Iris Pro vs Intel HD vs r7 APU - Post Processing - Cores vs iGP for DX12

[jamesgalb]

So I'm looking for some insight as to the power of Iris Pro when it comes to post-processing.

I am considering an ITX build and am weighing the balances of x99 vs z170, mainly thinking of the processors...

With DirectX12 around the corner, utilization of more cores is going to get better, which weighs in favor of the x99 platform.... HOWEVER, it will also bring in multi-GPU configurations, allowing iGPs to take over things like post-processing, weighing in favor of something with an iGP...

So without any real Skylake/Iris Pro benchmarks existing, is there any insight as to how the Iris Pro compares to the Intel HD and AMD APUs in things like post-processing? For frame of reference.

 

[Vittra]

I'd be careful about expecting multi-GPU configurations gaining traction anytime soon. DX12 will bring a lot of good things, but I suspect there will be resistance to some of them.

With that being said, there is a line of current processors (Crystal Well) that currently use the Iris Pro 5200 which already utilizes eDRAM (the Pro 5100 did not):

http://ark.intel.com/products/codename/51802/Crystal-Well#@All

The 5200 runs 40 EUs whereas the 6100 will purportedly run 48 EUs.

Here's an older article on Anandtech that discusses the performance of the 4950HQ:

http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested

I believe it would be reasonable to expect similar performance, perhaps within the 5% range. 15-20% would be a stretch.

 

[jamesgalb]

performance looks strong from those benchmarks... Iris Pro looks like it would deliver enough power to take over many post-processing effects...

I'd be surprised if multi-GPU configurations lagged behind. Without a doubt AMD and Intel will be pushing this, and nVidia would be hurting themselves in future purchases if they did anything to stop some standard set of post-processing effects to be offloaded to another GPU...

I dont believe we will get SLI/Xfire on different GPUs, even within the same brand... But I would think specific post-processing tasks will be coded so they can be offloaded to any GPU. At that point I dont think more than 1 additional GPU, or an excessively powerful 2nd GPU, will really present any gains... Overall performance will still be determined by the main GPU (or SLI/Xfire), I would think.