The IP paradox: such a low payback for such useful piece of design

[Eric Esteve]

The IP market is a niche market, with revenue of about 1% of the overall semiconductor business, but it is highly strategic for the chip makers. IP helps the chip maker to drastically improve their TTM, when buying a commodity function they can concentrate on the part of the design which differentiate from the competition. When buying an advanced processor core, or a specific controller for the latest Ethernet protocol, to name a few only, they can enter a certain market more rapidly and make a breakthrough. Bringing so many advantages, the IP vendors should be very successful! In fact, that we see is an IP market dominated by either real innovators (ARM, Rambus or Silicon Image), either by (a) large EDA company who has built a strong port-folio by running successive acquisition (USB, PCIe, DDRn, NVM, Memories, MIPI and mixed signal IP). You like it or not, but these are successful and share more than 50% of the market. Then you find a few dozens of companies (maybe less) doing well, but no more, on niche markets. What about the others, several hundred companies? Trying to survive…

What is the best way to buy an IP? When you need for example an interface function like USB or PCIe, and you do not have in house expertise to design it, you have to acquire it. The need for high bandwidth data exchange, and for interconnect your product with other chips or systems, is not a "nice to have" feature. It is a "must have". Then, you have to look for a supplier for this IP function. There are two major issues that you will face: you need a high quality product, fully verified and easy to integrate, and you want to pay it the right price, or the market price.

As you probably do not have a specific knowledge of the IP market, you can go on an IP portal (like "Design&Reuse" or "Chip Estimate"), type the name of the function, and get access to numerous companies claiming that they support the required function. You will get a data sheet, but no idea about the IP quality, the level of Verification, the number of customers already using the IP. What about the vendor commitment to this product line, does is has a solid roadmap or will he give up soon? It is very easy to start selling IP, some design service companies who have developed a function for a single customer decide to market it as a real IP product. Because the entry cost is low, the IP market is very dynamic, there is a lot of new comers, who can end up being a real IP vendor, or stay on the market for a couple of years and finally exit. If you start using an interconnect function, say PCI Express, you will probably use it on several generation of products, so your goal is to select a vendor who can support your product roadmap, stay on the market, and follow the PCIe protocol roadmap, generation after generation. With such requirements, you will discover that the short list reduces considerably…
Our goal at Semi Wiki Design IP is to help you find your way across the IP market, to find the best way!

 

[mngardon]

So it is a tough market for small IP companies. It is especially difficult to deal with when as engineers we all think we have a better mousetrap. So what is needed to break this market open? There are areas of opportunity. Companies have outsourced fabrication, and are now looking at outsourcing IP as well to cut cost, but how and when that will be mainstream is a crapshoot. Additionally, we have seen some large companies actually lose capabilities especially in analog/mixed signal application. As talent comes and goes, managements have not instilled systems to capture and retain that knowledge and expertise with the company no matter where the talent goes. These are opportunities for IP, but an effective conduit is needed between internal need and external solution. Free flowing information sites can greatly help. I look forward to participating.

 

[Pieter]

Dear Eric,

GSA once started a working group and made a tool to measure the quality of IP. It's my understanding that this group wasn't successful and that the tool was ignored by the bigger players in the market.
We welcome all initiatives to support the IP vendors.
IP vendor like us, spend a lot of time and money in proving the IP works. Getting silicon access is for us the first priority.
getting the IP then in products is the next one. So far we've succeeded in this in over 700 IC's in the market. It remains a challenge though.
AN IP vendor like we license our P and with this license comes also our warranty. This should give confidence to the market.

 

[Bill Martin]

industry groups and participation..

Actually the GSA IP group continues to make progress. Warren S is driving it this year. GSA does have various tools that can be used to help with Make vs Buy decisions as well as a Quality tool. Regardless of what group is involved with IP (SPIRT, VSIA, GSA, etc), the progress depends on people contributing their time/energies to various initiatives. Once these are available, then the various eco-system members need to use and critique how to improve. 2007-201? was a very difficult time in high tech world as it was in turmoil. Various layoffs, offshoring, etc caused the staff that remained to pick up jobs that their co-workers were performing. Many companies were pushed to the limits trying to correct revenue falloff vs stable/increasing expenses (which resulted in layoffs). So 'additional' initiatives that were not part of their 'main' job were on the backburner (or even off the stove!). I hope as the industry remodels itself, various initiatives get volunteers that can help accelerate these works. Similar efforts with GSA EDA/Design for 3D are in progress (Herb R) as well as with Text (Steven S). Many other groups are also driving change and standardization on many fronts. If you have the intitiative, I would encourage you to participate.

 

[simguru]

This is one of the reasons I keep campaigning for open-source tools. A common delivery platform and free tools would make life a lot easier for IP vendors.

http://www.semiwiki.com/forum/f2/time-open-source-design-platform-420.html

I have spent over a decade trying to get better support for mixed signal (hard) IP through the Verilog-AMS and SytemVerilog committees, and at the current rate it isn't going to happen anytime soon if the big EDA companies are left in charge.

 

[DaveW]

Fair value in IP?

The issue is about commoditization of design. IP is design content. We can all argue that chips could not be developed without it, but the problem how much of the R&D budget is allocated to IP costs! 1% of chip revenue equates to 4-5% of R&D budget. When that is shared out, not all IP is equal. Some technologies get the lion's share of that (ARM for instance).

There is data elsewhere predicting significant growth in the use of IP. At the same time, growth in IP revenue is predicted as modest. This means further price squeezes.

With complex IP (subsystems?) there is a tendency for integrators to want 'a few tweaks' that essentially means a design service engagement with free IP!

It's a tough business for sure.

 

[Bill Martin]

There are lots of 'ifs' with IP but if you find a quality vendor and continue to reuse and expand the IP purchased from that vendor, IP is the 'one' tool that can have significant impact on development schedules. If you do choose the wrong vendor, you can make you costs and time to market much worse. So there is significant downside on choosing the wrong vendor. But if chosen correctly, there is no tool (implementation or verification) that can have as much impact on productivity as IP will. Choosing the right IP can insure that functionality is correct, that SIG compliance is attained, that engineers are focused on other development/verification other than purchased IP blocks....again, the key is choosing the correct IP vendor regardless of price (well almost...but price should not be the largest factor in purchasing an IP block. If it is, buyer beware.

 

[barun]

I think a buyer will be confident about the performance of an "IP" only when he sees some demonstrable example in the applications where IP is supposed to be used. For example if someone claims their video codec (which can be used in a broadcasting application) can decode "x" number of frames of "y" resolution on "z" platform then it may be wise for IP vendor to have an hardware board with "z" platform where compressed video (generated from a real life source like broadcasted content) of the above specification can be fed and output can be fed into displays like TV in real time. Just storing of compressed video in computer, getting it decompressed in an simulated environment and then display the uncompressed stored video in computer may not make sense.

The vendors need to keep in mind that an IP will not work as stand alone. It will work in an application and it has to communicate with several other sub-systems in that application and hence demonstration of IP in a real time applicaiton is very important. And hence the development of IP not only involves engineering cost but also involves substantial hardware cost to set up the demonstration environment.

Hence it is very important that an IP vendor should develop an IP in collaboration with one of their customers. Stand alone development may not make sense and it may be tough even to first customer to convince the quality of IP

 

[Daniel Payne]

Is there an open forum where users of IP can publicly rate their experience with specific IP?

I'm thinking of something similar to Amazon.com or BestBuy.com where they have forums where people that have bought a product can rate it on a scale of 1 to 5, complete with detailed comments on the Pros and Cons of owning the product.

Would such an open forum be in violation of the IP licensing agreement?

 

[nachumk]

IP designer

As a contractor designing IP for semiconductor companies, I have encountered many old IP blocks, and I have had a chance to design some new ones.

On the FPGA design front, I have seen comparisons between Altera and Xilinx, and almost none of those comparisons talk about the IP. There is a huge difference in quality of IP between these companies.

I have also spent time using Texas Instruments DSPs, and I have been almost consistently impressed by the clarity and the straight-forwardness of the IPs that they have on their C6x class of DSPs.

To the extent possible:
Good IP is straight forward in its design.
Good IP is obvious in its usage.

 

[Daniel Payne]

Nachumk,

So which FPGA company has better IP, Altera or Xilinx?

Just curious.

Thanks.

 

[nachumk]

Daniel,

Xilinx has better IP. Simulations run faster. The modules are better written and easier to use.