Synopsys, ARM, Samsung, GLOBALFOUNDRIES (Part 1 of 2)

Synopsys, ARM, Samsung, GLOBALFOUNDRIES (Part 1 of 2)
by Daniel Payne on 06-14-2011 at 12:26 pm

Intro
The 28nm nodes is ready with foundry silicon, IP and EDA tools. Tuesday morning at the DAC breakfast I learned more about the 28nm eco-system.

Notes
Why 32/28nm
Lower power, high integration requirements, mobile applications

What is Ready?
IP is qualified (ARM, Memories, Foundation IP, SNPS IP, PDKs)
August 2010 SNPS and GLOBALFOUNDRIES at 28nm
June 2011 SNPS and ARM at 28nm (A15 core)
June 2010 Samsung at 32nm with SNPS tools
Common Platform – Lynx tool flow is ready, January 2011
June 2011 GLOBALFOUNDRIES ready at 28nm
Samsung qualifies 28nm
Samsung at 35 tape outs at 32nm to date

Anna Hunter, VP Samsung
Technology Roadmap
32nm LP: ready, HKMG process
o SRAM at .149um*um, tiny size
o Good yield at 86%
o Matches SPICE results
28nm LP: ready
o Same HKMG as 32nm node
o Works with ARM IP and SNPS tool flow
28nm LPH: under development (low power, plus higher performance modules)
o Will be up to 50% faster (with more leakage, 2.3X)
o Same HKMG
o Added strain to silicon
o Shuttles starting now
20nm LPM: in development, PDK evaluation now. Ready by end of 2012.

Lynx – flow of SNPS tools and IP management, used by Samsung internally too

ARM CPU – 45nm >1GHz on Cortex A9
32/28nm >1.35GHz on Cortez A15
28nm LPH, >2.0GHz Cortex A15

IP Portfolio – High Speed, Memory, Mixed Signal
ARM, SNPS<

Going from 45nm to 32nm more than 50% improvement in SRAM bit cell size

Turn key solutions from Samsung
Design, Fab, Wafer Sort, Assembly, Final Test
Working on TSV technology for higher integration on packaging

MPW – Run every quarter for 32nm and 28nm
Will start 20nm in September

Fab sites – Korea( 20nm), Texas (40K wafers per month)

Jim Ballingall, VP Marketing at GLOBALFOUNDRIES
AMD lead product used HKMG technology, quad core CPU with GPU integrated, 500GFlops, for notebooks
Llano powered laptops later in June

Super Low Power – 28nm SLP (doesn’t use stressing), about 2.3GHz

High Performance Plus – 28nm HPP (uses stressing), about 3.1GHz

Global Solutions – Design Solutions, Technology, Design Infrastructure<

IP – in place

Fabs – New York, Germany, Singapore

MPW – 4 shuttles in 2011

20nm – working with Common Platform partners, area scaling of 50% from 28nm


Comments

There are no comments yet.

You must register or log in to view/post comments.