Audio/Video, Business, Hardware

Marseille Networks brings a new business model to the chipmaking world

In the world of consumer electronics one of your major costs is development. When developing a new chip you have to jump through a few hoops. Each one of these can cost you time and money. Let?s take a GPU for example; most of the time a GPU is designed on paper then it is taped out, set into test silicone and tested thoroughly. 9 times out of 10 the first runs will need to be tweaked [AMD’s Juniper i.e. Radeon HD 5700 was that 1 out of 10]. If it’s something minor you can reuse the same basic design and save a few bucks, but that is rare. Most of the time you are taping the GPU out again and then trying your tests over. These repeated cycles are very expensive a company get quickly run up costs into the millions of dollars just to put out a single consumer chip.  Even if you are a big company like nVidia or Intel that can hurt your bottom line and also your cost to customer.  If you are a small company delays and re-spins can bankrupt you.

So what do you do to avoid all this? You can dream and hope for a perfect product right off the drawing board, but as that is so rare you might as well wish to win the lottery while you are at it. Still all hope is not lost; there is something that can be done. One company in particular has done something that is not just smart, it is brilliant. The company I am talking about is Marseille Networks.

Marseille Networks is a relatively small and new company; but they are certainly a smart one.  Instead of plugging away using the old methodology of tape out, build silicon, send to customer, rinse and repeat Marseille has invested in a new method for developing and delivering their chips to customers.  This new method uses hardware virtualization. Now this is not the virtualization that you get with VMware, Citrix or any of the other popular operating system virtualization products.  This is a proprietary method for virtualizing the actual hardware [chipset, DSPs, etc.] in software while still maintaining their connectivity to existing I/O [input/output] technology for the device.

To get a good idea of the thought process behind this new [and brilliant] move we spoke with Marseille President and CEO Amine Chabane.

Our conversation was interesting Amine started out the discussion with the move to 4K TVs [including 3D 4K]. He spoke about the inevitability of this transition and, in very broad terms, talked about Marseille?s own developments in this market.  Then the talk shifted to their business model, as the company was founded by an engineer with experience in the industry Amine was able to walk into the market with a very solid idea of what works and what does not.

He realized that doing business "the same as everyone else" would not make for a successful company.  Marseille needed something to make them stand out from the crowd and also to prevent the catastrophic losses that can happen if there is an error in development [errors that can cost into the millions or $ for even a small redesign]. He then began the work on the virtualization of silicon.  

Now, designing a prototype in a virtual environment is not really a new concept. When you think about it cars, planes, and many other complex machines are developed and tested in a virtual environment long before they are ever built and put to the test. These models are subjected to simulated environmental conditions to make sure the design is as close to final as possible before the first weld is made.  This common practice in manufacturing has not really been used in development for consumer electronics. Computers are generally used to design the layouts of the silicon and the boards they are attached to, but for the most part once the design phase is done all testing and validation is done with the real thing. Amine and his team wanted to change this; they developed their own methods for virtualizing the Silicon while using the existing infrastructure.  Doing this means that they can refine and validate their designs before the first wafer gets pushed out.

This move has allowed them to make even major mid-stream changes without the expense of re-spinning a new chip. Marseille can make these changes in the virtualization system and send it right back out to their customers for validation with very little expense and time lost.  Amine told us that they have documented more than one incident where this method has saved them time and money during the development phase.

This new design model can benefit more than just emerging technology. As standards change using virtualized silicone can allow Marseille to test out these new technologies with existing hardware to see if any core design changes are needed, they can then implement them right in the model; again before heading to the FAB. The same is true for their customers; Marseille can allow them to validate their products using these systems without the need to stamp out a few wafers [and spending millions of dollars]. It truly is a brilliant move.  For those of you wondering; Amine and Marseille Networks have no plans for license their virtualization software or model at this time. We were told they have been approached with offers to buy the company and the software, but they have turned all of them down.

This kind of thinking is exactly what we need in the industry. With the big guys like AMD, Intel, NVIDIA, and Sony Dominating the market it is not easy to get a technology company off the ground. You have to have a new idea or a new way of thinking to make it. Marseille Networks have that. Their way of thinking by-passes the traditional way of doing things which in turn allows them to save money on their development cycle speed up their time to market and provide their customers direct usage of the product without the time and money expenditures normally associated with the development phase.

We are certainly going to keep an eye on Marseille; they have a great concept and look set to shake things up in the consumer electronics world. The list of their partners tells a bright story.