ESL may rescue EDA, analysts say

Gary Smith, chief EDA analyst at Gartner Dataquest, and Daya Nadamuni, senior analyst, both presented their views of the EDA industry at the annual Sunday evening Dataquest forecast meeting. Smith opened by saying that the two were using a "good cop, bad cop" strategy and that he was the "bad cop" giving the EDA industry a "reality check."

Smith said that the EDA business model has always been an outsourcing model, but people forgot that in the late 1990's. Because it's an outsourcing model, Smith said, in-house tool development both precedes and defines commercial tool development and pricing. "Customers really don't care how much value you [EDA vendors] bring to the design process," he said.

Further, Smith said, EDA tools today are undervalued because of an "incredibly stupid price war" among EDA vendors.

Smith also said that mainstream EDA users, who bought half of EDA tools in 2001, are getting out of IC physical design, and leaving that to ASIC vendors or consultants. These designers, said Smith, would really like to work at the electronic system level and sign off at the register-transfer level.

But Smith foresees a pickup in the EDA market after five years of nearly flat growth. He predicted EDA worldwide revenues will rise from $3.97 billion in 2005 to $7.46 billion in 2009.

Figure 1 — Gartner Dataquest projections of EDA industry revenue

Even so, Smith said, the use of in-house tools is growing. He said that 27 percent of designers use in-house tools today, up from 20 percent in 2001. One reason, he said, is the lack of available tools for the 65 and 45 nm process nodes.

"Tools should be on the market 18 to 36 months before the first production on the new node," he said. "We're late. We should have the 65 and 45 nm tools now, and we're just starting to see them."

Setting the stage for an optimistic view of the ESL market, Dataquest's Nadamuni said that there's a change in EDA methodology every 10 to 12 years. In the late 1980's, there was a shift to RTL design, and in the 1990's, a move to in-house IC physical design. Both drove EDA growth, which has since turned flat.

"What will it take to drive growth to double digits?" she asked. "It really seems like ESL is going to be the savior."

While ESL was around 10 years ago and didn't take off, there are differences today, she said. Nadamuni noted that the tools 10 years ago were domain-specific and industry specific and didn't offer much return on investment. Today, with 90 nm processes providing millions of gates, that return is potentially there.

The software challenge is a key driving force behind ESL, she said. And it's a tough one, with multi-processor systems-on-chip and multiple threads that must be programmed in parallel. The goal of ESL, she noted, is concurrent hardware/software design from a single high-level model.

Nadamuni predicted a significant upsurge in ESL revenues in the next few years, going from around $200 million today to $1.6 billion in 2009. At that point, it would be about equivalent to RTL tool revenues.

Figure 2 — Gartner Dataquest projection of ESL revenues

Nadamuni said that Dataquest has identified three major ESL methodologies, centered around algorithms, processors and memories, and control logic. Each includes behavioral-level and architectural-level design, and each is served with different tools and vendors.

A possible stumbling block to Dataquest's rosy ESL projections is the relatively low average selling prices (ASPs) of many ESL tools today. "You can't grow EDA on a $2,500 tool," Nadamuni said.