DAC Special Section

Results of the 1996 USE/DA Survey on EDA Business Practices

The deadline pending--blood, sweat, and gallons of Coca-Cola surged through Steve's veins as he cradled his little girl in one arm while typing the results of this survey with the other.

by Steve E. Schulz

During the 1994 DAC Executive Forum, chief executive officers of the largest EDA supplier companies publicly lamented the existing "purchase/maintenance" pricing model for its inability to sufficiently fund new research and development on 12-percent maintenance contracts. Therefore, it seemed reasonable that both users and vendors might benefit from a revision of current practices.

With this backdrop, this author published an "EDA Platform" column in the November 1994 issue of Integrated System Design to generate industry-wide discussion. The column proposed a number of potential alternatives to the present purchase/maintenance business model and an approach for building "win-win" consensus between users and vendors. The overwhelming reader response to that column, along with Electronic Design Automation Companies' (EDAC's) full support of the process it described for reaching bipartisan agreement, served to mobilize USE/DA's Business Practices working group. The working group developed a user survey--the results of which were presented at DAC '95. A capacity crowd during the USE/DA-sponsored panel session generated much discussion, suggesting that a clarification of several questions would improve confidence in the survey results. In response, USE/DA rewrote the survey to ensure more consistent interpretation by users and, again, secured the support of Integrated System Design to mail out the surveys and collect the response data.

The 1996 survey was mailed to 1,000 of Integrated System Design's readers who met the selection criteria, qualifying them as "EDA tool users." Sales and marketing responses were specifically excluded from the analysis, as were EDA supplier responses. A statistical analysis performed on the 140 responses found extremely good correlation with known profile data for all Integrated System Design readers who use EDA tools. This validates the article's inferences to a much broader user population.

Composite survey results In this section, we'll discuss the survey results in three distinct parts: (1) EDA user profile data, (2) how EDA tools are being used today in industry, and (3) users' specific preferences for business practices in EDA. All results in this section represent an overall composite view of EDA users.

Profile data Most of the users included in this survey analysis told us that they are primarily responsible for engineering (65 percent) or engineering management (23 percent), with relatively few representing non- engineering management (7 percent) or "other" (5 percent). As mentioned previously, sales and marketing positions were excluded from this analysis. The demographic data shows that 66 percent perform design and development engineering functions, 8 percent perform engineering services, and another 13 percent are involved with system or software engineering. Our survey found that nearly three-fourths of these users (74 percent) are on the Internet.

Figure 1. A wide variety of EDA tools are used to support industry design processes. Schematic entry, simulation, and PCB layout are dominant, along with growing use of timing analysis tools.

Since the intent was to address EDA business issues, we asked users for their level of responsibility in making recommendations and negotiating contracts. We found that 22 percent have some direct responsibility for negotiating EDA contracts, and that 60 percent make overall product-level recommendations. Over 80 percent make technical recommendations, and nearly as many select the products which are to be evaluated. This profile is a good fit with the intended target population for this survey.

Electronics companies are a very diverse group. Nearly half (46 percent) of all respondents state that their companies are geographically dispersed and are responsible for a broad range of products. 20 percent represent the IC and semiconductor industry, another 18 percent provide communication systems, and 11 percent produce electronic instrumentation, ATE, and design and test equipment. Approximately 10 percent provide hardware for the personal computer industry, while military equipment represents just 3 percent.

The Unix/PC debate is unlikely to go away any time soon. Most companies use PC platforms as all or part of their EDA environment (83 percent), while Unix is being used by just over half of companies (54 percent). Though one might assume from this that Unix EDA tools are in less demand than PC tools, that conclusion is not necessarily valid. Many large companies use PCs as a supplement to Unix workstations to reduce overall system cost--mostly connected to other Unix tools in a network configuration. Unix systems dominate for almost all compute or memory-intensive applications, such as global IC routing.

EDA tool-usage within companies When we asked respondents to report how they used various EDA tools within their company's design group, some very interesting facts began to emerge (see Figure 1). In spite of the tremendous growth in HDL-based design over recent years, schematic entry is still used in twice as many design groups as HDL entry (79 percent and 39 percent, respectively). Note that the use of logic synthesis, at 42 percent, is very similar to the use of HDL entry, an expected result. Fault simulation's low usage-level (30 percent) may well be an indication of the emphasis towards more structured, built-in design-for-test methodologies. From an analysis of IC versus PCB layout, one can conclude that the sample population represents roughly a 60/40 mix of PCB/system design to IC design, as reflected by the EDA tools being used.

However, just knowing which tools are used gives little insight into their relative priority within the design process. Therefore, users were also asked about the percentage of time spent on various parts of the design process (see Figure 2). The composite data shows that, overall, schematic entry, PCB layout, and logic simulation consume approximately 57 percent of design work, with schematic entry the most time-consuming task (23 percent). By contrast, HDL entry and logic synthesis combined required only 10 percent of the design cycle. This implies that HDL-based design flows are truly a much more efficient means to rapidly generate both RTL- and gate-level design representations.

Business preferences This section of the survey asked users to respond to the following:

* Level of satisfaction with current EDA business practices.
* The importance of various business issues.
* Factors they use to assess overall tool quality.
* Preferences among a set of alternative EDA pricing models.
* Preferences among a set of alternative EDA software distribution schemes.
* Preferences among a set of alternative EDA contract options.
* The single most important change in EDA business practices, and why it is
important to their business.

This survey attempted to minimize the chance of misinterpretation by explicitly defining a set of terms used to describe various alternative pricing models (see Definitions). In addition, the survey emphasized a "revenue neutral" assumption. This means that, for any chosen model, the user would not pay any more or less money for their EDA tools on a year to year basis. The intent behind this assumption was to avoid confusing genuine business preferences with preconceived assumptions of lower tool prices. The revenue-neutral assumption does not preclude savings to both users and suppliers resulting from greater overall efficiency.

Figure 2. Percentage of total design time used by EDA tools for each process. HDL entry, synthesis, and timing analysis combined consume much less design time than schematic entry alone.

We asked users to rate the various business issues in order to understand the relative importance among the issues (see Figure 3). The results indicate that the EDA pricing model ranks as the single most important business issue to users today (chosen by 64 percent of respondents). However, floating license restrictions are nearly as important at 61 percent, and EDA sales practices tied with floating licenses for third place (60 percent). Right-to-use restrictions came in fourth place at 50 percent. Contract issues were clearly fifth in overall importance at 40 percent, which was less than anticipated. It appears that, although contract issues are intertwined with floating licenses and right-to-use restrictions, more emphasis is needed on making improvements to the pricing model, floating licenses, and sales practices.

As shown in Figure 4, 43 percent of survey respondents expressed their dissatisfaction with network licensing restrictions, followed by enhancement support for existing customers (37 percent) and lack of maintenance/support options (33 percent). The purchase/maintenance pricing model only ranked as a moderate priority (30 percent). The current need to negotiate different terms and conditions with different vendors was a slightly lower priority to respondents, at 26 percent--possibly because only 22 percent had stated that this was a part of their responsibility. Surprisingly, product/service quality was a minor area of dissatisfaction. When assessing overall tool quality, after-sale support was an extremely important factor to users (80 percent). The cost of doing business with that vendor also ranked high (77 percent), as did in-house support costs (72 percent).

Next, the survey zoomed in to detailed preferences within three areas: alternative EDA pricing models, contract options, and software distribution schemes. As shown in Figure 5, nearly three-fourths of the users (73 percent) still selected the "shrink wrap" pricing model among their top three alternatives (much as they had done in last year's survey), even though this survey clearly separated the conceptual model from assumptions of "cheaper tools." Of course, actual prices are not determined by models, but rather by R&D costs, support overhead, and total sales volume. The conclusions from this result are that, either users truly want to buy complex EDA software under this model at whatever price, or that users expect less complex EDA software as a result of this model.

Also notable is the strong support for the "mini-bar" pricing model (60 percent). This is most likely a reflection of the need for greater flexibility in the ever-increasing array of tools and need for adjustments to short-term, peak design levels. The "cable TV," "smorgasbord," and "lease" models were roughly tied for third place, while the "mortgage" and "revenue sharing" options were clearly the least popular (38 percent and 19 percent, respectively). An analysis of the features for the top two preferred models shows they serve complementary roles in the marketplace. A shrink wrap model reduces the number of choices and creates a more static resource allocation, while the mini-bar model serves to create new choices with dynamic resource allocation characteristics.

Figure 3. EDA pricing models, floating licenses, and sales practices are the most important business issues to users. EDA contract issues and product packaging options were of lesser importance.

Evaluating the users' top three preferences for EDA contracts is less clear. Having a single industry standard contract, or having no contract at all, were the most popular options (70 percent and 67 percent). Yet, support for a more flexible "modular standard" contract (building block approach), as well as a limited number of (complete) standard contracts, both came in a very close third (64 percent). The sole unsatisfactory option was "contract terms only," which garnered just 33 percent. Many industry leaders believe that agreement on standard contract terminology is needed; however, these results indicate it should be considered merely a starting point for more aggressive long-term goals.

Figure 4. EDA tool users expressed their greatest dissatisfaction with network licensing restrictions. Pricing models rated as a moderate priority, and product service quality did not surface as a key issue with users.

When asked about preferences for software distribution, users told us that they generally prefer direct sales channels with their vendor (92 percent) but also see good value from local technical distributors (87 percent). Catalog sales and Internet-based software distribution are somewhat less popular (71 percent and 62 percent, respectively). This validates the argument that, despite the noted popularity of PC shrink wrap pricing, users still recognize the need for access to specialized application support.

Figure 5. Survey respondents stated their primary preference for the shrink wrap pricing model. There was also strong support for the mini-bar model. Revenue sharing was not popular.

Results segmented by IC/PCB After reviewing the initial survey responses, some data-points appeared out of balance with expected industry responses. For example, how could one explain that HDL entry (VHDL and Verilog) is used by less than half the number of design groups as schematic capture? What happened to the often-celebrated "HDL revolution?"

To gain insight, I asked for a re-cut of the data--based not upon size of company but, instead, based upon whether the design groups performed IC/ASIC design or PCB/system design. Since this could not be extracted directly from the companies' end-products, responses were classified by whether or not they used IC layout and/or logic synthesis. Those users performing IC layout or synthesis were then considered to be designers of ICs or ASICs; those who performed PCB layout were considered to be "system" designers (some overlap exists where IC and PCB layout occur in the same design group). From this segmentation, several clear differences emerged.

Although PCs are a rapidly growing option for many EDA environments, we found that IC designers (as defined above) rely on Unix platforms for EDA--substantially more (65 percent) than do today's system designers (39 percent). This can be explained by noting that VLSI chips often require EDA tools to run for days or weeks on multi-processing servers, requiring several gigabytes of virtual memory. Such tasks are not yet feasible on PCs, and critical EDA tools for IC layout and synthesis are still not available on a PC platform. System designers often rely less on exhaustive simulation and have other means to partition jobs.

When we looked at which EDA tools were used on these mixed platforms, even more distinctions could be made. HDL entry is being used by 57 percent of IC designers, while only 14 percent of system designers reported using it. And, while 87 percent of IC designers use logic simulation, only 48 percent of system designers simulate. Similarly, two-thirds (67 percent) use timing analysis on their ICs, but just over one-third (36 percent) noted using timing tools on their systems.

In terms of business practices, the data showed generally similar views on EDA pricing models, except for minor differences. One such difference is that IC designers appeared somewhat more concerned for product service/quality, while system designers emphasized greater concern for consistent terms and conditions across EDA vendor contracts. Also, IC users seemed more inclined to accept the "smorgasbord" EDA pricing model than systems users (58 percent to 40 percent). For software distribution schemes, systems users favored using local technical distributors (97 percent), while IC users were much more moderate in their support (76 percent). In addition, IC respondents chose Internet distribution more frequently in their top three preferences (71 percent), compared to systems designers (52 percent). Another difference found was with regard to EDA contracts: eliminating them entirely was the favorite choice for systems users (72 percent) and only the third choice for IC users (61 percent).

The data also show differing industry profiles for IC vs. system organizations. We found that IC companies tend to be larger on average, with 55 percent having annual revenues of $50M or better, compared to just 29 percent of systems companies. This is not surprising, as the cost of capital for designing and building ICs (especially wafer fab costs) generally requires greater supporting revenues. Similarly, the data shows that although 50 percent of IC design groups are geographically diverse, only 39 percent of systems groups have multiple geographic sites.

Definitions

Cable TV A monthly lease to a set of tools, applicable to a single vendor. Thus, multiple subscriptions to sets of tools would often be required for full best-of-breed tool flows. By using a short-term lease to a set of vendor tools, package price breaks are likely, even though one of the tools may not necessarily qualify as the best-in-class offering. Separate subscriptions for specialized tools, such as design for test, might make sense to improve flexibility, although to be "revenue neutral" such tools would average out at the same user cost over time. Shrink Wrapped Pre-packaged software sold with user manuals but no free training, hotline support, or upgrades. Subsequent versions are purchased again, and old versions are thrown away. Applying this model under the "revenue neutral" assumption means that EDA software, which has much less sales volume and greater R&D costs than PC-based office automation tools, would average out at the same cost to users over time. Mini-bar This is the pay-per-use model. There is some abuse of this model in the market today. For this discussion, we asked that you assume that a $48k simulator (or synthesizer) would cost $4k a month--as it would be in constant use. On the other hand, a $48k DFT tool would cost $48k a month--as it is only used one month out of the design. Smorgasboard A fixed yearly lease fee for the unlimited use of all tools in the agreement. Again, in today's environment, you would need multiple vendor agreements to get a full set of tools. This would allow you to change tools within a vendor contract and change vendors yearly. Mortgage Purchase a design capability (not necessarily including engineers) that would provide you with the tools and support to accomplish a predetermined design task for a specific length of time. The payment schedule would be spread over the life of the contract. Periodic lease In essence, a rental of tools for a short period of time. Every time you leased, you would get the newest tool available. Also, this would allow the cost to be shared by multiple users. Useful in a low-use environment. Revenue sharing Lower purchase price with some sort of royalty being paid to the EDA vendors based on the end-products' sales.

Conclusions The 1996 USE/DA survey adds further weight behind the argument that EDA customers are dissatisfied with current EDA business practices and want selective change. The EDA pricing model is the most important business issue to users, and floating license restrictions are a close second. EDA contract options ranked less important overall, but clearly contracts remain key in facilitating the changes desired. Users clearly favor the "shrink wrap" and "mini-bar" EDA pricing models and are not at all excited about merely agreeing to standard EDA contract terminology. Most users are satisfied with the current practice of direct sales and local technical distributors.

These results will be forwarded to EDAC for their analysis and consideration, hopefully resulting in a set of improvements to today's practices over the coming 12 to 18 months. USE/DA welcomes your support behind these initiatives, which seek to refine and mature EDA industry practices to the benefit of users and vendors alike.

For more details, contact the USE/DA worldwide user group at www.useda.org/useda.html

Contributing editor Steven E. Schulz, P.E., is a senior member of the technical staff at Texas Instruments (Dallas, TX).

To voice an opinion on this or any Integrated System Design article, please e-mail your message to: michael@asic.com.

integrated system design  June 1996