Physical IP Group at ARM

After lunch on Monday I met with John Heinlin, Ph.D. – VP Marketing of Physical IP Division



Back in the day I knew the founders of Artisan (VLSI Libraries) when we worked together at Silicon Compilers (Mark Templeton, John Malecki, Scott Becker).


Q: Do you favor any EDA tools for creating your IP?
A: No, we don’t really endorse a specific EDA vendor tool or flow.

Q: What's new at ARM for 2011?
A: Just started Process Optimization Packs. Linking IP that is tuned for ARM cores.

Artisan – Generic cells. Used to wait for a node to become stable.

ARM – Focused on leading edge cells, now at 28nm nodes. Now doing test chips quite early to tweek and optimize cells, as requirements into the process.

2010 – 32nm last year announced.

2011 – 28nm IP is now ready as soon as announced. TSMC, SMIC, Samsung, IBM, GlobalFoundries (CP), TSMC.

Q: How to reach Power Performance and Area for your design?
A: We have a User Guide on how to get best results.

Q: Any preference for PDKs?
A: Support whatever PDK is available, no preference.

Synopsys – renewed tools agreement for a period of time.

Cell Library – about 1,000 cells to choose from, with about 100 different functions. Simple gates, flip-flops (power saving modes)

Q: What is your royalty model?
A: Royalty Model – same as always, no upfront costs. Foundries pay for library development.

Memory compilers cells – Standard RAMS, then lower VDD RAMs for an extra price. Typically 5 to 7 compilers per node.

Foundries – Some are offering their own libraries.

ARM Artisan IP (re-using)

Memory Compilers – Virage (Synopsys) has a piece of it. ARM invests heavily in these at leading edge nodes. Area efficient nodes too.

SOI – French based (SOISIC), acquired about 5 years ago, used at IBM.

Physical – GP IOs, 40nm Low Power. Will add DDR libraries soon. Have controller, PHy and IO in one piece.

Q: What do you think of Intel's TriGate?
A: FinFET – At 20nm planar CMOS still works, 14nm then FinFET looks viable.

Q: What is success for ARM?
A: Success – We think rich solutions across a wide range of foundries at 28nm now, 20nm soon (test chips with Samsung).

Q: How do you use RDR?
A: RDR – We’ve been dealing that for a few nodes now (20nm needs double patterning). We work closely with foundries for all nodes.

Common Platform – ARM is the IP you get.

Q: Where do I find out more?
A: Visit our SOC Design Community