Using collaborative simulation process to …...Simulation Process and Data Management (SDM) 24 - 25...
Transcript of Using collaborative simulation process to …...Simulation Process and Data Management (SDM) 24 - 25...
© 2010 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary© 2010 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary
NAFEMS European Conference:
Simulation Process and Data Management
(SDM)
24 - 25 November 2010, Frankfurt, Germany
Using collaborative
simulation process to
accelerate the Thermally
Aware Chip Design
Thomas Lehnhäuser,
Kamal Karimanal,
Shantanu Bhide,
Vamsi Krishna,
Kapil Sahu,
Pankaj Gupta
© 2010 ANSYS, Inc. All rights reserved. 2 ANSYS, Inc. Proprietary
Motivation
• Well-known for its
simulation software
products
• Data management
software (EKM)
Chip
Manufacturer
• Chip design expertise
• Seeks for
improvements in
simulation process
Collaborate on a new
approach for
simulating the thermal
behavior of chips
© 2010 ANSYS, Inc. All rights reserved. 3 ANSYS, Inc. Proprietary
Challenges in simulation
Multiple analysts working on same project
Search and reuse previous simulations
Global deployment and Collaboration
?
Collaboration Search and Reuse
Deployment
© 2010 ANSYS, Inc. All rights reserved. 4 ANSYS, Inc. Proprietary
Simulation Collaboration
EKMManage data, process,
and deployment
Multiple analysis runs?
What-if study ?
Simulation decisions ?
Reports ?
© 2010 ANSYS, Inc. All rights reserved. 5 ANSYS, Inc. Proprietary
end user
Compute
serverFile & web
servers
Engineering Knowledge Manager
File Repository
Meta-Data Extraction
Advanced Search
Data Mining
Report Generation
Automate Processes
Manage Workflows
Design Systems
E-mail Notification
Track Progress
PROCESS MANAGEMENTDATA MANAGEMENT
Enterprise Access
Web Enabled
ACCESS MANAGEMENT
Application Portal
Job Submission
© 2010 ANSYS, Inc. All rights reserved. 6 ANSYS, Inc. Proprietary
Technical Background
• Semi conductor Gate lengths are
shrinking – 28 nm
– Higher concentration of heat dissipation
– Leakage current
• Function of temperature
– Higher temperature -> more
leakage
• High level of integration on CPU/GPU
– Multi cores, Memory, other functionality?
• Heat generation is not uniform
• Keeps moving based on end user
application
• Employed Simulation techniques
– Chip design simulation
– Thermal simulation
Intel Pentium 4 (2000)
(42 million transistors)
Intel386™ (1985)
(0.275 million transistors)
© 2010 ANSYS, Inc. All rights reserved. 7 ANSYS, Inc. Proprietary
Traditional Thermal Management
Process
Chip Designer
Power map
Thermal Engineer
EDA Software
Temperature
Map
ANSYS Icepak Supplier
Issues with this approach
- Data exchange
- Data consistency
- Reponse time
© 2010 ANSYS, Inc. All rights reserved. 8 ANSYS, Inc. Proprietary
Enhanced Thermal Management
Process
Chip DesignerT-Estimator
Pre-Characterization
Passive Heat sink Cooling
(Telecom FPGA & ASICS)
Active Heat sink Cooling
(CPUs and GPUs)
Liquid Cooling
(High End GPUs & Supercomputer CPUs)
Pre-Characterization
Thermal Engineer
© 2010 ANSYS, Inc. All rights reserved. 9 ANSYS, Inc. Proprietary
Pre-Characterization
• Suppose that the relation between power
load and temperature is described by a
linear system: [T] = [C] * [Q]
• [T]: m X 1 matrix of temperature
responses
• [C]: mXn matrix of self and mutual
influence coefficients
• [Q]: power map represented in
nX1 matrix form
• Pre-characterization is the process of
determining the CmXn coefficient matrix
for one cooling design
• Thermal simulation was performed to
power the n sources of the power map
array in a linearly independent manner
1 W 0 W 0 W
0 W 0 W 0 W
0 W 0 W 0 W
0 W 1 W 0 W
0 W 0 W 0 W
0 W 0 W 0 W
370K 350K 330K
360K 340K 330K
320K 330K 330K
350K 350K 340K
340K 380K 340K
340K 330K 340K
© 2010 ANSYS, Inc. All rights reserved. 10 ANSYS, Inc. Proprietary
Validation of approach
System level
Validation Setup
Thermal EngineerANSYS Icepak
© 2010 ANSYS, Inc. All rights reserved. 11 ANSYS, Inc. Proprietary
Enhanced Thermal Management
Process
Chip Designer
Pre-Characterization
Passive Heat sink Cooling
(Telecom FPGA & ASICS)
Active Heat sink Cooling
(CPUs and GPUs)
Liquid Cooling
(High End GPUs & Supercomputer CPUs)
Pre-Characterization
Thermal Engineer
Data Life Cycle
Version Control
Global Data Repository
Custom Applications
Reporting
Searching
T-Estimator
© 2010 ANSYS, Inc. All rights reserved. 12 ANSYS, Inc. Proprietary
Generate new libraries
Publish new libraries on
EKM
Request new life cycle
stage
Change of stage for outdated libraries
Generate new libraries
Publish new libraries on
EKM
Request new life cycle
stage
Change of stage for outdated libraries
Generate new libraries
Publish new libraries on
EKM
Request new life cycle
stage
Change of stage for outdated libraries
Generate new libraries
Publish new libraries on
EKM
Request new life cycle
stage
Change stage for outdated libraries
Workflow in EKM
Thermal Engineer
© 2010 ANSYS, Inc. All rights reserved. 13 ANSYS, Inc. Proprietary
Check out Project and provide new
powermap
Execute Thermal estimator
Generate report
Check in project
Check out Project and provide new
powermap
Execute Thermal estimator
Generate report
Check in project
Check out Project and provide new
powermap
Execute Thermal estimator
Generate report
Check in project
Check out Project and provide new
powermap
Execute Thermal estimator
Generate report
Check in project
Workflow in EKM
Chip Designer
© 2010 ANSYS, Inc. All rights reserved. 14 ANSYS, Inc. Proprietary
Advantages
• Thermal result obtained from EKM based solution is more reliable since there is less chance file corruption
• Better risk management using EKM based solution
• Results can be shared easily among analysts/designers
• Thermal result library created in Icepak can be maintaied and enhanced in EKM in a user friendly way. Leads to collaboration between Analysts and Designers
© 2010 ANSYS, Inc. All rights reserved. 15 ANSYS, Inc. Proprietary© 2010 ANSYS, Inc. All rights reserved. 15 ANSYS, Inc. Proprietary
Thank you!