Summer is when people buy electronics, don't ask me why, but that's why summer has more than one popular electronics show where everyone breaks out their best new stuff. CES is the consumer show that just happened and AMD stole the show with their newest Ryzen processor, while Intel played coy. All of Intel's newest ultralight laptops about to come on the market were displayed, but with their batteries drained so nobody could take any benchmarks. Nonetheless, one benchmark was leaked that showed their new chip to have off the scale physics. I'm talking a stupid 15 watt chip that is otherwise unremarkable appearing, that crunches physics equations like cotton candy. Intel has also hinted that we may require new synthetic benchmarks, and the physics numbers hint that they have done Nvidia one better and integrated AI circuitry into their graphics so efficiently they run on this wimpy chip. All of Intel's new chip contain small AI circuits already, and we can possibly expect them to soon announce a completely new AI driven gpu architecture that makes Nvidia's tensor cores look wimpy. That might sound strange, but Nvidia's tensor cores are merely added on IBM asics, while Intel can integrate all the fpga circuitry they want any way they want. If I'm right, they have compiled the data for what physics run easy on a processor or gpu and reduced it to a minimal set of equations for what looks good in video graphics and, then, built it into their gpu and processor as fpga circuitry that can do it all faster and more efficiently. The new Ice Lake chips could produce incredible graphics on 15 watts using this kind of approach, but its so complicated I can see why Intel has suddenly hired half of AMD's graphics department. Ice Lake includes 5g capability, meaning it could produce high resolution graphics with smooth frame rates and outrageous physics for things like explosions and vehicles without an internet connection and, with one, the sky is the limit. We'll just have to wait and see, but those physics numbers haunt me. Extrapolating on the same design they could easily add ray tracing or anything they want to their discrete graphics cards and produce significantly better results more efficiently than anyone else.
