Upload
christine-townsend
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
Paradigm Shift to 3D and its Impact on our EcoSystem, Standards and EDA Tools
Herb Reiter, eda 2 asic ConsultingChairman of GSA EDA Interest Group
& 3D IC Working GroupEDPS, Monterrey, April 8, 2011
Agenda
“2D ICs” make continued innovation very expensive Need EcoSystem changes to enable many of us to continue innovating
3D demands closer cooperation AND additional standards Data exchange formats & tools interoperability foster cooperation Complex assembly processes and much higher packing densities
demand more rules (… like living in a condo complex vs on a 50 ac ranch…)
3D technology offers new opportunities for EDA 3D is a SYSTEM integration technology, not limited to IC design Opens up a wide range of EDA opportunities:
- Engage in multi-level modeling- Make proven 2D tools “3D aware”- Solve complex system-level H/W & S/W planning, implementation &
verification challenges with EDA tools, flows and methodologies- DFT for economical testing of 3D ICs
Discussion, conclusions, action items
Challenges with the proven 2D SoCs
• 2+ years development time • Software is ~ ½ of the TTM• NREs of $ 100M and more
demand very high volumes to reach breakeven point
• RISKs: Design error Market changes Competitor moves 1st
Synopsys Keynote, @ LSI, Oct 6, 2010
Major Implementation Alternatives
Source: eda2asic article in Yole’s 3D Packaging Magazine, Nov 17, 2010
2½ + 3D Commercialization Schedules
http://www.i-micronews.com/lectureArticle.asp?id=6351 Dr. Phil Garrou, YOLE= this year
2½D with
Characteristics of our current EcoSystem ICs, packages and PCBs are designed (almost) in isolation Monolithic “2D ICs” are hitting economical and technical limits Several modular solutions (PoP, PiP, SiP,… ) are production proven New 2½D* and 3D configurations are emerging as alternatives
to integrate ICs and entire systems (*: 2½D = interposer-based)
What do most of our end customers want ? Lower cost, smarter, faster, smaller systems with longer battery life Following latest trends quickly ( shorter time-to-market in design & manuf.)
How can we meet these requirements ? Architect for 3D implementations (utilize wide I/O, A & D, MEMS, passives)
- Higher level of integration allows reduced power and increased speed- Lots of memory & S/W to increase user friendliness and update systems- IP reuse and modularity to reduce development time and –cost
Add’l standards to simplify cooperation & drive economies of scale Investments in our design- and manufacturing EcoSystem
Semiconductor EcoSystem
Cooperation Within the “Food-Chain”
Complete Product
Package,Interposer,Substrate
WafersOS
Appl. S/WDrivers
Soft/Hard IPand/or KGD
Packaged3D stacks
EDADesign Tools/Flows
Foundry contributes:“2D” PDK,die COST,
TSV models, TCE data,thermal characteristics, …
Materials vendor:COST, Pkg dimension
thermal characteristics,…
OSAT contributes:COST projections for
assembly, test, & shipping.Production test requirements, …
Fabless/light vendor: Idea, architecture, targets for COST, Power, Performance, FormFactor, …
IP Partner contributes:“2D” license & royalties,
models of IP blocks.And for 3D:
Licensing & KGD COST,die-level models, …
Software Partner:H/W & S/W
co-design needs,memory footprint,
performance,power dissipation, …
EDA can play an essential role in building the 3D EcoSystem!
SYSTEMDesigners
SOFTWARE Designers
The Case for Standards in 3D Design
CIRCUITDesigners
ManufacturersFab, Assembly, Test
DesignRules,
Libraries,Models
DesignFiles and
TestPrograms
Design tools
Modeling tools
Standards for technical
hand-off criteria &business responsibilities
Standards to express
material characteristics &manufacturing capabilities
EDAVendors
Interoperability
3D/TSV Modeling & Design Steps
“3D”
PDK
Major 3D design steps to “hand-off”at fabless / fablight silicon vendors
3D System Designand Partitioning
3D-aware IC Layers ImplementationWafer, TSV,
Interposer,Passives &Package’s:Electrical,thermal &
mechanicalcharacteristics
“2D” PDK data
and
From FABs, OSATs and materials suppliers
Standards for modeling and data exchange formats between companies
Interoperability between tools
Die- & Block-level
IP Models
3D-Layers & Stack Verification for
Design “hand-off” To Manuf.
System Planning & PartitioningIncluding die-level IP Reuse
New EDA Opportunities in 3D EcoSystem
H/W Dev. Tools for: S/W Dev. Tools:
OperatingSystem
.
ApplicationsSoftware
.
.
Drivers.
.
Field Updates
Verification of new design portion(s) and entire system.Test program development
Hand-off to manufacturing
Logical & Physical 3D-aware Implementation
of new design portion(s)
EDA Opportunities in 3D Design
Modeling tools/flows to expand 2D PDKs to “3D PDKs” Capture electrical-, thermal-, magnetic-, mechanical characteristics,…
Modify proven 2D tools to make them “3D aware” PI, SI, noise, temp analysis, TSV strain, P & R ?,…
Create tools for 3D specific new challenges 3D DFT, die+stack+I/P+passives+pkg co-design, warpage, magnetics,…
Create system-level H/W planning and partitioning tools “Pathfinding” for lowest cost, lowest power, smallest formfactor,…
Create system-level hardware and software co-design tools System partitioning in H/W and S/W, emulating, debugging, updating,…
-------------------- Create tools for OS-, appl. S/W-, drivers development & debug
Multi- and many-core, handling of very wide busses, error detection and recovery, managing hardware redundancy (TSVs, memory, logic?, …)
A Closer Look at System Design
EDA vendors say about system-level tools:They have to be application specific and utilize a high-level of abstraction to offer significantly higher productivityJoint development of application-specific tools requires very good cooperation with the actual users at the EDA customersThe number of system designers is much smaller than the number of circuit designers
Points above are valid concerns, but keep in mind:As system complexity and value increase, the need for system-level tools increases beyond C++ to RTL,… large opportunity for EDA !Higher abstraction level tools will increase number of system designersToday’s EDA tools sell per seat for the contract period. Their price is not tied to the value they create New business model for system tools?System designers who implement large “2D SoCs” can benefits from 3D system tools as well increases user base and revenues for these tools !Systems’ software content is increasing and extends systems’ useful life significantly Tools for H/W & S/W co-design & S/W development needed!
3D Challenges & Opportunities for EDA
• Incremental minor model formats
• Incremental minor DRC decks
• Minor upgrades of 2D tools to recognize incremental features
• Standards for Design Exchange Formats
• Pathfinding tools and estimation methods, to make it worthwhile for users
• New Biz Model, to make it worthwhile for providers
Design Kits
include Design Rules DRC/LVS Decks Spice Models DFM Kits
Standard physical RTL 2 GDS design tools and flow
• Private domain of IDMs and system design entities
Interface toManufacturing
PhysicalDesign
SystemDesign
3D
2D
• Incremental EDA features and problems associated with migration to 3D• Top-down and very broad-brush perspective
• Key challenges are on the System Design Side of things
Source: Riko Radojcic, Qualcomm
Discussion Topics
When will 2½D cross the chasm? When will 3D …. ?
Which applications will use 2½D first ? // … 3D/TSVs first ?
Are EDA vendors jumping into the 3D drivers seat ?
How are IDMs vs fabless/fablight vendors approaching 3D ?
Would a different EDA biz model increase support for 3D ?
Can we jointly define open EDA standards for 3D or do we want to wait for de-facto (= proprietary) standards ?
What are bigger opportunities (vs 3D) to grow EDA revenues?
NEXT technologies: 3D monolithic ICs and Carbon Nanotubes