Graphcore Uses Synopsys Design Platform to Implement Colossus Chip to Accelerate AI Computing
Synopsys Fusion Technology Enables Superior Power, Performance, Area for AI Chip Design
MOUNTAIN VIEW -- Calif., Sep 18, 2018 -- Synopsys, Inc. (Nasdaq: SNPS) today announced that Graphcore has used the Synopsys Design Platform to successfully implement its Colossus™ intelligent processing unit (IPU) to accelerate artificial intelligence (AI) computing compared to existing processors (CPUs and GPUs). Comprehensive Synopsys digital and custom design flows, enhanced with key AI-focused optimization technologies, enable designers to achieve target quality-of-results (QoR) and time-to-results (TTR) goals on their AI chips for cloud computing, data center, automotive, and mobile applications. Synopsys' Fusion Technology™ has been augmented with optimization capabilities for AI chip design, including interconnect planning, multiply-accumulate (MAC) topology optimization, and complete AI IP reference flows for maximum speed, smallest area, lowest power, or an optimum balance of the three.
The demand for AI processing potency has been doubling every 3.5 months since 2012 (source: OpenAI), however, the escalating performance, power, and latency required to achieve the processing level are pushing the limits of existing CPU and GPU processors. Graphcore's Colossus IPU is optimized for AI workloads, and can hold entire neural networks on the chip. It employs graph computing with massively parallel, low-precision floating-point compute, and provides much higher compute density than other solutions to accelerate both training and inference for machine learning.
"The Synopsys Design Platform with Fusion Technology provides all of the capabilities we need to achieve superior processing performance for artificial intelligence and machine learning," said Phil Horsfield, vice president of Silicon at Graphcore. "Our partnership with Synopsys has been instrumental in developing the Colossus IPU in a fast-moving, dynamic market for driving current and future AI applications."
The typical AI chip consists of thousands of replicated processor cores, characterized by high throughput, energy efficiency, and low latency. The processor core is the critical building block, with challenging connectivity requirements that can lead to routability-limiting congestion. Synopsys Fusion Technology is the preferred choice for AI chips, providing the best power, performance, congestion/area, yield, and design closure. It has been enhanced with several key AI-focused optimization technologies, including AI chip interconnect planning for hundreds of replicated modules, MAC topology optimization, complete AI IP reference flows, full-flow concurrent clock and data optimization, wire synthesis, and logic restructuring. Comprehensive digital and custom design flows comprised of Design Compiler® Graphical synthesis, IC Compiler™ II place-and-route, PrimeTime® signoff, StarRC® extraction, Custom Compiler® custom layout, SiliconSmart® characterization, and HSPICE® circuit simulation deliver best-in-class QoR and TTR.
"As a leader in AI chip design, Synopsys has been working with pioneers in computer vision, object and speech recognition, and big data analytics," said Sassine Ghazi, co-general manager, Design Group at Synopsys. "Graphcore's Colossus IPU is a major achievement in machine learning, empowering AI applications to move data across the chip orders-of-magnitude more efficiently, while significantly reducing any wasted processing power. Synopsys' Fusion Technology is enabling AI chip companies like Graphcore to quickly bring to market highly-differentiated, industry-defining Super Chips with Synopsys."
About Synopsys
Synopsys, Inc. (Nasdaq: SNPS) is the Silicon to Software™ partner for innovative companies developing the electronic products and software applications we rely on every day. As the world's 15th largest software company, Synopsys has a long history of being a global leader in electronic design automation (EDA) and semiconductor IP and is also growing its leadership in software security and quality solutions. Whether you're a system-on-chip (SoC) designer creating advanced semiconductors, or a software developer writing applications that require the highest security and quality, Synopsys has the solutions needed to deliver innovative, high-quality, secure products. Learn more at www.synopsys.com.
Related Semiconductor IP
- Multi-channel, multi-rate Ethernet aggregator - 10G to 400G AX (e.g., AI)
- Multi-channel, multi-rate Ethernet aggregator - 10G to 800G DX
- 200G/400G/800G Ethernet PCS/FEC
- 50G/100G MAC/PCS/FEC
- 25G/10G/SGMII/ 1000BASE-X PCS and MAC
Related News
- Synopsys and TSMC Streamline Multi-Die System Complexity with Unified Exploration-to-Signoff Platform and Proven UCIe IP on TSMC N3E Process
- TSMC and Synopsys Bring Breakthrough NVIDIA Computational Lithography Platform to Production
- SEMIFIVE joins Arm Total Design with plans to develop Arm Neoverse-powered HPC Platform
- SignatureIP Makes Network-on-Chip (NoC) Design Widely Accessible with Cloud-Based iNoCulator™ Platform
Latest News
- How CXL 3.1 and PCIe 6.2 are Redefining Compute Efficiency
- Secure-IC at Computex 2025: Enabling Trust in AI, Chiplets, and Quantum-Ready Systems
- Automotive Industry Charts New Course with RISC-V
- Xiphera Partners with Siemens Cre8Ventures to Strengthen Automotive Security and Support EU Chips Act Sovereignty Goals
- NY CREATES and Fraunhofer Institute Announce Joint Development Agreement to Advance Memory Devices at the 300mm Wafer Scale