3D IC Design Tools and Applications to 3D IC Design Tools and Applications to Microarchitecture Exploration Microarchitecture Exploration Jason Cong Jason Cong UCLA Computer Science Department UCLA Computer Science Department cong@cs.ucla.edu cong@cs.ucla.edu http:// http://cadlab.cs.ucla.edu cadlab.cs.ucla.edu/~cong /~cong Outline Outline � Thermal Thermal- -Aware 3D IC Physical Design Flow Aware 3D IC Physical Design Flow � � Thermal Models and Assumptions Thermal Models and Assumptions � � 3D Routing with Thermal Via Planning � 3D Routing with Thermal Via Planning � 3D Placement � 3D Placement � 3D � 3D Floorplanning Floorplanning � 3D Architecture Exploration 3D Architecture Exploration � � 3D Component Modeling and Testing 3D Component Modeling and Testing � � Concluding Remarks and Future Work Concluding Remarks and Future Work � 2 Page
Thermal Challenges in 3 Thermal Challenges in 3- -D ICs D ICs Temperature distribution along z direction � Key Challenge of 3 Key Challenge of 3- -D IC D IC T � 150 o C Design: Design: 135 o C � Higher power density � Higher power density 100 o C � Inter Inter- -layer dielectric layer dielectric � layers layers � High Temperature High Temperature � 30 o C Effects: Effects: Z Si 1 � Longer interconnect Longer interconnect Si 2 � Si 3 Si 4 delays delays Temperature increases � Functional failure Functional failure � dramatically along the z direction 3 3 3- -D IC Cooling Schemes D IC Cooling Schemes � Heat Sink Optimization Heat Sink Optimization � � Air cooling fans Air cooling fans � � Heat radiating fins � Heat radiating fins � Thermal grease, AC, etc � Thermal grease, AC, etc .. .. � Chip Chip- -Level Temperature Level Temperature � Optimization Optimization � Microchannel � Microchannel cooling cooling � Floorplanning Floorplanning � � Routing Routing � � Thermal via insertion Thermal via insertion � 4 Page
Thermal Thermal- -Aware 3D Physical Design Flow at Aware 3D Physical Design Flow at UCLA (2002 – – 2005) 2005) UCLA (2002 Technology Design constraints Technology Netlist (LEFDEF) (LEFDEF) Design constraints Netlist Thermal- Thermal -Driven Driven 3D Floorplanner 3D Floorplanner Thermal Thermal Compact Thermal- Thermal -Driven Driven Compact Simulation Simulation Compact Open Open Thermal Thermal Open Thermal 3D Placement 3D Placement model model Access Access model Access Timing Timing Thermal- -Aware Aware Analysis Analysis Thermal 3D Router w/ 3D Router w/ Thermal Via Planning Thermal Via Planning Parasitic Parasitic Extraction Extraction CIF/GDSII CIF/GDSII Layout Layout Verification Verification 5 3D Physical Design Flow (IBM, UCLA, and PSU) 3D Physical Design Flow (IBM, UCLA, and PSU) (2006 (2006 – – present) present) Layer & Cell & Via* Netlist Design Rules definitions (HDL or DEF) (LEF) (LEF) PSU PSU UCLA UCLA Thermal Thermal- -Driven Driven 3D RC extraction 3D RC extraction 3D Floorplanner 3D Floorplanner 3D OA 3D OA Tech. Lib Tech. Lib Timing Timing Thermal- Thermal -Driven Driven EinsTimer EinsTimer Ref. Lib Ref. Lib 3D Placer 3D Placer Interface Interface Design Design 3D DRC & 3D LVS 3D DRC & 3D LVS 3D Global Router 3D Global Router Tier Tier Thermal Thermal- -Via Planner Via Planner Export Import Detailed Routing Detailed Routing Layout 2D OA 2D OA by Cadence Router by Cadence Router (GDSII ) 10/8/2007 UCLA VLSICAD LAB 6 Page
Thermal Resistive Network [Wilkerson04] Thermal Resistive Network [Wilkerson04] � Circuit stack partitioned into tiles P 5 5 R 5 � Tiles connected through P 4 4 thermal resistances R 4 P 3 3 � Lateral resistances: fixed R 3 P 2 2 � Vertical resistances ∝ 1/#via R 2 P 1 1 R 1 � Heat sources modeled as ± current sources - R lateral � Current value = power ( a) Tiles stack (b) Single tile array stack � Heat sinks modeled as ground nodes Accurate and slow 7 Thermal Resistive Chain Model Thermal Resistive Chain Model � One One- -Dimension Heat Flow Analysis Dimension Heat Flow Analysis � � � Elmore delay Elmore delay- -like formula [Chiang01] like formula [Chiang01] P 4 4 4 4 ∑ ∑ = T ( R P ) R 4 4 i j P 3 3 = = i 1 j i R 3 P 2 4 i 2 ∑ ∑ = T ( P R ) R 2 4 i j 1 P 1 = = i 1 j 1 R 1 ± � Reduce R: thermal via insertion (routing) - � Permute P: floorplanning Fast and rough 8 Page
Through Through- -the the- -Silicon Vias (TS Silicon Vias (TS- -Vias) in 3D ICs Vias) in 3D ICs Pad Metal Routing Block 1 Block 2 Layer Dielectric Through-the-Silicon Via Layer (Signal TS Via) Block 3 Through-the-Silicon Via (Thermal TS Via) Block 5 Block 4 Silicon (Device Layers) � Effective in heat dissipating Effective in heat dissipating � � Regular wires have almost no effect (size/direction) � Regular wires have almost no effect (size/direction) � Two types of TS Two types of TS- -vias vias � � Signal TS � Signal TS- -vias, part of the vias, part of the netlist netlist � Thermal TS � Thermal TS- -vias, with no connections, introduced to reduce vias, with no connections, introduced to reduce temperature temperature 9 Thermal- -Aware Aware 3D Routing Problem 3D Routing Problem Thermal � Input Input � � 3- -D floorplanning (placement) result D floorplanning (placement) result � 3 � � Technology Technology � � Netlist Netlist � Required temperature, such as 80 O O C C � Required temperature, such as 80 � Output Output � � � Routed nets Routed nets � � Thermal TS- Thermal TS -via number and locations via number and locations � Objectives Objectives � � Minimum wirelength � Minimum wirelength Minimum TS- -via number via number � � Minimum TS 10 Page
Multilevel TS- -Via Planning and 3D Routing (TMARS) Via Planning and 3D Routing (TMARS) Multilevel TS G 0 (1). Power Density Calculation Thermal Resistive G 0 Network Model (2). Heat Flow Estimation (3). Routing Resource Estimation G i level 0 level 0 G i (1) Routing Refinement (1). Power Density Coarsening (2). TTS Via Planning (2). Heat Flow Estimation (3). TTS Via Number Adjustment G k (3). Routing Resource Coarsening level i level i Upward Pass Downward Pass (1). Init Routing Tree Generation (2). TTS Via Planning (3). TTS Via Number Adjustment level k 11 Thermal TS- -Via Planning Problem Via Planning Problem Thermal TS Determines the thermal TS via density for all tiles Determines the thermal TS via density for all tiles � � Minimizing #total thermal TS via Minimizing #total thermal TS via � � Meeting capacity and temperature constraint Meeting capacity and temperature constraint � � Solving through Solving through � � � Via planning proportional to ∆ ∆ t t (VPPT) (VPPT) � Via planning proportional to • • ∆ ∆ t: vertical t difference t: vertical t difference � � Alternating direction via planning (ADVP) Alternating direction via planning (ADVP) 0 3 4 a 5 8 ∆ t a =t a -t b 1 6 2 5 b 0 10 8 12 Page
Thermal TS Via Planning [Cong & Zhang, ICCAD Thermal TS Via Planning [Cong & Zhang, ICCAD’ ’05] 05] Non- -Linear Programming Formulation Linear Programming Formulation Non � Variable Definition, for tile Variable Definition, for tile L L i, j, k � i, j, k � a i,j,k : TS-via number � R i,j,k : vertical thermal resistance � P i,j,k : current source � Ύ : constant R i,j,k = Ύ / a i,j,k � t i,j,k : temperature I i,j,k : heat flow � Objective Objective � γ N I ∑ ∑ t i,j,k = = i,j,k #total _ via a − i,j,k t t ≥ − � Constraints Constraints k 2 i,j,k i,j,k 1 � I i,j,k � Capacity constraint Capacity constraint � � Temperature constraint � Temperature constraint R i,j,k = Ύ /a i,j,k � Kirchoff's � Kirchoff's current law current law � Constrained NLP Constrained NLP � � Can be solved by general NLP solver Can be solved by general NLP solver � � But very time consuming But very time consuming � Fixed R ± 13 Alternating Direction TS Alternating Direction TS- -Via Planning (ADVP) Via Planning (ADVP) � Decompose the NLP into simplified sub Decompose the NLP into simplified sub- -problems problems � � Optimizing the via distribution at one direction at a time Optimizing the via distribution at one direction at a time � � � Alternating between vertical via planning and horizontal Alternating between vertical via planning and horizontal via planning at each level via planning at each level � Updating the heat flow after every step � Updating the heat flow after every step 14 Page
Recommend
More recommend