In a nutshell, we're probably going to see an 8-16 core SoC with high-bandwidth interconnect, utilizing high-bandwidth memory. Given the timeframe of its arrival (2014), we'd wager it's too early for GDDR6 memory and it will probably have to rely on DDR4, expanding it with memory space on parts such as ioFX (both NVIDIA and Fusion-io recently started to work together on expanding the memory addressing directly to GPUs/SoCs). Details about the actual Project Boulder silicon are scarce right now, and we would rather wait and see instead of throwing out guesstimations.The site also has some news about Project Denver, NVIDIA's effort to create a CPU/GPU processor with an ARM-based design. Valich heard Project Denver originally was a RISC architeture capable of executing ARM, x86 and MIPS instruction set architectures (ISAs). The goal was to create a dual-core part with a cluster set of Fermi cores to take on Intel's Arrandale in 2010, originally a dual-core x86 processor with built-in Larrabee GPU. Both companies' plans got cancelled however, Intel killed the consumer version of Larrabee and NVIDIA decided to abandon the work to focus on low-power processors. NVIDIA's first Project Denver part is anticipated to be the T50 (Tegra 5), it will combine Project Denver cores with the Maxwell GPU. This chip will reportedly be manufactured using a 20nm-SLP process at Samsung's fab in Austin, Texas.
On a related note, Bright Side of News also hints that Apple has a similar hardware development program in the works to replace the Intel products from its consumer notebook and desktop products. The plan is to create a self-made processor, but this won't be possible until iOS is mature enough to replace the Mac OS.