Performance Efficiency AI Accelerator for Mobile and Edge Devices
IBM, GLOBALFOUNDRIES Enhance Si2 Unified Power Model Standard
Patented Technology Supports System-level Power Models, Temperature Co-Dependencies
AUSTIN, Texas -- October 17, 2018 -- Silicon Integration Initiative, an integrated circuit research and development joint venture, announced today that IBM and GLOBALFOUNDRIES have contributed patented technology to support the Si2 Unified Power Model standard, the industry’s first significant power model enhancement in many years.
Early stage estimation of System on Chip power consumption is fundamental to ensuring new SoC designs meet or exceed power specifications when fabricated. For a credible estimate, the power models must comprehend the target implementation technology and circuitry, along with voltage and temperature conditions. At the same time, power estimation results are needed quickly to perform rapid “what if” scenarios.
UPM’s multi-level power modeling capability provides the necessary level of modeling detail required at various stages of design. Abstract high-level equations to gate-level characterization tables can be accommodated through the same, standard interface. Beyond this, the UPM interface, upon acceptance and approval by the IEEE’s P2416 working group, will be a direct plug-in to the widely-used IEEE 1801 stub created for power models.
The IBM and GF contributions enhance UPM by providing a new and unique approach to power modeling. Rather than storing pre-characterized, process-voltage-temperature specific data, UPM models store power proxies that represent different contributors to overall power consumption, such as sub-threshold leakage, gate leakage, and dynamic power. Appropriately entitled “power contributors,” this approach vastly simplifies and reduces the power modeling effort, and allows the power model to be voltage and temperature independent, enabling a single power model to be used at a multitude of voltages and temperatures.
SoC designers using UPM with contributor-based modeling will ultimately be equipped with thermally-aware, system-level power estimation. In addition, the late-binding of specific PVT conditions at simulation run-time will provide accurate, early estimates of leakage power, which increases exponentially with increasing temperature. The donated technology covers key aspects of contributor-based power modeling including model abstraction, generation, compression and evaluation.
Contributor-based modeling will be fully integrated into UPM, which forms the basis for P2416, the planned IEEE standard for developing and maintaining interoperable, IC design power models. P2416 is scheduled for balloting in early 2019.
Jerry Frenkil, director of Si2 OpenStandards, said the IBM and GF contributions bolster UPM and provide P2416 with proven and ready-to-use modeling methods. “These power proxies enable voltage and temperature-independent modeling which greatly reduce the model generation and support effort,” Frenkil explained. “They also enable late binding of voltage and temperature conditions at simulation run-time, a major benefit for both IP developers and SoC designers.”
“IBM is pleased to donate this advanced modeling technology to Si2’s UPM development to facilitate interchange of IP power data,” said Dr. Leon Stok, vice president of EDA at IBM. “We have used contributor modeling internally on several generations of IBM micro-processors to great effect. We look forward to seeing UPM contributor models being provided by IP block developers so that entire systems, consisting of both internal and external IP, can be modeled efficiently using a common modeling standard. Additionally, the combination of power contributors and multi-level modeling structures promises major cost and resource improvements in creating and supporting IP power models.”
“UPM directly addresses a major industry need—accurate and efficient system-level power models,” said Richard Trihy, senior director of design enablement at GF. “Since IP providers need only produce a single model for a multitude of PVT points, these models enable significant productivity gains in model generation. Our clients will also get a good early estimate of their systems’ total power, including leakage, which can operate at high temperatures.”
“These contributions from IBM and GF come at a fortuitous time,” said Dr. Nagu Dhanwada of IBM, chair of the IEEE P2416 Working Group and the Si2 UPM development project. “The P2416 Working Group is rapidly gathering momentum towards IEEE standardization. We anticipate going to ballot early next year.”
For more information about this project, contact Jerry Frenkil at jfrenkil@si2.org. For information about the Low Power Working Group and other OpenStandards programs, visit http://www.si2.org/openstandards/.
About Si2
Founded in 1988, Si2 is a leading research and development joint ventures that provides standard interoperability solutions for integrated circuit design tools. All Si2 activities are carried out under the auspices of the National Cooperative Research and Production Act of 1993, the fundamental law that defines R&D joint ventures and offers them a large measure of protection against federal antitrust laws. Si2’s international membership includes semiconductor foundries, fabless manufacturers, and EDA companies.
|
Related News
- Dolphin Design announces partnership with Fraunhofer IIS to enhance power management in edge AI
- GlobalFoundries Files Lawsuit Against IBM to Protect its Intellectual Property and Trade Secrets
- ARM, IBM team on low power analog AI chip
- Vidatronic Announces Power Management IP Now Certified on Globalfoundries' 22FDX® Platform for IoT Applications
- Arasan announces the immediate availability of its ultra-low power MIPI D-PHY IP for the GlobalFoundries 12nm FinFET process node
Breaking News
- HPC customer engages Sondrel for high end chip design
- Ubitium Debuts First Universal RISC-V Processor to Enable AI at No Additional Cost, as It Raises $3.7M
- TSMC drives A16, 3D process technology
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- TSMC drives A16, 3D process technology
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
E-mail This Article | Printer-Friendly Page |