LEXINGTON, Mass.--(BUSINESS WIRE)--Adapteva, a privately-held semiconductor technology start up, today announced that it is sampling the fourth generation of its Epiphany multicore architecture, making the ground-breaking performance and energy efficiency of the Epiphany platform widely available for co-processing solutions to be implemented by industries requiring the next leap forward in parallel computing. Adapteva’s new OpenCL SDK, also announced today, provides a portable API for accessing the compute capabilities of a platform, accelerating performance in a wide spectrum of applications in market categories from gaming and entertainment to scientific and medical software.
The Epiphany IV chip was implemented in GLOBALFOUNDRIES’ leading-edge 28nm-SLP process technology. The product includes 64 independent high-performance RISC cores on a single chip, effectively offering over 50 GHz of achievable programmable performance while consuming less than 2 Watts.
“This is a watershed moment for low-power parallel computing, pushing manycore computing far ahead of what the market thought possible within this decade. With our latest OpenCL programmable 28nm chips we have created an easy-to-use multicore processor solution that boosts energy efficiency for critical applications by an order of magnitude in a broad range of markets,” said Andreas Olofsson, CEO of Adapteva. “The Epiphany efficiency breakthrough will enable significant savings in total cost of ownership for high-performance computing and will enable server level performance in mobile devices such as smart-phones and tablets.”
About the Epiphany-IV Multicore Implementation
A key reason for the unmatched efficiency of the Epiphany-IV chips comes from early access to GLOBALFOUNDRIES leading-edge 28nm low power process. The industry to date has been challenged by the 28nm process node but the Epiphany chip was able to exceed expectations in terms of performance.
Initial testing of the Epiphany-IV chips has demonstrated unprecedented performance numbers:
- 800MHz peak operating frequency
- 51.2 GHz of equivalent microprocessor performance
- Peak performance of 100 GFLOPS
- 77,912 CoreMark Score
- Peak core energy efficiency of 72 GFLOPS/Watt at 500MHz.
- Max 2 Watt “worst case” chip power consumption
- 80mW chip idle power
- 10mm2 silicon die packaged in a 324 ball 15 x 15 mm BGA package.
“Early engagement with our customers helps to deliver innovative products with industry-leading performance and energy-efficiency,” said Mike Noonen, executive vice president of worldwide marketing and sales at GLOBALFOUNDRIES. “Our close collaboration with Adapteva is helping them bring a whole new class of multicore microprocessors to market on the foundry industry’s only production-proven HKMG technology.”
Programming with Standard Open Source Tools
Epiphany is the first massively parallel processor to achieve leading performance metrics while fully supporting standard programming languages like ANSI-C, C++, and IEEE floating-point data manipulation natively. Adapteva’s programming tools are based on standard open source packages, including GCC, GDB, Eclipse, and Newlib. The latest Epiphany product will also be programmable using an open source OpenCL SDK, allowing programmers to quickly scale their applications to take advantage of thousands of Epiphany cores within a system. For developers, Adapteva’s open source programming tools means lowered risk and faster time to market.
A Powerful Application Accelerator
The Epiphany multicore architecture was designed to sit next to an x86 or ARM microprocessor to accelerate computationally intensive tasks. Epiphany is only multicore IP offering that supports true task-based execution. By virtue of being a general purpose multicore RISC architecture, the Epiphany can take on complex tasks from start to finish without any involvement from the host. To demonstrate the effectiveness of the Epiphany accelerator, the company has developed a low-power parallel face detection solution based on the open source Open CV image processing library.
“Amdahl’s Law implies that if you need to accelerate an application by orders of magnitude, the programmer needs to take all the serial parts of the code out of the application,” continued Olofsson. “It’s very hard to do this if the accelerator is limited in the types of operations it can do well. Our CPU-based accelerator architecture allows for better host-accelerator bandwidth conservation and higher scaling factors by more effectively decoupling the host from the accelerator.”
Target Markets and Availability
The Epiphany-IV multicore IP is the ideal technology for enabling the next level of performance in a wide range of battery constrained mobile markets including, high performance tablets, smart-phones, embedded vision systems, software defined radio, radar, security and medical diagnostics. Adapteva is currently sampling Epiphany-IV chips and evaluation kits to qualified early access customers.
About Adapteva
Adapteva, Inc. is a privately-held semiconductor technology company based in Lexington, Massachusetts. Adapteva has developed the world’s most energy efficient multicore microprocessor architecture, immediately boosting by an order of magnitude the number of cores that can be integrated on a single chip. Adapteva’s breakthrough architecture will have an immediate impact in a wide range of end-user products from compact mobile devices to next generation of supercomputers. For more information on the company visit http://www.adapteva.com or via twitter @adapteva.