Sun releases their new processor as open source.
Read the press release on Jonathan Schwartz’s blog (Sun’s CEO).
Its definitely a first time a major producer of microelectronics has released their designs to the public under a free license (free as in speech), and one can’t but wonder what are the ramifications of such a bold statement. One might argue that the SPARC design has lost to its competitors of the x86 type a long time ago and the Niagra models (The nickname for the UltraSPARC/OpenSPARC T1 and T2) are just the death throes of an dying product line. Granted, the T1 and more so the T2 are indeed of the top performing chips available right now – if not the top, but it can’t be argued that they met with little commercial success.
Even given the situation where UltraSPARCs are not selling as well as they could, what could possibly Sun gain from “open sourcing” the design ? Jonathan Schwartz talks about selling processors to external integrators to use in their own server room computing products, but they could have done that without the GPL. Why here and why now ?
Its not as having an open source chip makes any difference to the average consumer directly – as open source software does. While the average Joe can’t be expected to compile and patch their software themselves, there are many many hobbiests that can do just that – you probably have a few in your own neighborhood – and there’s also many professional software developers that use open source in their everyday work as well as a hobby. The entry level for working on open source software is very low – anyone can do it and many actually do. The benefits to the average consumer when such an incredible amount of available work is leveraged toward open source software are immediately apparent.
Not so with open source processors. Even though Sun makes most needed tools for developing the OpenSPARC process available to the community for free (free as in beer, in this case), the entry for barrier is still incredibly high, just for helping with the development. But worse – unlike with open source software where almost anyone can compile and package a software product, actually putting a processor to silicon – which is the only way to get to a finished product with even an open source processor – takes billion of dollars and is something only a handful of companies in the world can do.
While its very unlikely that Intel or even a 2nd tier semiconductor manufacturer would try to produce and market OpenSPARC chips (and I’m very skeptic that Sun’s license for the OpenSPARC actually permits this), its quite possible that Sun’s competitors in the microprocessor industry will take some hints and copy design features from the OpenSPARC “source”. It has been done in the past many times in open source software and its very difficult to discover even when you have easy access to the product (in the case of software). The situation is far worse in hardware – if Intel would copy a good design for some sub-component of OpenSPARC for their new CPU, its very unlikely that this would ever be discovered, and even if its possible to detect such infringing use, there are ways to mimic important ideas expressed in code (source code or microchip blueprints) w/o actually violating any copyright license. And I’m sure Intel can be very careful.
So what does Sun has to gain from such a bold move ? Possibly better compilers that leverage the strengths of the UltraSPARC architecture – with better understanding of the underlying hardware, software developers can make better use of it, and that in turn may drive sells of the processor. But will the benefits outweigh the disadvantages ?