This project is an industry-centric project coordinated ARM with its Vision and Objectives determined by the marketing trends in industry.

The low-power cloud server concept relies on the observation that emerging Cloud computing applications (e.g., video/music streaming, song recognition and data mining) favour an increase in thread- and memory-level parallelism and benefit little from instruction-level parallelism. Each client request is serviced by a single or multiple independent threads each running on a dedicated processor core. These applications need many simple processor cores with high-bandwidth/low-latency access to very large memories. Using standard off-chip DRAM is bandwidth-constrained due to limited pin count, slow due to chip crossing, and power-hungry due to I/O pads and driving circuitry. To eliminate these inefficiencies and address these issues, we propose to 3D-integrated  DRAM chip on top of the ARM Cortex-A9 processor cores and  hardware accelerators (e.g. GPU, video and crypto engines).

The domination of the high-end high-performance servers is slowed down by the emergence of lower cost and lower power PC-based commodity servers around 2000. We expect a major shift to new servers designed with embedded CPUs around 2010 timeframe. These embedded-based servers will be equipped with off-chip DRAM, and emerge in the server market to meet the demand for energy-efficiency and green Cloud services. Our forecast is that  servers based on many-core embedded processors will come to the server market around 2015, promising many low-cost low-power server cores with high-bandwidth on-chip memory subsystem.

With the emergence of Cloud computing, significant amount of software services will be provided by these new servers located in green data centres. The new servers will need to address the requirements of billions of thin clients or handheld mobile devices that will demand Cloud services from these servers. Such a vast number of clients require a significantly large number of servers, and the servers have to be very cost and energy efficient.