Gary Smith once again wore his signature white coat and extolled the virtues of system level design and automation. The room at the Intercontinental was packed, and the folks in the hallway outside were noisy as usual, eager for the party to start after Gary finished.
System design - this is the new trend, all OEMs don't even manufacture any more. One large example is Apple (vertically integrated, but not really a manufacturer).
Changing Relationships - who is my competitor? Ideal - be as stable as ARM and have an ecosystem. Intel hasn't figured this out yet for SoC.
1996 DAC Slide predicting system level design - 20 years at DAC talking about system level design. Maybe I've missed by a few decades on ESL predictions.
System design needs new skills, like targeting vertical industries.
- Industrial/Medical (big enough for its own vertical)
- Mil Aero
Who is doing this vertical marketing well? Mentor is doing this with their automotive focus.
Each vertical requires an expert in something to set requirements, they come from that industry. Their name is mentioned and it influences that industry.
IP is key to productivity, lowered cost of design by 44% in 2011, the acronym impresses Wall Street. IP as a patent tool is still powerful.
10 types of IP now being categorized, there's quite a range available. Wall Street guys are calling GARY Smith and saying that EDA is a stable market now, no surprises, worth investing.
Physical IP - 2,400 (UART) to 7,999 gates
Library IP - bundled with tools, i.e. free with tools
DesignWare IP - silicon IP from Synopsys, very good job, #3 in IP sales
Large IP - 75K gates to 1,000,000 gates
Verification IP - adds SW for large IP
Functional Platform - compute, NOC, etc
Foundation Platform - snapdragon, i.MX
Applications - audio, GPS (bought from 3rd party IP ), etc. 1st generation uses bough IP, then 2nd generation uses own IP to eliminate royalty.
Average Platform design team is 50 people. Intel is too late to catch up to Qualcomm, because of momentum in cell phone market.
ARM has low power and an ecosystem, so how can Intel catch up to those two head starts? The old PC ecosystem was INtel and Microsoft, much simpler than the mobile markets today.
We have an ESL Flow! See EDPS 2014 session on this flow. Not perfect but it is working.
Architects are handing off the design for implementation and then lose control over their design intent. With an emulation system you can pass models between HW and SW engineers, but you cannot go back up to the architecture level. Emulation is too expensive for many teams at$2-4M. The ITRS cost chart shows most designs are spending under $50M, many under $25M. In 2011 emulation took a large percentage increase in budgets. You just need to get more engineers using each box, like 140 engineers on a single box now is possible. Emulation is a good bug hunter, accelerator and verification tasks.
Take your respin budget and use it on emulation instead, pay it forward and reduce respins. The emulation market is growing for emulation.
Forecast, $9B market in 2018 for EDA and IP combined, Lori does the forecasting now, not Gary.
IoT - it's just another Wall Street acronym.
Gary Smith EDA (GSEDA) is the leading provider of market intelligence and advisory services for the global Electronic Design Automation (EDA), Electronic System Level (ESL) design, and related technology markets.