Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
Using SIP: How they know what they know
February 14th, 2013 by Peggy Aycinena
After the euphoniously monikered IP provider, Uniquify, announced several weeks ago that the more whimsically monikered organization, Pixelworks, is using Uniquify’s DDR memory controller subsystem IP for multiple distinct processors that Pixelworks is, in turn, providing to TV makers who make 4Kx2K ultra high-def systems, one question still remained: How did Pixelworks know to use Uniquify’s offering?
According to a January 2013 article in IEEE Spectrum, knowing what IP to use in a project here in the 21st century is fairly easy knowledge to come by. I don’t know what planet the author of the op-ed piece, “Other People’s Knowledge”, lives on but it doesn’t seem to be the one that I hear about from the folks who make or buy third-party IP.
In fact, those people seem to indicate that knowing what IP to use in a particular project continues to be far more art than science. In particular, because until a system, or sub-system, is fully defined, modeled and simulated – let alone, manufactured and deployed in the field – one can never really know how a piece of IP is going to work in the environment into which it’s been placed.
No matter that all knowledge in our highly-connected/wired/always-on world is readily available to anybody with access to an Internet Appliance – that wired world is not always forthcoming with simple/available answers about which IP to use, or what part of the stack would be better off with IP rather than something invented in the here-and-now for the this-or-that.
To prove it, try Google-ing DDR memory controller IP and see what you get. Yes, there’s Uniquify. But there’s also Cadence, and Synopsys, and TI, and BarcoSilex, and Eureka Technology, and even Xilinx. Which one of these companies is providing the proper offering for the project at hand?
Well, clearly these sample offerings are not all the same, in design, intent, or use case. Nonetheless, there are offerings out there, among these listed and others, that at least superficially seem to warrant further examination before a final choice is made. So it’s tough. Not all IP is created equal, even if it’s created to be the same.
What I take particular issue with then, in the Spectrum article, is overlooking the importance of the vendor-customer interaction as part of the process of choosing and using IP to create the type of superchips described there.
It’s not just the quality of the IP that’s important in the process; it’s also the ability of the IP vendor to understand the customer’s needs. And then it’s the ability of that vendor to either provide the exactly proper IP for the project, or to provide the exactly proper mix of IP and services, that will fulfill the needs of the customer. The customer who, in turn, hopes to deploy that IP-laden project successfully as it moves down the design chain and on into the ultimate end-product.
People who use IP need not just the products of their IP vendors, they need their knowledge as well.