Oracle APS

1
Oracle APS

• Mark Hopkins
• Manager, Operations Systems
• Xilinx, Inc.

2
Todays Topics

• Introduction to Xilinx
• Xilinx Supply Chain
• Drivers and Supply Chain focus areas
• Our APS Implementation
• Benefits and Lessons Learned
• Summary

3
Introducing Xilinx

• Leader in fastest growing semiconductor segment
• Invented programmable chip in 1984
• 7,500 customers, 50,000 design starts/year
• Leader in semiconductor process technologies
• First to 180nm, 150nm, 130nm and 90nm
• Enable hardware to change its spots
• Fastest time-to-market field upgradeability
• Pioneer of fabless semiconductor model
• Focus on design, marketing, support
• Partner for everything else
• A well-managed company and great place to work

4
Xilinx Revenue
1600
1400
1200
1000
800
600
400
200
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Fiscal Years
5
Xilinx Supply Chain
Distribution Mfg Network
End Market
Assy Test 14 Days
Wafer Fab 60-90 Days
Wafer Fab
Assy Test
Distributor
End Customer
EMS
Wafer Fab
Assy Test
Wafer Fab
6
Fundamental Supply Chain Drivers

• Xilinx products are generic until the customer
  programs them
• Customers require leadtimes 4 weeks or less, lots
  of upside flexibility
• Total Manufacturing Time is 75 Days
• Inventory is a significant cost due to yield
  improvements and technology advances
• Finished inventory costs 2x as much as inventory
  held 2 weeks back in wafer form
• Large BOM fan-out from 1 wafer type to multiple
  finished products (120 typical)

7
Xilinx Supply Chain Focus

• Supplier-facing Processes
• Integration of planning processes and
  communication with key suppliers
• Internal Planning Processes
• Inventory Management thru Replenishment,
  Build-to-Order, Postponement strategies
• Advanced Planning, decision support systems
• Customer Aligned Processes
• Better information for critical decisions
• Improved Delivery and Inventory

Todays Topic
8
Key Drivers for APS

• Capabilities of Existing Planning Applications
• Legacy systems didnt meet our changing business
  needs
• Complexity of Planning Environments
• 9 different planning modules involved
• 3 main sources Xilinx, Oracle, Legacy
• Support
• Legacy isnt supported
• Leverage Standard Oracle for Enhancements

9
Why Oracle

• Functionality in key areas for semiconductors
• Binning and Downgrades
• Wafer to Die Conversion
• Capacity Modeling
• Flexibility
• Sourcing Rules
• BOMs Can include masks, piece parts, preferences
• Integration
• Leverage data, interfaces in ERP BOM, Items,
  etc...
• Partnership
• Oracle is committed to Advanced Planning
• Learning from each other

10
Supply Chain Systems Flow
11
Whats In ProductionAPS and GOP

• Monthly / Weekly APS Plan for Wafers and Piece
  Parts (since April 03)
• EDD
• Unconstrained Plan
• Wafer Plan, Piece Parts, Assy, Test, Sort, Bump
  Capacity
• Daily Plan for Assembly Releases (since August
  04)
• ECC
• Output is used for all products assembly starts
  planning
• Daily Plan for GOP (since August 04)
• ECC
• Ramping all products to GOP by Jan 05 -
  Currently Scheduling about 40 of incoming sales
  orders

12
Our Implementation Timeline

• Source Data Maintenance
• Model Construction
• Monthly Planning Cycle
• Wafer Starts
• Piece Part Procurement
• Assy, Bump, Sort, Test Capacity Planning

• Daily Planning Cycle
• Material Constrained Planning (piece parts and
  die)
• Assembly Starts
• Mark Starts

Advanced GOP Assembly Starts Interface Vendor
Forecast Management
6/01 Phase 1 Begins
4/03 Phase 1 Ends
11/03 Phase 2.1 Ends
2/04 Phase 2.2 Ends
5/04 Phase 2.3 Ends
8/04 Phase 2.4 Ends
Upgrade and regression test to latest version -
11.5.9 Foundation for phase 2 Weekly Cycle for
Piece Parts

• Basic GOP
• Small portion of products

13
Our Model(example from our Assy Plan)
Plan Type Cost Based Optimized Plan - HP 64bit
Exe Enforce Material Constrained Plan Holistic
Manufacturing Plan with Project Planning Time
Buckets 35 days, 10 Weeks Inputs Orgs
56 Items 22k Demands 19k BOM's
17K Outputs Demands 64k Supplies
56k Pegging 78k Res Reqmts 200k
14
Customizations

• Operations Parameter Repository - Data
  Maintenance
• Demand Management Form - Viewing APS and OM Data
  together
• Schedule Execution System - Matching Wafers with
  APS recommendations
• Customer Forecast Management - Committing to
  customer forecasts

15
Benefit Unified Plan

• Before
• Separate Wafer, Piece Parts, Assy, and ATP Plans
• Separate plans for SJ and IR
• Today
• One integrated plan for wafer, piece parts,
  assembly, and order scheduling, one data source,
  one model
• Piece parts planning moved from spreadsheets to
  APS
• Example Orders are scheduled taking in to
  account piece part availability

16
Benefit Plan Quality

• Quality of Plan
• Enabled mask specific planning
• Better netting binning, downgrades, scavenging
• Better loading of factories alternate materials
  and locations
• Example wafer starts plan accounts for binning
  yields

17
Benefit Constraints

• Choices of constraints
• Unconstrained
• Constrained by Materials
• Constrained by Capacity
• Example Assy starts not recommended unless die
  and ppart are on hand. We are using APS daily to
  recommend source material and full routings for
  all production releases

18
Benefit Improved Customer Response

• Improved Customer Service
• GOP commits based on unified, high quality plan.
  Not necessarily more aggressive, but more
  accurate.
• Better visibility and linking of exceptions to
  sales orders - Pull in and Push out
• Real time order scheduling. Before Majority of
  orders scheduled within 6h. Today with GOP,
  Majority within minutes of order entry. Some
  orders will still require planner review and take
  longer.

19
GOP Date Acceptance By Week
85 of our line items are being scheduled by
GOP. Planners agree with GOP dates 72 of the
time. On Time delivery maintained at above 90
during rollout so far.
20
Lessons Learned

• Partnership is important
• Xilinx, Oracle Consulting, Sales, Support, and
  Development
• Absorbing deep technical functionality takes time
  and organizational commitment
• Understand the model, start small, and build on
  that
• Tradeoffs - Complexity vs Time to Run vs Ability
  to Comprehend
• Planners and Support Team require hands on
  experience
• Testing
• Functionality Testing, Full Data Suite Testing,
  Comparison Testing, Regression Testing
• Data Maintenance
• Data Generation / Management
• Data Quality
• Measure your results

21
Next Steps

• Stabilization and rollout are our number one
  priority we do have ideas about what to do next
  with APS
• Capacity constraints for test and assembly
• Enhanced BOM to accommodate revision specific
  planning (automotive, newer products)
• 24/7 ATP