Package, Assembly and Thermal Challenges for Future Microprocessors Corporate Fellow Chief Technologist C4, Packaging and Back End Technologies 1 Scope Flip Chip Package Technology and Manufacturability Assembly Technology Thermal Low K Challenges 2 www.cpmt.org/scv/ 1
Defying Laws of Nature Lid Si Organic Substrate C4 TIM1 Low K Films Underfill Lid Adhesive Loading scenario I: (PKG level) Si/Substrate CTE Mismatch -Die Bowing -Underfill shearing/peeling -BEoL film CPI Heat Sink Loading Scenario II: (Platform level) Socket insertion/heatsink clamping -Die flattening -Possible Die/Lid separation -System over-rigidity Socket/Motherboard Technology and Challenges Overview Flip Chip Bumping Pitch Decreasing Bump count increasing Max. current increasing C4 Probe package Package Technology at limit C4 probing High I/O count solution?? Thermal Power Going up Thermal solutions at limit Cost $$$ Full Low K Current organic technology broken underfill, sawing optimization Ceramic - Current limited Alternate technologies needed Low K adhesion needs improvement µp RoHS Compliant BGA Passives Package Technology Construction 2/2/2-1/4/1 2/4/2 4/4/4 No stack/2 stack/4 stack vias Supplier capacity Cost Pitch limit Manufacturability Ceramic coplanarity with increasing C4 Solder Printing for pkg. C4 pads Design features at limit Assembly Fine Pitch Increasing bump count New UF materials/processing Passives smaller, more in number Low K Underfill, sawing, stress mgmt. 4 www.cpmt.org/scv/ 2
Flip Chip Challenges Pitch Decreasing from 350 µ --- 150 µ --?? No. of bumps increasing --- 5 10 K per die Bumping Choice Electroplating --- Uniformity, extendibility, low defect rate Screen printing --- Limited by Solder paste, stencil printing, voids Current Carrying capability Current Density increasing - limit pitch decrease Bump and die at UBM via opening Joule Heating significant Choice of Solder Eutectic Solder not suitable for EM High Lead solders should be used Alternate bumping/interconnection technologies required Bump less Pillar bumping Compliant springs, Nano links 5 Bump Geometry 6 www.cpmt.org/scv/ 3
Bump Considerations Joule Heating (constant via dia ) Joule Heating ( deg.c ) 80 70 60 50 40 30 20 10 0 Bump Current Imax (ma) via size effect 125 µm 150 µm 100 µm 0 0 10 20 30 40 50 via size (um) 7 Max Bump current -- Electro Migration High Lead Solders Imax ma (11.4 years) Imax ma (7 years) Imax ma (5 years) Imax (ma) 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 sigma Max Current = Joule Heating + EM 8 www.cpmt.org/scv/ 4
Package Choice Criterion Electrical Performance Cost Consistent supply of packages Reliability 9 Packaging Choices Reliability Board Level C4 Board Level Organic ----- CTE = 18 Solder Columns for ceramic High TCE Ceramic ---- CTE = 12 (Single Source) C4 Organic www.cpmt.org/scv/ 5
Package Selection Ceramic Advantages Mature Technology Multiple mature suppliers Higher moisture resistance Closer TCE match to Si Low K Large die Higher mechanical strength Thermal solutions with higher load Higher reliability Disadvantages Performance << organic Mo vs. Cu conductor Higher dielectric constant Pitch limited to 200 m at best for high volume Cost has matured Package Selection Organic Advantages Higher Performance Copper conductors Lower dielectric constant Cost cheaper then ceramic Design ground rules advanced and more manufacturable for higher performance designs Multiple reliable Suppliers Disadvantages Moisture sensitivity Lead free High TCE mismatch Low k Large die Sensitivity of assembly processes to surface consitions Lower mechanical strength Lower reliability www.cpmt.org/scv/ 6
Organic Structure C4 bump solder Insulator ; Film insulator PTH filling Core Solder Resist Pin solder Pin 13 Package Challenges Increased functionality and larger die Finer Pitch Ceramic. 200 µ Organic. 180 150 µ Larger body sizes to route Ceramic.. Finer features not manufacturable Organic Finer features possible requires material and process changes Need for increased decoupling capacitance Higher cost. Larger body size Embedded capacitance breakthrough needed Suppliers have different material sets 14 www.cpmt.org/scv/ 7
Package Complexity@ fine pitch Solder Printing Printing solder at fine pitch approaching limit Jetting?? Plating?? Ball Attach??? 15 Key challenges 2/4/2 4/4/4 4/4/4 Crack Unstack 3 stack 4 stack Build up materials -TCE - Temperature resistance for the solders used PTH material - TCE excessive z motion can cause delamination cracking 16 www.cpmt.org/scv/ 8
Package Manufacturing Challenges Fine Pitch and Large no. of I/O require Finer features Body Size Routing Drives substrate manufacturing to limit New Manufacturing processes required Cost pressures make it difficult Embedded capacitance???? 17 Package Design Challenges Core Current rules (um) Future rules (um) Impact min PTH diameter 300 200 Finer drill & thin core min PTH pitch 600 400 Finer drill & thin core Line width on core 90 50 New core patterning max # of core Cu layers 4 6 New process Build up Micro Via diameter 85 50 Need UV laser Fine line/space 20/20 15/15 New dry film/pattern process # of cu layers 2 4 Reliability/cost/capacity Stacked vias 3 4 Reliability risk Solder mask opening 90 60 New s-mask/higher resolution 18 www.cpmt.org/scv/ 9
Assembly Chip Placement Finer pitch drives higher accuracy Slows down UPH Large die and finer pitch Stringent bump and package co planarity New fluxes required Low residue for larger area Compatibility with under fills Oxide removal 19 Die footprints Die and Bump counts getting larger JH-F (90nm) JACKHAMMER CH 7.98 x 8.71 2821 CLAWHAMMER DRILLHAMMER SLEDGEHAMMER DH 9.71 x 8.59 3490 SH 12.89 x 8.75 4676 JH 13.1 x 15.5 8876 JH 13.1 x 17.5 10000 What will they think next!!!!!!! 20 www.cpmt.org/scv/ 10
Assembly UF materials Finer particle size for lower standoff as pitch decreases Solids loading decreases due to high surface area Effects physical and mechanical properties Optimized physical / mechanical properties Large die Low K UF processes Finer P.S.D. slows down the conventional UF dispense Sensitivity to Solder mask surfaces increases New approaches for high volume manufacturing?? Novel dispensing methods Jetting Multiple needles ( 5-10 simultaneously ) No Flow UF?? Wafer applied UF?? Defects become more critical 21 Passives Roadmap 6.0 300 Mounting Area (mm 2 /pc) 5.0 4.0 3.0 2.0 1.0 Mounting Area Mounting Density 250 200 150 100 50 Mounting Density (pcs/cm 2 ) 22 0.0 Courtsey: Murata Corp 1206 0805 0603 0402 0201 Case Size 0 www.cpmt.org/scv/ 11
Assembly Passives attachment Numbers are increasing Evolution SMT Multi terminal LLA, IDC Resistors Smaller sizes Processes and defects Dispensing at limit. Shorting,tombstoning;unbalanced fillets Printing??? Embedded Package capacitance required 23 Thermal Issues The 2002 ITRS says power increases gradually and die size remains constant at ~300 mm 2 : But that s not the whole story POWER DENSITY (W/cm 2 ) POWER (W) 300 3000 250 2500 200 2000 150 1500 100 1000 50 500 130 nm 90 nm 65 nm 45 nm 32 nm POWER (W) Power (W) Power density (W/cm 2 ) 130 nm 90 nm 65 nm 45 nm 32 nm 3000 2500 2000 1500 1000 500 POWER (W) 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 20160 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 YEAR YEAR because the power dissipating area (core) continues to shrink! (and as # of transistors increases, so does total switching current required!) 24 www.cpmt.org/scv/ 12
Thermal Interface Materials Traditional Curable Grease / PCM with Solid Fillers Ease of Application Understood Reliability Low Cost Thermal Resistance does not Meet Future Requirements Solder Low Thermal Resistance Best Thermal Resistance Uniformity Concerns with Temperature and Power Cycling Reliability Surface Preparation and Added Manufacturing Steps Liquid Metal Hybrids Low Thermal Resistance Needs Containment Poor Understanding of Reliability Materials Under Research Carbon Nanotubes Graphite Derivatives 25 Thermal Issues: Solutions Power density cannot continue to increase at this rate, but in the near-term we need to address up to 500 W/cm 2 (Hotspot cooling)! Very thin die to reduce thermal resistance through the Si New thermal interface materials (TIMs) High thermal conductivity metallic fillers Phase change filler materials with high heat of transformation Low contact resistance (wetting on difficult surfaces) High elasticity to account for mechanical stress Refined surfaces at thermal interfaces Polished to reduce asperities and contact resistance Reactive to improve wettability of TIMs New heat spreader materials Maybe diamond finally has a chance New cooling technologies Heat sinks and fans losing efficiency (increasing volume/watt removed) Liquid cooling with miniaturized pumping systems Need to bring cooling technology directly to the die 26 www.cpmt.org/scv/ 13
Constrained Power Density Projection: Limitations of Thermal Packaging POWER (W) 300 250 200 150 100 50 0 Solid Filler TIM1 Multi-Core Processing Application Specific Processing POWER (W) Solder Liquid Metal TIM1 130 nm 90 nm 65 nm 45 nm 32 nm 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 YEAR Die Thinning? Alternate Handle Substrate? Direct Die Cooling? the power density becomes independent of feature size and is restricted by available packaging technology 500 400 300 200 POWER DENSITY (W/cm 2 ) 27 Thermal Issues: Heatsink and Fans Thermal Resistance (C/W) 1.5 1 0.5 Healing Length Model AAVID Dual-Fan Sinks Power ~ 3 W Noise > 35 db COFAN KC- Commercial Heat Sinks With Top -Mounted Fans Fan Power ~ 1 W Noise ~ 25 db INDEK COFAN KF- Courtesy Dr. Ken Goodson 0 0 10 20 30 40 Heat Sink & Fan Volume (cubic inches) Half the volume Of a laptop computer 28 www.cpmt.org/scv/ 14
Thermal Issues: Direct die cooling (liquid) 29 Micromachined Pump-Driven Two-Phase Capillary Evaporator Cooling Loops Designs Energy storage unit with phase change material (PCM) overcomes power surges (CMU) Electrokinetic pumped cooling system enables highpressure evaporators, eliminates moving parts (Stanford) Vapor chip Courtesy Dr. Ken Goodson Multilayer micromachined Electrokinetic Liquid Pump evaporator increases capillary pressure (U. Maryland) Condenser Region Microchannel Evaporator + V - High-Performance Evaporators Energy Storage Unit: Phase Change Material vapor liquid Fully micromachined evap/condenser (UC Berkeley) Impingement structure Ni Kovar chip pump Condensor Exchanger Single-layer optimized wicking structure (Sandia / Texas A&M) Low-K Overview Ceramic Package Alumina ( Std) Better CTE match for full Low K Reliability data supports Fine pitch problem Performance marginal Glass Ceramic (LTCC) / Cu Not ready for high volume Cost= 1.5-2X current Low K Organic Package Organic Current organic technology broken for full Low K Die adhesion improvement Key Low CTE package required No successful approaches Cost=1.5-2X current package Assembly Current state-of-the-art 150µ Laser saw possible Assembly Laser saw needed Full low K drives material and process solutions stress management 30 www.cpmt.org/scv/ 15
Technology and Challenges Overview Flip Chip Bumping Pitch Decreasing Bump count increasing Max. current increasing C4 Probe package Package Technology at limit C4 probing High I/O count solution?? Thermal Power Going up Thermal solutions at limit Cost $$$ Full Low K Current organic technology broken underfill, sawing optimization Ceramic - Current limited Alternate technologies needed Low K adhesion needs improvement µp Lead Free BGA Passives Package Technology Construction 2/2/2-1/4/1 2/4/2 4/4/4 No stack/2 stack/4 stack vias Supplier capacity Cost Pitch limit Manufacturability Ceramic coplanarity with increasing C4 Solder Printing for pkg. C4 pads Design features at limit Assembly Fine Pitch Increasing bump count New UF materials/processing Passives smaller, more in number Low K Underfill, sawing, stress mgmt. 31 Emerging Thermo-Mechanical Issues for High Performance uprocessor Flip Chip Packages Silicon: Increasing die size (multicore CPU) Increasing metal levels Mechanically weaker, fragile BEoL low-k dielectrics Package: Increasing power consumption presents harsh thermal stressing; heat spreader (lid) reduces flexibility C4 pitch shrink; Lead-free implementation Aggressive organic substrate technology (thin core, coreless, tight routing) Platform: Thermal preloading (socket insertion, heat sink clamping) adds more constraints to package/die MCM (multichip module) introduces die to die interaction and complicates the socket/package interaction Failure Modes: Chip-package interaction, BEoL film delamination Underfill delamination C4 fatigue Underfill voiding and C4 shorting TIM degradation and Thermal run-away List 32goes on. www.cpmt.org/scv/ 16
Skill Set Required Material Science Chemistry Physics Mechanical Engineering Thermal Engineering Fluid Dynamics Analytical techniques Surface Science Adhesion Science Metallurgical Engineering Mathematics Electrical Engineering Fracture Mechanics Modeling 33 Summary Approaching limit on many technology fronts Conflicting direction Thinking outside the box required Collaboration between Users, Suppliers and Academia 34 www.cpmt.org/scv/ 17