Deal allows ARM to create virtual prototypes for performance and power analysis of its cores.
ARM acquired Carbon Design Systems and its staff for an undisclosed amount of money, adding virtual prototyping capabilities for ARM cores.
The deal is the latest in an onslaught of mergers and acquisitions that have racked the semiconductor industry over the past 18 months, propelled by low interest rates and relatively low valuations of target companies. For ARM, this deal solves a growing problem for its IP, namely how to verify it in large, complex chips.
The functional verification of these chips has always relied on the usage of abstract models for the processors within the chip. These models enable large amounts of software to be executed in a reasonable time. However, when you add any significant amount of hardware, where the only models are RTL implementation models, the performance degrades very quickly. A virtual prototype integrates abstract models of all of the pieces such that performance can be maintained. The problem is where do you get all of the models and how are they integrated together?
Back in 2004, ARM acquired a virtual prototype development company called AXYS Design Automation, Inc. . At that time ARM was toying with the ideas of becoming a tool provider as well as an IP provider. It didn’t work too well for them and by 2008 they had decided that they needed to divest that group.
They chose Carbon Design Systems, a company founded in 2002 who had a tool that could create an abstract model from an RTL model. The fit of the two tools enabled Carbon to create virtual prototypes without having to create new abstract models for the parts of a design that were being reused. This had been one of the biggest barriers to virtual prototype adoption.
Since that time, Carbon has been one of EDA’s middle class. A stable, viable company that has a defined niche in the industry, is well respected, and makes money.
While Carbon was free to create virtual prototypes for systems other than those based on ARM processors, the dominance of ARM in certain industries has meant that Carbon kept this as a focus. Over time, Carbon has progressed from analyzing performance to adding power analysis as well. The other company who were also working on power optimization at the virtual prototype level was Docea Power, a company acquired by Intel earlier this year.
Last year (2014), ARM decided to get back into the tool business with the acquisition of Duolog Technologies Ltd., a company that specialized in tools for IP integration. Now they are adding to that with their asset purchase of Carbon and they also expect to take on all Carbon staff.
EDA’s middle class is quickly disappearing. Just this year, Calypto, Docea Power, Tanner EDA and Atrenta have been acquired and last year saw Forte, Nimbic, Berkeley DA and Berkeley Design Automation, Inc. acquired. With very few companies left for the larger EDA companies to acquire, they will be forced to up their R&D spending because they can no longer buy innovation, it will have to be developed organically.