When the news broke last week about Oracle’s intent to buy Sun Microsystems, my thoughts turned to the Sun design and verification engineers I’ve spoken to in recent years. Will Oracle, a newcomer to the world of EDA and chip design, keep that team?
The outcome depends on two questions – Oracle’s plans for the Sparc architecture, and Oracle’s views about keeping hardware design in-house. I think it’s in Oracle’s best interests to keep Sparc and keep the team.
From all indications, Oracle does intend to keep Sun’s hardware business. A FAQ document at the Oracle web site states that “Oracle plans to grow the Sun hardware business after the closing” and notes that “key to this strategy will be our plans to develop software-optimized hardware that integrates all of the enterprise components: hardware, database, middleware, and applications.” (Interestingly, Oracle entered the hardware business in September 2008, when it announced two servers developed in conjunction with Hewlett-Packard, a company it will now apparently compete with).
The server business is tough, and Sun’s differentiation has rested with its Sparc processors. Will Oracle continue Sparc development? A recent EE Times analysis article noted a few of the downsides:
Maintaining the proprietary Sparc processors represents a significant cost
- Sun is a year late with its next-generation Rock processor
- Sparc faces tough competition from Intel’s Nehalem processor
- Sun just lost one of its lead microprocessor architects, Marc Tremblay, to Microsoft (what Microsoft will do with that expertise is another question).
As also noted in that article, however, it would take a long time to wind down Sparc, and Sun has made considerable investments in its next-generation Niagara and Rock processors. Oracle should think long and hard before throwing that investment away. Keeping Sparc, and keeping Sun’s processor expertise, might be the best way to develop highly differentiated machines that are optimized for running Oracle software.
So what about keeping Sparc, but outsourcing some or all of the chip development work? In my view, that’s not a good idea. Consider:
Processor design is specialized work that requires considerable expertise, as well as a state-of-the art CAD methodology (and Sun is on the leading edge of verification technology, based on discussions I’ve had with people there). It’s one of the last things you would want to outsource in IC design.
- Outsourcing portions of the chip design flow is really tough. How do you decide what to outsource and what to keep? Do you keep the RTL team and outsource the layout, or do only high-level chip architecture and depend on somebody else for the rest? What do you hand off and to whom?
- There’s an information gap between what is done in-house and what is done externally. If IC layout is outsourced, for instance, the RTL team doesn’t get the same kind of instant feedback and guidance they would receive from people in their own company. For processor design, that could be crucial.
It is certainly true that the systems business is becoming more software-centric. However, Sun’s hardware design expertise is part of the core competence of the company, just like Java and Solaris. I hope Oracle recognizes that.