Originally Posted by moshpit
But now I've gotta ask you, just how much use will Dx11 on iGPU be? I know you don't think you'll be playing Dx11 games in full Dx11 glory on Llano's iGPU. It's going to be lucky to play BFBC2 in Dx9 mode with any kind of decent performance. I honestly cannot think of one Dx11 game title I would even WANT to try on any iGPU, including Llano's leaked specs. That is, unless one's playing games on a 1024 res monitor.
You're talking about first-generation DirectX 11 games, and you're right. Currently, the most obvious uses of DX11 are for ultra quality options and high performance cards. That doesn't mean there aren't hidden ones, like the use of Direct Compute for physics in the EGO/EGO 2 engine.
Most of the benefits of DirectX 11 have yet to be realized, especially as driver support for command lists/multi-render target etc. have yet to be implemented (except for NVIDIA with Forceware 270, which is useful for Civ5).
However, BF3 and newer engines are looking beyond just using direct compute for AO and are exploring using the tessellation, HD texture support, multi-render target/command list features.
If your advice for people now is to wait to upgrade their 2-3yr old computer because you don't like the CPU core in Llano, because DirectX 11 isn't used fully yet and anyway by the end of 2012 Intel CPU's will support that API, too, at less performance than you can buy now? Spectacularly bad, completely short sighted and self serving for an agenda only a handful of people even partially care about. You place overwhelming emphasis on CPU, yet consider your computer - how much did you spend on graphics? and how much on CPU? Why is that, which is more important to you?
As someone here so eloquently says, "Don't let great be the enemy of good". I'd love to see Llano feature new x86 architecture cores, but it doesn't. We have to wait until BD+/BD 2 cores hit for that product. Thats the timing. I don't think we need to go off onto why that timing is the way it is, its a combination of events not all of them under AMD's control.
Another API - OpenCL. Intel, NVIDIA, ARM and AMD support it. AMD have 400-500TFLOPS of GPU cores to run it on in an A-series APU. Intel has... hmm. So the questions is, does that matter? Will there be a broad infrastructure of OpenCL accelerated applications? Yes. By the time Ivy Bridge launches, it's going to be very obvious you need horsepower for OpenCL.