On the eve of CES 2016, NVIDIA held a press conference, highlight of the program where a car supercomputer DRIVE PX 2. We know that this computer system, promises to be six times more productive than GeForce GTX Titan X, based on Tegra processor new generation
The initial development plan assumed that Parker will combine the core Denver and Maxwell graphics generation , but the system-on-chip delayed in a way that seems to have made the NVIDIA substantially revise the device processor.
Parker will be manufactured by TSMC 16 nm technology FinFET, they will be the first products NVIDIA, released on the process technology. The processor has four cores ARM Cortex A57 core and two Denver. In the previous generation, we know that Denver relies on its own architecture with a very long machine instructions (VLIW), which are obtained by binary translation and optimization of conventional ARM-code. It looks like the core Denver should provide high speed processor in single-threaded tasks while the cluster core Cortex A57 will undertake a multi-threaded calculations and tasks, where the "optimizer" Denver is inefficient.
We can be sure that the processor will receive the schedule generation Pascal, moreover cluster CUDA cores will be quite large: NVIDIA in this case does not limit itself to the scope of power consumption and heat dissipation. Given the significantly increased performance, it can be assumed that Parker include at least 1024 cores CUDA.