As I have already written more than once, the 20 nm process is introducing a new reality in IC design and manufacturing. This is just a preview of how much harder things are going to be at 14 nm. Third party confirmation did not make me wait long. A press release from Synopsys contained the following statements.
A few weeks ago I had the opportunity to visit Blue Pearl Software, one of the emerging EDA companies in Silicon Valley. Blue Pearl Software was started six years ago with the goal of doing design analysis at the functional level and providing designers with control of their design from RTL through synthesis. Another goal as important as the first was to automatically generate timing constraints and allow the verification of their correctness.
Someone's numbers do not make sense
Just a day after the news of the proposed acquisition by Synopsys, Magma reported results for its second quarter of the fiscal year (May to April). The data shows a company that is rebounding well from its 2009 fiscal year low. The significant dip in Magma's performance for fiscal 2008 and 2009 has been attributed by almost all analysts to the marketing and sales problems the company experienced while embroiled in the legal battle with Synopsys over intellectual property rights.
The behavioral scientists that analize the EDA industry were sure of one thing above others: there would never be a deal between Synopsys and Magma. Their wisdom said that Aart and Rajeev despised one another, there was too much bad blood between the two companies after the legal action by Synopsys pushed Magma to the brink of oblivion, and Magma was a small player in the EDA business that could never again be a serious contender for number 3. The two companies could never do business together was the strongly held belief.
When reading the Silicon One paper from Magma Design Automation, one is struck by the style of the document. In the EDA industry we are accustomed to technical papers that assume strong technical knowledge of the semiconductors industry. We talk to ourselves expecting that those outside our industry will somehow appreciate both the complex problems and the elegance of the solutions. Magma does no such thing. Its audience, which includes financial and business professionals, is given an accurate, yet understandable description of the requirements of our industry and the possible solutions.
The first paragraph of the official release from Apache promises new help for designers. Power consumption is not just a problem for portable devices, but, translated in heat, is also a major issue with electronics that remain plugged to a socket. Therefore one cannot but rejoice from the announcement. Or can we? Based on the contents of the press release, many doubts remain starting with the title that implies a new model of not just power distribution but also a dynamic impact of power consumption on the circuit behavior.
The acquisition by ARM of Prolific underscores, if there was still a necessity, the difficulties of bringing to manufacture designs targeting the 20 nm process. It has become clear that the cooperation between EDA tools suppliers and design houses must be practically air tight in order to succeed. Let's not forget that just a few weeks ago ARM had announced the successful tapeout of its Cortex-A15 MPCore achieved with the close collaboration of TSMC and Cadence. The experience certainly thought them something and I believe the Prolific acquisition is one of the results.
As many know, Accellera and OSCI are in the process of merging. The work on the merger, announced in June, should be completed by the end of the year. Concluding legal matters takes almost as long as developing a standard. Through all of the complexities that the representatives of the two organizations are dealing with, one stands out for its negativity. The desire, by some of the people involved, to develop a new name for the resulting organization. What is proposed is truly bad business practice.
There are two interesting items in the most recent email letter from John Cooley.
One is about the attention given to Steve Job's passing versus that of Dennis Ritchie. The other is a note by Andrew Stevens observing that C/C++ and SystemC are not capable of describing parallel computation.
Today I received three different press releases about a tapeout at 20 nm. The first important item about the releases is that what is announced is a tapeout. Not a successful fabrication or even more remarkable the achievement of commercially viable yield levels. When a tapeout becomes news, the relevant news is that it is worthy of a press release at all.