NVIDIA Unveils Next-Generation GPU

New Pascal GPU architecture supports stacked memory and eliminates bus bottlenecks.

At the 2014 GPU Technology conference in San Jose, CA, NVIDIA has announced details on their new Pascal GPU architecture, which they will officially release in 2016. The new Pascal GPUs will support stacked memory, a technology that allows multiple layers of DRAM memory components on a single unit. According to NVIDIA, stacked memory provides "several times greater bandwidth, more than twice the capacity, and quadrupled energy efficiency" compared to present alternatives."
The increased capacity of GPUs requires a higher-bandwidth connection between GPUs or between the GPU and CPU. NVIDIA has also unveiled a technology called NVLink that it says will replace PCI Express as a conduit for communication between CPU and GPU components. NVIDIA says NVLink will provide "a more energy-efficient, high-bandwidth path between the GPU and the CPU at data rates 5 to 12 times that of the current PCIe Gen3. NVLink will provide between 80 and 200 GB/s of bandwidth, allowing the GPU full-bandwidth access to the CPU’s memory system."
The first Pascal GPU products will appear in 2016.