Need for CPU Accelerators
The advent of multi-core CPUs and GPUs has helped to increase the performance of modern datacenters. However, this performance is being limited by a non-proportional increase in energy consumption. As workloads like Big Data analytics and Deep Neural Networks continue to evolve, there is a need for a new computing paradigm which will continue scaling compute performance but drastically lower power consumption and at an affordable cost.
The eASIC Platform provides the ability to cost effectively offload the main processor with applications that are performance and power sensitive. The eASIC Platform uses a highly parallel architecture to efficiently implement compute elements with local storage. This combination allows applications such as big data analytics, searching, pattern-matching, encryption and signal and image processing to be implemented with extremely efficient performance per watt footprint.
The eASIC Platform provides the following key features:
- Up to 80% lower power than FPGAs
- Up to twice the performance of FPGAs
- Embedded one bit fast adders for high-speed arithmetic and DSP functions
- High speed SERDES up to 28 Gbps
- High speed memory interfaces (including DDR4)
- GreenPowerVia disables power to unused logic (no leakage for unused logic)
- eFUSE for traceability and storing special keys
- FPGA like design tools and flow
To read the press release go to eASIC Joins the OpenPOWER Foundation to Offer Custom-designed Accelerator Chips page.