On Cores
Meditations on the semiconductor and IP industries
By Warren Savage, CEO, IPextreme



Thursday Sep. 25, 2008

One on One: Stephen Padnos

Next in our interview series, Stephen Padnos of Atheros offers insights on the IP market.


Steve,

Thanks for meeting with me. I have a few questions for you.


<hr>

First I have a few meta-comments that will influence the answer to all the questions.

Treating IP as one market is difficult. The challenges and opportunities for standards based soft IP, say a USB Controller, are different then for a piece of hard IP like the USB PHY. And both of those markets are very different from the star IP (dominated by CPUs) and the foundation IP (standard cells, memories and IOs).

IP has some roots and is often lumped with the EDA industry. I think that is also one of the problems with IP. The IP business is very different from the EDA business. Trying to force an EDA business model on IP has not been terribly successful.

Many of the companies providing IP do so because of expertise with a particular standard or standards. That is necessary, but it is also necessary to have the tools and methodology for developing the IP. In my opinion there should be more focus on the IP development process. they should be very good at customizing the IP. This applies mostly to hard IP and the ability to rapidly and easily port to new process nodes, foundries or process options. Second, they should emphasize the IP verification, not the creating. It is well documented that verification is often over half of the effort on chip development. The same should be true for IP.


1. What are your biggest challenge as an IP buyer?


As a consumer of IP it is hard to find the IP we need, where we can be confident of the quality, when we need it, at a price that makes sense. Going back to Outsourcing 101 we look for an IP provided when someone else can do it better/cheaper/faster then we can do it internally. We have frequently resorted to developing IP internally that is not core to our business simply because the lead time and cost quoted by providers was 2-3X what we calculated it to be internally.

Another major issue is quality. Quality of IP obviously varies tremendously. And while working with a more established vendor might help, we have had good and bad quality IP from both large and small vendors. I don’t know of any practical way to evaluate the quality.

The problem is most IP has basic functionality. Problems typically arise in corner cases. IP providers often respond to customers by talking about how their IP has been certified by a standards body or they have done a test chip. Unfortunately the goal is a checkmark on a PowerPoint presentation. IP vendors seem more focused on proving their IP can work then proving that it will always work. The goal of verification and test chips is not to show the design can work but to show it can’t be broken.


2. Putting the shoe on the other foot, what do you perceive as your IP provider's # 1 challenge and what should they be doing about it?


It is certainly easy to underestimate the difficulty of developing a piece of IP. This in turn creates unrealistic pricing, schedule and quality expectations from IP providers.


3. The IP business model undergoes periodic attack from the media, do you think the IP business model is broken? Please elaborate.


Yes, I think the model is broken. Companies are struggling to make money in IP. A number of IP providers have recently announced layoffs or exited the business altogether.

IP companies seem to struggle to find ways to provide clear value to their customers. IP has a number of submarkets and each has different challenges.

Hard IP has the problem that customers can be on a huge variety of processes. Although a particular piece of IP, such as a USB PHY may be sold many times, it will likely only be sold in small single digits for a particular process with all the same process options. IP companies have not found a way to quickly and cheaply port IP. The result is that the IP providers often cannot provide proven IP sooner, cheaper and of higher quality than the customer can develop it on their own.

Soft IP that is standards based is easily portable, but is also the easiest for a small outfit to develop and bomb the price. Large IP providers have not clearly established an ability to develop higher quality soft IP.

I’ve heard if floated that for much IP the expectation of the provider is to make money by the 3rd sale. To a large extent this makes IP a services business. I think that is a mistake. For much IP there should be an opportunity to sell many, many more copies. That would also the price to drop, while the IP provider can still make a good return on the investment. That would put the value proposition back in the IP business.

The cost of sales is also currently very high. This partially offsets the value proposition of externally developed IP.

Finally, the business requirements of customers vary. Some are startups with a single product, some are large IDMs. Some customers are working on high cost low volume chips, others are doing high volume low margin designs. It is difficult to satisfy the business requirements of each of these customers with the current business model.


4. The IP market is dominated (revenue-wise) by a few large players, yet there remains hundreds of small outfits. Why is that?


The barrier to entry is often perceived to be small. It is pretty easy for someone to go out and create some IP. It is more difficult to come up with high quality IP, but it is also difficult for an IP consumer to evaluate the quality.

For some IP a couple of people with expertise in a particular area can produce better IP than a larger company without the expertise in that area.

I think this creates an opportunity for an IP publishing model where the larger company has expertise in packaging and distributing IP that was developed by the small shops. The larger company can also help with providing a certain level of quality filtering.


5. Should there be more or less IP companies?


That depends. There need to be fewer talking to end users. It needs to be easier for IP consumers to find IP and judge IP. Having too many vendors makes both of those hard. Web search helps, but is often not sufficient since IP is often under development and hasn’t made it to web pages. It is necessary to know the industry and call around and ask who might have something.


6. What role should EDA companies play in the IP market?


EDA companies offer a good channel. But they also bring an EDA mentality to IP. I think this is one of the problems with the lack of success of IP business models. IP is not EDA.

EDA companies can provide tools to support IP development. One reasonably successful area for this is standard cell characterization tools.


7. What role should Service companies play in the IP market?


Hard IP often looks more like a service then a commodity business since the IP often needs to be ported.

See also the comments above about IP often turning into a services business.


8. To what extent do you think the IP industry has come to grips with its quality issues?


Not terribly well. See earlier answers.


9. Is there a need for greater standardization in IP?


Yes, in several ways. Providing some commonality in the delivery of IP will make it easier for customers to adopt. Providing more standards based IP will


10. Ten years from now, what does the IP market look like?


If I knew the answer, I’d be putting that company together. :)

I don’t think anyone has figured out a good business model yet. Artisan did something innovative with the free libraries and that worked very well for a while, but they failed to evolve the model to take advantage of what they achieved.


Posted by Warren Savage on Thursday Sep. 25, 2008 | Add a Comment




RSS Feed

Add this blog to your RSS newsreader!

Archives

About the Author

Warren Savage, President and CEO of IPextreme, is a well-known and published authority in the field of semiconductor intellectual property. He has a long history of pushing the envelope of design methodology from his work in fault tolerant computing at Tandem Computers in the 1980's and driving reliable design metholologies into commercial practice at Synopsys for its DesignWare IP product in the 1990s. Much of his thinking became embodied in the seminal book on IP reuse, the Reuse Methodology Manual. Warren is taking his vision to the next level with his latest company, IPextreme, which is focused on enabling broad commercialization of IP captive in large semiconductor companies.