Mohtalim

http://www.anandtech.com/show/8367/inte ... -in-detail

New node hard but Moore's Law scaling achieved, including $/transistor. Which, as you will recall, is not what TSMC is promising.

http://semiaccurate.com/2014/08/11/inte ... m-process/

Charlie is the only one I can find who actually analyzes what is said rather than regurgitates it. Sigh.

http://arstechnica.com/gaming/2015/08/d ... -nvidia/3/

There's your DX12 improvement, best case. (NVidia flat, AMD up substantially on a benchmark originally designed as a showcase for their hardware).

a benchmark originally designed as a showcase for their hardware

Exactly, designed for them. That particular alpha (pre-beta as they call it) "benchmark" is a DX11 style game converted to DX12. This DX12 migration happens to help out AMD with things they've been really bad at while nvidia was doing fine. Saying nvidia loses perf from already being really really good and comparing that to extreme suckiness transitioning to finally not sucking so much is a specious argument. nVidia's perf is still better in the end. nVidia's driver team is better, so this particular measurement won't be the last word for this benchmark and other benchmarks are in the works that will be designed more natively for DX12 will likely be more in nVidia's favor. This benchmark hammers on a particular resource which AMD has more of than nVidia cards which helps them out a bunch and really AMD should be beating nVidia cards, not just nipping at their heals.

Jonathan wrote:DX12 is FUD. Don't worry about that.

Forgot to quote myself.

Jonathan wrote:

Jonathan wrote:DX12 is FUD. Don't worry about that.

Forgot to quote myself.

Says the guy who works for the CPU company that DX11 helped keep up workload for. DX12 is supposed to reduce CPU workload so you can get by with a cheaper CPU.