A recent, informal survey of several sellers, buyers and users of semiconductor IP cores produced an interesting view of current business conditions and what the future might bring.
Overall, it is reported that a very slow recovery is already underway, although IP providers still complain of excruciating pain extracting money from users.
When IP cores are acquired by SoC developers, they undergo an evaluation phase ahead of any purchase, which largely decouples the function's quality issues from the critical development path.
In most Tier 1, 2 and some Tier 3 accounts, after initial budgetary scouting, an engineering group will exercise IP for typically 2 to 4 weeks by verifying, testing, exploring its behavior and functionality in their systems.
The contractual process will then pick-up again later, just before the main project engages and the core needs to be deployed. Interestingly, most Tier 1 suppliers have extensive internal IP standards (for packaging, delivery, testing, verification etc. etc.).
However, it is almost always the reported case that no attempt is made by such companies to impose such standards upon suppliers, rather the focus is upon "what do you have and what will it take to integrate."
This lack of pressure from users may even be fortuitous, as there is no real uniformity between the internal IP standards held between these companies and although this undoubtedly eases the burden on suppliers, it unfortunately just as certainly affects quality.
Generally, the central R&D groups pushing internal standards in such user-companies seem to be viewed as out-of-touch with the real integration needs and operation of SoC development teams.
Overall, the industry adoption of this off-line evaluation process can (to a degree) "protect" SoC developers, certainly from gross quality issues, but will not always avoid issues related to exhaustive application requirements, nor protect those user-communities unable make any serious level of pro-active evaluation.
What is most alarming are the gross repetition, risk and duplication of effort wasted in this process adaptation, and all for the lack of industry cooperation on shared standards.
Despite common beliefs, the rationalization of royalties versus license fees is not a general impediment in IP sales. Providers either have IP that commands royalties, or they do not, so this conflict should always self-rationalize in the purchasing process.
Also, the evaluation/purchasing process is normally run outside the critical path of SoC development. Despite this, the most serious and common problem reported today in contractual delays is the legal resolution of nuances relating to IP rights, indemnification etc., and this particularly so where Asian and Western buyer-sellers are engaged.
The reported purchasing cycles for technical evaluation phases by buyers of IP run typically 2-6 weeks with the overall purchase commonly completing after 3 months. Purchases in the Military segment typically take 3-6 months from start to finish. A common supplier complaint is that customers will not pay for quality and low-quality (and often lowest-bid) suppliers continue to set the pricing expectations in many sales transactions.
What might change over the next 5-10 years?
It's difficult not to envision a progression of consolidation in the SoC IP supplier base. Increasingly larger IP blocks or IP subsystems will be offered as suppliers strive to add value and expand their opportunities in the video, sound and general multi-media space.
Expect to see large groupings of currently stand-alone IP cores offered with software, drivers and debugging schemes, etc. It is also likely foundries will take on an expanding position and leverage their closeness to silicon hardening with pricing and capacity advantages, much as Synopsys bundles IP products with Design Compiler today.
Quality should continue to improve with vendors either providing improved product quality or fall off supplier lists.
As more users cease to be able to afford the costs of SoCs, growing FPGA usage will most likely play a prominent role in driving standards increasingly heavily into IP core deliverables, since increased ease-of-use will prove essential support for a huge and growing FPGA design community.
Ian Mackintosh is President of OCP-IP.