greatest at it and don't have great driver/software support for it. It's kinda like how AMD's GPUs are physically capable of running PhysX, but Nvidia doesn't want to let AMD be capable of it anymore, so they don't let AMD video cards run current versions of PhysX. These aren't hardware problems, granted the hardware could be optimized for it in order to be better at it. The problems are in the drivers and software allowing this or not allowing this to be done because either they don't support it or they support it, but are coded to not allow it.
This isn't like an Phenom II CPU not supporting AVX instructions when an i5 or i7 do support them. This is more comparable to say if MS would suddenly not let Silverlight code run on an AMD CPU. The Phenom II would still be physically capable of it, but the software refuses to allow it.
Those're still not architectural differences that separate APUs from any other GPU/CPU hybrid chips. My point does still stand. Furthermore, other integrated GPUs (including Intel's) have support for some such instructions. That they don't support them all at this time is not an architectural difference, just a matter of supporting some code or not. It isn't even hardware problems, it's just Intel's drivers. Intel's GPUs are physically capable of running OpenCL and such and we know that because they have been tested with it. They simply aren't the greatest at it and don't have great driver/software support for it. It's kinda like how AMD's GPUs are physically capable of running PhysX, but Nvidia doesn't want to let AMD be capable of it anymore, so they don't let AMD video cards run current versions of PhysX. These aren't hardware problems, granted the hardware could be optimized for it in order to be better at it. The problems are in the drivers and software allowing this or not allowing this to be done because either they don't support it or they support it, but are coded to not allow it.
This isn't like an Phenom II CPU not supporting AVX instructions when an i5 or i7 do support them. This is more comparable to say if MS would suddenly not let Silverlight code run on an AMD CPU. The Phenom II would still be physically capable of it, but the software refuses to allow it.
Facebook
Twitter
Instagram