You haven't been looking if you haven't seen that Linux is a virtual tsunami poised straight overhead, about to crash down onto the beach all around you. Whether you're a designer, an EDA vendor, or an IT guy - if you don't know how to swim, now would be a good time to learn.
And Rob Lucke ('Dr. Linux') might be just the guy to teach you. (Sponsored by Hewlett-Packard, he'll be appearing at DAC 2003 in Anaheim.) Lucke comes to his work honestly, both as an EE and a former Senior Technical Consultant at HP. He specializes in analyzing compute environments for improved systems and applications performance, and implementing heterogeneous architectures for enhanced interoperability. He's Red Hat Certified, author of Designing and Implementing Computer Workgroups, and currently writing a second book on Linux. Who better then, to address issues surrounding Linux, EDA, and IT?
Lucke says it's important to remember that “Linux is a technology and an operating system - it's not a solution in and of itself. You need take the technology and configure it, massage it, and turn it into some kind of solution. To do that of course, first you have to identity a problem and pick the best technology to solve it. Linux is a solution to lots of problems, but not all problems.”
He says that the whole Linux discussion has become confused these days because different constituencies have different needs, and the problems that each one is addressing may or may not be best tackled with Linux: “Engineers, for instance, think Linux is great. For one thing, they can use it on hardware that might otherwise have been outdated - systems sitting around in a closet or in an empty cubicle which can't run Windows XP, for instance. Engineers can put Linux on those systems and do useful work. A cluster of 10 or 15 left-over machines can be used to run an application faster and at very little cost.”
“Also, engineers have inquisitive minds and want to explore the next and best thing - it's part of their makeup. They often have home networks where they're using Linux and becoming comfortable with the technology. Next thing you know, they're bringing Linux into the office and sticking it on the company's network. Engineers are enamored with Linux and want to play with it, control it, and believe they can use it to get a lot of work done.”
“But engineers are often focused on technology for technology's sake and that's causing a growing disconnect between the engineers and their IT support guys. Of course, there's always been a divide between traditional business-oriented issues in IT and the issues surrounding engineering IT - partly arising from the old joke that IT isn't rocket science.”
“Well, as it turns out - in the case of engineering IT, it is rocket science. And it doesn't help, that you often have IT people looking down their noses at the engineering people, while the engineering people don't think they're getting the kind of help they need - which in this case is migrating over to Linux. So, Linux is quickly becoming a flash point between the engineers who want to use it and their IT support people are going crazy trying to keep up.”
“Unfortunately, there's an underlying message here for the IT guys. Whether you like it or not, the engineer is going to bring Linux into the office and it's going to show up. So you might as well get used to looking for ways to use Linux to solve problems. And yes, you'll have to learn something new. But like everybody who works in high-tech, if you ever stop learning in this business, you'll be behind so fast it will make your head spin.”
“Actually, everybody needs to step back and look at the overall picture. The engineer's desktop is actually mutating today; its role is changing. Previously, everybody had to have a supercomputer on their desktop and that system had to be swapped out every 18 months to follow Moore's law. Today the situation is entirely different, however.”
“The brightest customers are now using the desktop as a display device to hook up to racks of powerful machines in the backroom - Solaris, HP-UX, Unix machines, or whatever - while still having enough power on the desktop to run things like Vmware [a virtual machine for MS Windows compatibility], which in turn allows the engineer to get a certain amount of work done locally. Meanwhile, there's load balancing and scheduling software that manages and runs the jobs on the backroom systems - the engineer's desktop being the display device capable of connecting to multiple environments and resources.”
“Don't misunderstand - we're not talking about a thin client here. One of the problems we found with configurations like [Oracle CEO] Larry Ellison's NC [network computer], for instance, was that the graphics performance was nowhere near what you needed to do useful work - even for the 2D performance associated with EDA work. The resources in terms of RAM and local storage were so thin and sparse that you were constantly going back and forth to the servers to get stuff across the network. But the network to most people's desktop has never been sufficient to handle that kind of traffic. Those NC-type devices were only designed for downloading small Java applets and so forth. So, it turned out
that for engineering design, thin clients ended up serving as productivity inhibitors, not as productivity enhancers.”
“Today I prefer to think of the ideal engineering desktop in terms of a “very fat X-terminal,” or better yet, an X-terminal on steroids - one which talks to a Unix system using X-windows as a primary interface, but can also handle VNC, Windows remote desktop, HP's Golden Gate, ThinAnywhere, or other such display technologies. This type of overall system configuration should allow the customer to jump off the treadmill of updating the supercomputer on the engineer's desk every 18 months. Meanwhile, it should allow the engineer, in conjunction with the engineering IT staff, to decide which applications sitting on the engineer's desktop will be productivity tools and which will serve primarily as a connection to resources out on the network.”
“Again, the big picture - there's a load sharing/scheduling facility that can launch and categorize jobs for the backroom. You put most of your compute power on compute racks in the backroom sitting on Linux, Solaris, HP-UX - whatever you choose to be the correct solution for the work you're doing. You collapse the high-speed network to the backroom and run cheap parallel fibers under the floor between the compute resources and storage. Therefore, by using the backroom resources, a virtual computer is created on the engineering workstation. The engineer can open a window on the desktop and the load balancing capabilities will choose which resources to run a job on. To the engineer, it looks as if there's infinite compute power with which to run the job.”
“The engineer is happy - running multiple jobs, all at the same time - and oblivious to what's happening in the backroom. Meanwhile, the IT people can monitor, schedule, track, and buy just enough hardware to handle the load - but not too much. You can get 80% utilization on what you've got using this 3-tier architecture.”
“When I was at HP, one of the functions I performed was environmental assessment for companies that were looking to improve, mutate, or migrate their engineering environments. I've been inside 200 to 300 different engineering environments, all the way from the top EDA and mechanical CAD companies to the largest semiconductor companies in the world. I started seeing the patterns that worked and those that didn't work.”
“If you look at the architecture that I'm describing, it's something that can save money if properly installed. Certainly companies like Infinion, Intel, and Philips Research caught on early and figured out how this architecture could save them money. Today, all of the really sharp companies are doing it, as well.”
“So how does Linux fit into all of this? Certainly there are some tools that run very well on Linux, although we are seeing some limitations right now. The 32-bit Intel chips don't want to go over 2 Gbytes per job. However, if you are able to restrict the job size, it's certainly less expensive from an operating system and hardware point of view to have a rack of Linux systems for certain types of jobs.”
“At this point, engineering IT people might be coordinating a rack of Solaris, a rack of HP, and the job scheduling mechanisms that sorts it all out in order to figure out the best place to use which tool. Of course, there are also file servers and other IT infrastructure matters that need attention. Linux fits well into this infrastructure - for running web servers on a commodity box, for instance. And importantly, Linux is providing functionality that's running on commodity hardware with a 'license-lite' open source operating system - often a distinct advantage over HP-UX, AIX, Solaris, etc.”
“Meanwhile, don't underestimate the power of a Linux compute cluster for software that can use message passing or RPC. Linux clusters are not the same as compute farms, because a single program can run across all CPUs. Imagine an Itanium 2 Linux cluster with 1900 processors simulating 200 atoms in a protein molecule at femtosecond intervals (10 to the minus 15th). That's a reality today. But even on a small scale - compute clusters of 100 systems or less - Linux systems are providing affordable access to some very serious computing. In fact, my next book is going to be about building clustered Linux systems.”
“So, maybe we can reconcile the different points of view regarding Linux by suggesting that engineering folks and IT folks may be at odds today for no real reason at all. They actually share lots of similar values and if they would just step back and ask what they are all trying to do, they would see that there are places where each technology is the appropriate solution for a particular problem - whether the technology is Linux, Unix, Windows, etc.”
“Maybe we should be asking, 'Why can't they all just get along?' Well, in the long run, I'm optimistic that they can.”
You can talk to Rob Lucke and friends in person at DAC. He'll be entertaining questions there and detailing what it takes to get Linux into an engineering environment - be it the desktop or the compute network. Rob says full disclosure mandates he point out that he's not a “real” doctor - Ph.D. or M.D. - nonetheless, he may know how to treat what ails you when it comes to implementing Linux.)
Requesting an update on Linux from EDA
By way of understanding where the EDA world stands currently, a random sampling of EDA vendors were asked to respond to the following statements:
- We have ported none/some/all of our applications over to Linux - OR - We're not there yet, but our future strategy with regards to Linux is
- We are developing our tools/products in one environment and then porting over to Linux - OR - There are timing differences between releases on one operating system and later releases of the same tool/version on Linux.
- Our customers are telling us that they would/would not/must/might like to have Linux-based versions of our tools/applications - The feedback we're getting from our customers vis-a-vis Linux is tremendous/fascinating/annoying/burdensome/they could care less.
- We are having to spend way too much/way too little/just the right amount of our precious resources chasing this Linux thing - Our customers find themselves in the same/different boat.
- True or False? Engineers love Linux - IT managers do not - EDA is ambivalent.
Twenty five companies respond
Applied Wave Research (AWR) - “Our EDA products are running natively on Windows today. We have ported our applications to run as a co-process on a Linux platform. We are presently performing a native Linux port to be completed later this year for those who only want to run Linux apps. We are developing our applications under Windows because of its advantages for software development. We will then port it to Linux, which is expected to be only a minor delay once the process is mature. We are getting requests to have our products running on Linux - it is Unix that people seem to be abandoning. The Linux porting effort seems to be very reasonable for us. What we see is that engineers
just want to run their apps and really don't care about the operating system. It is the IT and CAD groups that are more interested in what OS is running.”
Joe Civello, ADS Platform Product Manager for Agilent EEsof EDA - “Agilent EEsof constantly monitors customer OS requirements. Within the last year or so, we've seen a much more organized approach by our customers to the transition to Linux than in prior years. Based on this recent change and Agilent EEsof's Alliance with Cadence, we plan to have Linux solutions for both the RF Design Environment and the Advanced Design System within the next 12 months. Our approach to PC/Unix/Linux OS support is fairly unique in the EDA industry - we don't port from one OS to another. Our products are architected with a layered application programming interface, which allows us the ability to deliver both at the same time, for each product
release, without porting or duplicating effort. As such, we don't encounter the difficulties associated with porting from say PC to Unix or now Linux. Because of this architecture, the addition of Linux support, to the existing PC and Unix OSs supported today, is straightforward. In fact, internally, our R&D team has been using Linux for product development for several years. The benefits of Linux to the EDA customer is clear, reduced costs and enhanced performance. Agilent EEsof EDA's focus is on providing solutions that lower customers' total cost of ownership and increase productivity. ADS and RFDE support of Linux accomplish both.”
Eric Seabrook, Product Marketing Manager at Aldec, Inc. - “We have elected to port some of our products (based on demand) for use on Linux. Riviera is our cross-platform simulation solution without graphical entry currently available for Windows, Unix, and Linux. We offer graphical design entry for Windows only, but provide simulation on all platforms and also offer the migration of files generated in Windows to be verified on Linux for our customers that are utilizing server farms or remote resources. Our core technology is platform independent, which allows us to concentrate on the product and not the environment. Once completed and tested, it is then compiled to all platforms and available simultaneously to our customers.”
“We have not seen a tremendous request for a Linux version of our tool and get about 25% of the evaluation interest (based on download) for this operating system. We do however see Linux beginning to gain momentum for certain applications and anticipate future growth. We don't expend a great deal of time on Linux-related issues because of our development accomplishments and the ability to produce a Linux version simultaneously with other platforms. This allows us to support the Linux initiative with our tools and create better design methods without limitation. I don't see the Linux argument as a love/hate relationship, rather just another alternative for engineers and managers to use for certain
applications. I also don't see Linux replacing engineering environments, but rather augmenting the development team's available resources. I believe from a cost stand-point that Linux offers great benefit to the IT manager comparable to other alternatives, but tool availability and flow requirements by designer are still not where they need to be.”
Robert DiGrazia, Director of Marketing for Alternative System Concepts, Inc. - “We have ported some of our applications to Linux, and expect to support future applications on Linux. We use multi-platform building mechanisms to cover all platforms. We build and release on all platforms simultaneously. Some of our customers prefer Linux. Most use Solaris or Windows. We spend just the right amount of time on Linux. Linux is largely compatible with other standard variants of Unix. It's true - engineers love Linux and EDA is ambivalent.”
Axis Systems, Inc. - “Axis Systems recently announced its support of Linux, creating the first, complete Linux-based hardware acceleration and emulation systems. All of Axis' simulation, acceleration, and emulation products, based on our ReConfigurable Computing technology and software, have been ported to Linux. The development of Axis' products is nearly independent of the OS, whether it is Linux or Solaris. Both Axis and its customers feel that the time invested in Linux development has been worthwhile. Axis' decision to support a Linux-based verification workflow was based on a strong customer trend to supplement Solaris development with Linux. Initial feedback from customers
indicates that the availability of a Linux-centered design methodology is providing them with cost-effective performance gains. Axis' expansion into Linux fits into its Design Team Emulation solution, which not only provides emulation for verification teams, but also expands the usage of emulation to system, hardware, and software engineering in a development environment - whether Solaris or Linux. EDA stays close to its customers' needs and supports whatever will make them most successful.”
Mitch Weaver, Vice President of the Marketing, Functional Verification group, Cadence Design Systems, Inc. - “Cadence enthusiastically supports Linux. Engineers tell us they like the combination of price/performance and mobility that it provides. We're finding strong demand for both the Cadence Incisive verification platform and the Encounter Digital IC design platform on Linux.”
Jeff Jussel, Vice President of Marketing for Celoxica - “Celoxica's tools for Software-Compiled System Design primarily target field programmable SoC devices and hence run on Windows to better integrate with existing FPGA vendor tools. Some of our customers have expressed interest in Linux versions of our products. The focus of this interest varies across organizations, in some there is a corporate IT move towards Linux, in others the corporate preference is Windows with some engineers using Linux workstations. Overall we have not found that our current Windows-only approach has become a purchase-critical issue - even in 'officially' Unix-only organizations, there seems to be a Windows
PC available. Our current product set is developed and supported on Windows platforms, but we acknowledge in our product strategy that there is a growing interest in Linux. Our product code is readily portable to Linux and we are currently working on a commercial port to Linux, which we expect to have available as an early release version later this calendar year, with full commercial availability during 2004.”
Pete Hardee, Director of Product Marketing for CoWare, Inc. - “As a result of customer demand, all CoWare's tools are ported to Linux, and have been for the at least the last three releases. Our new ConvergenSC product family was our first release completely developed on Linux, then ported to Solaris. Customers investing in new hardware are buying Linux boxes. Customers love running three or four times faster on boxes that cost a fraction of the price of Unix workstations. Two or three years ago, there may have been some doubt. However, customers have been demanding Linux for years and have been getting it from most vendors for some time. If there are any managers who haven't woken
up to Linux, just call them Rip Van Winkle! You can get Linux boxes for hundreds of dollars, running Pentiums at 2.5GHz. Even IT managers have got to love that. One important development overlooked by these questions, however, is the impact of Linux on embedded software. Embedded Linux as a real-time operating system is gaining favor and changing the game in the RTOS market. We'll be booting Linux live at DAC on a SystemC transactional prototype.”
Lauro Rizzatti, Vice President of Marketing at Emulation and Verification Engineering (EVE) - “EVE's software runs under Linux, and we may or may not port it to other platforms. Our software development environment is based on Linux and we will continue to design our future products under Linux. Our customers overwhelmingly like Linux. We have not lost any potential business because of Linux. Customers either use Linux today, or plan to adopt Linux soon. Engineers love Linux. Some IT Managers do, others do not. All EDA vendors either support Linux today or will support Linux in the near future.”
Alec Stanculescu, President and CEO of Fintronic USA, Inc. - “Fintronic was the first company to deliver a commercial Verilog simulator on Linux in 1993. Our development platform has been Linux since 1992. EETimes wrote an article in November 1999 presenting Fintronic as the leading Linux supporter in the EDA industry. Fintronic provides its most advanced features on Linux and on Sun. Our customers love the quality, performance, open architecture, support, and price of the Linux platform. Over the years, Linux has become more and more manageable, reaching a state today where it is as manageable as any Unix environment, but at a much lower cost. The help from ASL Inc.,
which provides our pre-installed Linux workstations, makes it a pleasure to maintain all the Linux machines in working order. The icon-based user interface, provided by Red Hat, makes Linux easier to use than in the past. The high quality, free compilers from GNU together with the increasingly high performance of the Pentium-like platforms make Linux on PCs ideal for both software developers and hardware designers.”
Brett Cline, Vice President of Marketing for Forte Design Systems - “All of the Forte products are supported on the Linux platform and are released on other platforms at the same time. We use Linux throughout the company as both a development and regression testing platform. The use and support of the Linux platform has two major benefits for us: 1) Some of our customers require it, and 2) as a development platform, it provides VASTLY superior performance-per-dollar to proprietary Unix systems. While some people may consider the support of Linux a burden, we consider it a benefit, since we now have the kind of compute farms and servers that we could not afford 5 years ago. We
have heard the same thing from customers, some of whom have significant simulation-farms of Linux machines. The management of the Linux machines is, in our experience, no more difficult than the management of proprietary Unix solutions. Our engineers love them and our IT folks see no discernable difference (except that we now have a lot more of them).”
Jackson Kreiter, CEO at Hier Design Inc. - “We got our funding in a bleak time. It was during 9/11. So we wanted to stretch our budget as far as possible. At some point our little $20K Sun just couldn't keep up. My choice was a) buy another Sun, b) buy a much cheaper Linux box and port our software, or c) do nothing. Well you can't make any progress doing nothing, and we just didn't have money to buy another Sun, so Linux was the choice for us. And, on top of that, we got a 4x run-time improvement.”
David Knol, Engineering Manager for Infrastructure at Hier Design Inc. - “The majority of the Hier Design engineering staff - both R&D and AEs - use Linux as the primary platform for code development, as well as running live customer designs. A PC running Linux offers two huge benefits over a comparable Sun running Solaris - better than 4x the performance for less than 1/4 the cost. It boosts engineering productivity and keeps the CFO happy. Linux is preferred to Windows because it offers the security and stability that only a Unix environment can provide. We aim to release software on all our supported platforms at the same time. Most customers are already running on Linux. Most
who aren't seem to be thinking they'll make the transition soon. The cost/performance factor that benefits us, also benefits our customers. 'Chasing this Linux thing' is an inaccurate description of our situation. It's been a 'no-brainer' for us from the start. Linux won't likely obliterate Windows anytime soon, although the Linux/x86 platform does appear to be a real threat to Sun. Often times, engineers do love Linux. You'll find all kinds of IT managers. But from an EDA standpoint, the cost/productivity advantage of Linux is compelling to any company, especially in these hard economic times.”
Ron Burns, Vice President of Sales and Marketing at InTime Software, Inc. - “We've ported all of our applications to Linux. We deliver simultaneously on Solaris and Linux. Our customers like our tools on Linux - we have had tremendous response regarding our tools on Linux. InTime's combination of RTL timing analysis and the Linux platform has given our customers an advantage in getting their products to market. We are spending the right amount of time on Linux and our customers are also spending the right amount of time. We are believers in Linux and do not think EDA is ambivalent. We really believe in speed, and Linux platforms are delivering speed.”
Magma Design Automation, Inc. - “All Magma products (since we have a single executable) have been ported to Linux. All of our products are 'co-developed' on Linux and released at the same time as the other platforms. Our customers tell us they must have Linux-based versions of our tool and the feedback vis-a-vis Linux is tremendous. We have spent the right amount of resources on Linux. It's true that engineers love Linux, but false that IT managers do not, or that EDA is ambivalent towards Linux.”
Mentor Graphics Corp. - “Mentor Graphics has ported a significant number of products to the Linux operating system and will continue to respond to customers who believe that Linux is an optimal EDA environment choice. However, as with product releases on other platforms, Linux releases may not always coincide with product releases on other platforms. Some Mentor Graphics customers request Linux-based versions of EDA applications and tools because Linux allows a choice of lower-cost and higher-performance hardware. However, they find it confusing to identify software tools and hardware products that match the available Linux distributions. Mentor Graphics is investing in Linux to meet
these customer requirements. EDA has an insatiable appetite for computing performance, capacity, and technology. As Linux emerges in EDA, and other technical computing environments, IT managers must support the disruptive characteristics of an additional operating system.”
Cedric Iwashina, Director of Corporate Marketing at Monterey Design Systems - “We have ported all of our applications over to Linux, and many of our customers insist on Linux. We release our products on our two primary platforms at the same time - Solaris and Linux. Linux is a 'must' - the economic and performance considerations are too compelling to be ignored. EDA has always suffered the burden of having to support multiple hardware/software platforms; it is simply part of the cost of doing business. Our customers understandably seek out the most cost-effective platform solutions, but any savings on hardware are offset by the hidden cost to the customer of supporting application
software on multiple platforms. It's true, engineers love the performance, while IT managers' jobs are complicated by the lower reliability of the Linux platforms and by having to support multiple platforms on heterogeneous networks. EDA, however, is accustomed to supporting multiple platforms.”
Graham Bell, Director of Marketing at Nassda Corp. - “All of our verification and analysis products run on Linux. These are ideal applications for Linux since designers are always looking for faster results and can take advantage of low-cost hardware that runs at clock speeds up to 3 GHz. Linux is not a poor stepchild in our development process. We run regressions every night on our Linux versions. Linux adoption is continuing, and we find that customers are accelerating their interest now that Cadence supports both their analog and mixed-signal design environments on Linux. Resources are not a problem in supporting Linux. One question we do have is, 'How quickly will engineers be
moving to 64-bit Linux, and how popular will the Intel and AMD architectures be?' Yes, engineers love Linux because it feels like Unix, runs 2x-3x faster, and is a viable alternative to Bill Gates' Windows. IT people want to support fewer platforms, not more. EDA is looking for anything that will sell more licenses, and if that means customers' Linux support, then it will be broadly adopted.”
Mark Williams, CEO at Pulsic - “Pulsic has ported all its applications over to the Linux platform. The company develops its products within a single development environment that builds on all platforms simultaneously. So every release of every product is available on all platforms at the same time. Many of our customers have said they would like our products to support Linux, due to the great value-for-money it delivers, in terms of power versus cost. In general, the feedback from our customers indicates that currently the market is impressed with Linux, but we firmly believe (a belief we know is shared with many of our customers) that once a 64-bit PC-based system is available,
the Linux solution will become tremendously compelling. Pulsic has had to spend little time on adding the Linux platform to its platform portfolio. However, our customers have found that the transition to Linux is not quite as straightforward, for various reasons - for example, some other EDA tools do not support Linux, or they work less efficiently under Linux. Another issue - Linux is perceived by some groups as not as stable as the other more established hardware vendors, such as Sun and HP. It's true that engineers love Linux, while IT managers are ambivalent. EDA is seemingly ambivalent, but ultimately companies simply must aim to include Linux within their product strategy, or they will get left behind in the wake of those that already do.”
Rajiv Kumar, COO, Vice President and Co-Founder of Real Intent Corp. - “We support our products on Linux. We release everything at the same time on all Unix platforms - Linux, Solaris and HP-UX. Our Linux solutions are in demand due to the high-performance and low-cost solutions of Linux platforms. We spend just the right amount of time on Linux, since we recognized Linux demand from day one and have proactively built support for it. Engineers love Linux due to high performance at low cost, and EDA has embraced the idea that it needs to do it. But there are some business problems with Linux - lack of support, a very fast release train, lack of guaranteed forward binary compatibility.
These will delay the adoption. The EDA industry and the Linux providers need to address these problems together.”
Bob Dahlberg, Vice President of Development at ReShape, Inc. - “Linux is our preferred platform - everything we do is Linux first. A Sun [platform] is needed in a physical design shop because they support 64-bit. You need 64-bit to support large chips in doing physical design verification like extraction and DRC. We are developing our tools/products in one environment and then porting over to Linux and releasing everything at the same time. We develop in Linux first, and port to Sun second. Meanwhile, we have not talked to a single customer that does not believe that Linux in their shop is inevitable. The price/performance is too compelling - 2x to 3x the speed at one-half to
one-third the cost. Larger corporations have legacy with Suns and HPs to deal with, but even they are making way [for Linux]. As a new company, ReShape can support the new trend in hardware more easily than an established EDA house. Today is not unlike the time when Synopsys/Cadence were getting started in the late 1980's, and when Sun and X Windows were overtaking Apollo and DEC VAX as the preferred EDA standard platform. Synopsys never had to support Apollo/DEC as those standards faded. With apologies to Vince Lombardi, 'Linux is the thing. The only thing.' The price/performance advantage is just too compelling.”
“It's true that engineers love Linux, but for now, IT managers do not. They don't like anything new as a rule, but they're being hammered to reduce costs. EDA companies are not ambivalent. EDA companies in their youth have always loved the fastest box. In its earliest days, Mentor pushed the Apollo DN660. Valid was the first to push the Sun3, which at the time was the first 1MIP workstation. Now that fine tradition lives on with the Linux wave. ReShape loves it. Synopsys' VCS group hopped on to Linux at least two years ago. (VCS R&D knows that speed sells.) Magma is touting it. It's only the out-of-touch EDA companies that support last generation's fastest box, who remain ambivalent about
Mitch Mastellone, CTO for Synchronicity, Inc. - “Synchronicity has two main product lines, the Developer Suite for design collaboration and management and the Publisher Suite for design reuse, IP distribution, and support. We started shipping the Developer Suite on Linux last year and the larger, server-oriented Publisher Suite will be available on Linux before the end of 2003. Once we announce support for a particular OS like Red Hat Linux 7.3, new releases arrive simultaneously with versions for other OSs. We have a mixed OS development environment. Though it's mostly Solaris, we also use HP-UX, AIX, Windows, and Linux machines, so we can build and test every night on all the
platforms we support. We have 120+ customers and they range from Linux activists to being totally indifferent. With its strong potential performance/price ratio on 32-bit machines, many of our customers are adopting Linux and we have responded to this significant customer demand by supporting the common variants of Linux.”
“Since we primarily develop our software on other flavors of Unix, supporting Linux is not difficult, but the number of configurations adds to our quality test burden. The flipside is, the standardization of Unix-related platforms (e.g., a common underlying kernel) should help us reduce that complexity going forward. Both the complications and potential solutions apply to our customers as well. Again, there is a range of love for Linux in the marketplace. Some engineers love it because they have more control, which might makes some IT people nervous. At other companies, the IT department is trying to force Linux roll-outs as a cost-saving measure, but the engineers really love their proven
Solaris or HP-UX. The various support models evolving in the marketplace will also affect Linux adoption.”
Synopsys Inc. - “Synopsys has ported all our products to Linux. We were an early adopter and announced our first major products on Linux in 2000. Our product line is extensive, while developing and porting across environments varies internally. When our products are released on a primary platform such as Linux, all products are released at the same time. We've been selling Linux-based versions of our tools for a couple of years. Our customers are indicating their support for Linux by way of their purchases. We're spending the right amount of time to give our customers Linux-based products. With regards to the true/false questions - Linux has too many aspects to be categorized in
this simple fashion.”
Jeff Garrison, Director of Product Marketing at Synplicity Inc. - “We have ported most of our applications to Linux already, and expect nearly all, if not all, of our applications will eventually be ported to Linux. We are developing our applications on multiple platforms, but mostly on WindowsNT/2000. We port our applications over to Linux, Solaris, and HP-UX - and we generally release all platforms at the same time. We are clearly seeing more customers interested in the Linux platform over time. There is a small segment of users that are passionate about Linux and consider it a must, but for most it is a 'nice to have' capability. Regarding feedback from our customers on Linux
- they are glad we have it, but we don't really see this feedback as being much different than that of other platforms as our applications run well on Linux.”
“We are probably spending about the right amount of resources on Linux. There does seem to be more releases per year of Linux - like Red Hat, for example - which causes additional testing of resources. Also there are many different variations of Linux (Red Hat, SuSe, Debian, Gnome, etc.), so we can't fully test all of them, but that has not been a problem to date. One irregularity with Linux is that, due to its open-source nature, licensing security is weak. While our applications run on Linux, we do require an NT/2000 or other Unix machine in the network to host the license server. Meanwhile, basically it's true - engineers love Linux, IT managers do not, and EDA is ambivalent.”
Paul McLellan, Vice President of Marketing and Business Development at VaST Systems Technology Corp. - “VaST sells tools into three main spaces: SoC architects, chip designers doing co-verification, and embedded software developers. Architects and chip designers run primarily on Unix of one flavor or another, including Linux. Software developers run almost exclusively on Windows. The main challenge to running on Linux is that the GUI for our tools was written using Microsoft Foundation. But Linux users expect an X-windows flavor user-interface, so extensive changes are necessary. However, Linux has another role to play in our space since it is an operating system that can run on
the virtual platform. VaST can boot a simulated microprocessor running Linux in about 10 seconds. Real-time Linux is starting to make inroads as an embedded operating system for cell phones and other systems. In the embedded market, its competition is not PC-windows but rather Symbian and WinCE.”
David Crites, Director of Sales for Zenasis Technologies, Inc. - “We are developing ZenTime on Sun and then porting to Linux. ZenTime on Sun is available now. We have a preliminary port on Linux and will be releasing it to customers next quarter. After our first Linux release, we expect the delay between our Sun and Linux releases to decrease substantially. Many customers are requesting a Linux version, some require it. Since getting our first product out is most important, we have spent just enough time on Linux to understand the challenges and benefits of a Linux release. I'd say engineers seem to love Linux; IT guys seem to love what they know best; and EDA vendors, like us,
love the price/performance potential of Linux, but appreciate the beaten path available on Sun.”
Closing comments on migrating to new technologies
Karen Bartleson, Director of Quality and Interoperability at Synopsys, spent a few minutes providing a large EDA vendor's perspective on migrating to Linux and migrating from 32-bit to 64-bit machines.
“Synopsys introduced its first Linux product three years ago, in 2000. Currently, we support a 32-bit version of Linux. The 32-bit machine is one of our primary platforms and our products are delivered on it. Both HP and Sun are very important mainstream environments for us. However, today we are supporting Red Hat Version 7.2, and we also support 7.3. Meanwhile, we're paying attention to what's coming in Red Hat 8.0 and 8.1. Red Hat doesn't provide compatibility between 7.2 and 8.0, so we want to make sure that the newer versions of Linux are stable and compatible [before fully porting over]. We do not want to disrupt our customers with too many versions of Red Hat.”
“Rather than our asking our customers which operating system they want us to support, it's much more the case that our customers are asking us [for our suggestions]. A design engineer's job is to design a chip. They are not pushing forward new operating systems - they just want to work on the most current ones. We have a very synergistic relationship with our customers here.”
“The 64-bit platform is definitely emerging - today's high-end designs need the capacity. However, the 64-bit [machine] is more about capacity, not performance. And as we have products coming out on 64-bit platforms, we need to be sure not only that the operating system is working, but that the hardware is working as well. Meanwhile, Red Hat is coming out with its new AS (Advanced Server) product, Version 2.1. We're using that right now for our 64-bit products.”
“For a company like Synopsys, we have had some special 64-bit arrangements with specific customers, a situation that we approach on a customer-by-customer basis. For any EDA company these issues are complex - which operating system, which underlying hardware, how fast to migrate to newer technologies, how fast to ask the customer to migrate. We believe that it's easier for bigger companies to port and test on newer platforms, and believe that's something that can be very challenging for smaller companies. Start-ups are looking to the leaders in EDA to set the tone, and then follow that initiative.”
“There's definitely a real balancing act here. New platforms are the key to new technology, but are also disruptive at the same time. Anytime a customer has to move to a whole new set of computers, they have to perform a host of quality tests, etc. Venders never take that kind of migration lightly - and neither do the customers.”
Industry News - Tools & IP
Anasift Technology Inc. announced the company is entering the high performance op amp optimization and synthesis market. The company says the products will be based on Symbolic Based Optimization, and that existing optimization approaches such as simplifying and linearizing the transistor models or using parameterized compilers don't result in high performance analog. The company's initial products are aimed at optimization and synthesis of high performance op amps, the basic building block for analog design and the device that typically gates the performance of the overall analog design. Anasift's products are designed to fit into existing Cadence, Mentor Graphics, and Synopsys analog design flows.
Applied Wave Research Inc. (AWR) and OEA International, Inc. announced a technology licensing agreement that enables AWR to integrate into its high-frequency EDA design environment core interconnect-extraction technology into OEA's NET-AN 3D multi-net field extractor. The companies also announced a marketing agreement enabling AWR to market a number of OEA products through its worldwide distribution channels.
Cadence Design Systems, Inc. announced that Teradiant Networks Inc. used the Cadence RTL Compiler to accelerate development of the TeraPacket chipset, which includes a network processing engine and a traffic manager. The companies say the 200+ million transistor chipset is among the densest semiconductors designed to date. It took Teradiant eight months to design TeraPacket using Cadence RTL Compiler, which the company calls a high-speed, high-capacity tool for RTL synthesis targeting advanced foundry process technology.
InnoLogic Systems, Inc. announced that Centaur Technology, Inc. and NVIDIA used InnoLogic's ESP-CV functional verification tool for development of the VIA C3 processor cores and GeForce FX cinematic graphics processor units respectively. Additionally, the company says that it has “achieved a significant milestone in the number of successful tape-outs accomplished by customers using ESP-CV
over one hundred successful silicon tape-outs from industry-leading electronics companies including MIPS Technologies, National Semiconductor, and Sun Microsystems.”
Magma Design Automation Inc. announced the release of Blast Rail, described by the company as a “correct-by-construction” rail design product to ensure power integrity for nanometer designs. Blast Rail is integrated within Magma's RTL-to-GDSII implementation flow. Blast Rail combines power planning, power analysis, voltage-drop analysis, voltage-drop-induced delay analysis, and analysis of electromigration on rail wires and vias. The tool also provides capabilities for static analysis as well as dynamic analysis, and provides the ability to model network resistances, capacitances, and inductances - and uses these network parasitics to perform transient analysis.
Also from Magma Design - The company announced that WIS Technologies Inc. used Blast Fusion and Blast Rail to complete what the company calls single-pass design closure on a 5-million-gate, 100 MHz, 0.18-micron designs. Additionally, Magma said that WIS was able to tapeout the design in 2 weeks without iterations.
Nassda Corp. announced that Aeroflex Microelectronic Solutions has adopted Nassda's HSIM full-chip simulator and analysis tool for verification of memories and complex mixed-signal designs in Aeroflex's high-reliability IC families for the aerospace and defense markets. Aeroflex engineers have applied HSIM in pre-layout and post-layout verification of a 0.18-micron CMOS memory design, a 0.35-micron memory, and two 0.25-micron CMOS mixed-signal designs. The memory designs include a 128Kx32 rad-hard SRAM memory with 60 million transistors and a rad-hard 32Kx8 PROM with 600K transistors. The mixed-signal designs included a LVDS serializer and an LVDS deserializer, each containing a PLL.
Palmchip Corp. announced the availability of what the company describes as “the industry's first complete serial ATA IP solution.” The product includes the serial ATA target core, the host core, and the analog PHY developed by Astro Semiconductor (see Newsmakers). Palmchip says it announced its serial ATA host controller core and serial ATA target core earlier this year. Through a recent agreement with Astro Semiconductor, Palmchip will now provide the analog PHY IP.
Pulsic Ltd. announced the new release of its Lyric Physical Design Framework, which the company describes as a “next generation physical design tool suite that provides a flexible, high performance auto/interactive routing and ECO placement solution for all IC design types, including complex analog, custom, mixed-signal, and SoC designs.” the Lyric IC Router module has been enhanced to include improved features and performance in terms of speed and completion, and tools for routing DRAM and SRAM memory designs. New physical placement capabilities provide support for ECO, deep-submicron process rules, and support for 64-bit computer architectures.
Sagantec announced Anaconda, a schematic-driven, constraint-based compaction tool intended to accelerate analog physical design by automating repetitive manual layout tasks and allowing analog design reuse. Anaconda targets the manual effort involved in analog reuse today, correctly placing and sizing the layout details of the derivatives and variations of a given physical topology and the migration of circuits to new processes. Anaconda reads sizes and constraints from a schematic and then refines a given topology to automatically implement the specifications to produce a complete layout.
Sequence announced that NVIDIA Corp. used Sequence's PhysicalStudio in six successful tapeouts. The companies says that one of the tapeouts was the GeForce FX 5800 Ultra GPU, a chip which incorporates numerous “out-of-this-world features” such as real-time rendering, and that the project posed complexities and design challenges NVIDIA addressed with the Sequence tool. Additionally, Sequence announced that it is releasing the next-generation of PhysicalStudio, with new signal-integrity, power-optimization, and voltage-drop analysis features that the company says are tuned for 90-nanometer design. (More discussion to follow next week.)
Silicon Canvas, Inc. announced Laker 3.1, which the company describes as “the first connectivity-driven layout tool for controllable automated full custom design.” The company has incorporated its Controllable Automated Technology (CAT) and connectivity-driven layout methodology into the existing rule-driven Laker product. Laker 3.1 is intended to allow layout designers to reach “handcrafted” layout quality from the netlist or schematics in much less time than previously required, without undergoing the difficult process of manual full-custom layout. Additionally, the tool's ECO capability, stick diagram, cell template, and rule-driven technology permit different layout topologies to be
explored and manually modified. Designers can maintain control over the quality of the layout, while the tool automatically performs much of the formerly manual process. The company says that measured productivity gains conservatively range from 2x to 6x on chips designs that vary from analog/mixed-signal ICs to SoCs.
Silicon Metrics announced that it has joined Synopsys' Milkyway Access Program (MAP-inSM). The company says it has tightened the integration of its SiliconSmart characterization products with the Synopsys Galaxy Design Platform for timing, power and signal integrity design closure, and this will result in better-performing nanometer design flows when used with tools in the Galaxy Design Platform.
Synopsys, Inc. announced a partnership with Airbus that has developed a new tool, Saber RT, for the simulation of systems in the Airbus A380 - described as the world's largest passenger airplane. The companies say the new tool allows for interactive simulation of many systems including hydraulic components, and that Airbus has used the tool to save prototype costs and speed time-to-production. The companies also say that traditionally, multiple hardware prototypes costing up to $10 million are needed to simulate the thousands of functions in large systems - everything from an overhead light to the landing gear. Synopsys says that Saber RT performs “interactive” simulation, which
behaves identically to an actual hardware component and eliminates the need to build multiple components, while functioning as a bridge from simulation to production.
Also from Synopsys - The company announced that ATI Research, a subsidiary of ATI Technologies, Inc., has adopted Synopsys DFT Compiler SoCBIST to implement the design-for-test architecture for its upcoming visual processor. SoCBIST is an extension to DFT Compiler. By using SoCBIST, ATI says it is able to improve test quality and reduce test cost for ATI's visual processing unit with 200+ million transistors of digital logic. The companies says that a design of this size and complexity requires extremely high stuck-at fault coverage, and thorough testing for delay-related defects, the most common defect type in 0.13-micron process geometries and below.
Finally from Synopsys -The company announced Magellan, which a new hybrid formal verification product. The company says that Magellan combines advanced formal engines with the “strengths of the built-in VCS simulation engine to help engineers uncover bugs that may be buried thousands of cycles deep in the design.” The tool's hybrid architecture provides deterministic results that the company says are free of false-negative errors. Magellan supports Verilog and VHDL designs and is architected to work with the emerging SystemVerilog standard. It is part of the design-for-verification (DFV) strategy being pursued by Synopsys. The press release was accompanied by a testimonial from NVIDIA
Corp. who said they used Magellan for developing NVIDIA graphics processing units.
Tanner EDA introduced the latest release of its layout and verification software, L-Edit Pro version 10. The software runs on Windows and is used for analog and mixed-signal IC, MEMS, and integrated optical device designs. New features include hierarchical DRC and the ability to utilize hierarchical information to eliminate redundant checking of duplicated geometry. The hierarchical error browser allows users to view and correct errors at the level where they occur in the hierarchy. L-Edit Pro 10 also provides all-angle DRC for MEMS and optical designs, and supports all-angle polygons and wires, circles, arcs, pie wedges and tori curve editing, layer generation and Boolean/grow
operations, and improved dimension display and multi-language menu options (English, Japanese, Russian and German).
ViASIC Inc. announced a new one-mask modular array architecture and new physical design software for modular arrays, as well as a design kit that the company says includes everything a design team needs to implement a TSMC one-mask design, including the software, a technology license, and the master tile. ViaMask is a one-mask architecture for 0.18- and 0.13-micron foundry processes. ViaPath is a physical design tool for placement, optimizing, and routing a one-mask design, intended to reduce the number of custom masks required to configure logic from up to 28 masks to just one. The company says the new products will allow modular array technologies such as those from LSI
Logic and NEC to reduce customized masks per design by 4x or more and will address the gap between FPGAs and the increasing cost and complexity of standard-cell ASICs.
Zenasis Technologies, Inc. announced ZenTime, the first product to use the company's hybrid optimization technology. The new tool is intended to help ASIC and SoC design teams reach target performance by injecting large timing gains into their designs in “just the right places.” ZenTime can generate timing gains using a hybrid of transistor, logic and physical optimizations that are 2x to 4x larger than conventional timing closure tools. The company says ZenTime can gain over 50 MHz in additional performance in 130-nanometer processes, The tool operates simultaneously at the transistor, gate, and physical levels, and the hybrid optimization combines “the benefits of custom-cell
crafting, physical optimization, and placement accurate timing.” The company says that by optimizing at the transistor-level, ZenTime finds ways to improve timing using fewer or smaller transistors and fewer inter-cell wires, and therefore improves timing without introducing power, area, or signal-integrity penalties. The company adds that no change is required to existing synthesis or physical design tools when ZenTime is added to a flow.
Coming soon to a theater near you
Sonics & Flextronics Course - It may not be too late to sign up for this free seminar on Wednesday, May 21st (in Mountain View, CA) or on Thursday, May 22nd (in Westford, MA). The two companies are making a joint effort to teach you how to accelerate your product's time to market by discussing topics from concept through to volume manufacturing. (
SystemC Course - Forte Design Systems is offering a free on-line SystemC training course intended for designers who are investigating language alternatives for high-level design. Designers who complete the course will understand the fundamentals of SystemC and considerations affecting its choice as a language for high-level design. The comprehensive course can normally be completed in 4-6 hours and will cover methodology, language usage, as well as comparison of SystemC and C++. Content for the course was created in partnership between Forte and Willamette HDL. (
Astro Semiconductor is a new company being formally launched by Anjan (AJ) Sen (formerly an executive with eSilicon Corp.), and is an analog and mixed-signal IP company founded in January 2002. With the launch, Astro is announcing a series of physical layer (PHY) IP for the storage and communications markets, and analog IP for the consumer market. The company says the Astro technology combines a design approach known as Adaptive-Dynamic Biasing and a methodology that allows efficient porting of analog IP to any foundry that has a standard CMOS process. The company's “siliconized” analog IP includes the first storage product, a 1.5Gb/s Serial ATA PHY compliant with Gen 1 of the Serial
ATA 1.0a specification. Astro also announced “a SPI-4 Phase 2 PHY and a HyperTransport PHY as part of its communication line of products and added a high-performance amplifier platform and PCI-X 1.0 I/O to its consumer product line.”
Sen said, “I formed Astro Semiconductor to deal with the challenges of designing and integrating robust, low-power, high-performance analog functionality into large, deep-submicron ICs. Bleeding-edge analog functionality demonstrated in a lab is a big step away from the end goal of delivering robust analog IP integrated into a large DSM ASIC that has a high yield and works to specification. Our analog products are proven to exceed performance expectations in real-world applications - even under the toughest noise conditions encountered in low-cost, mass-produced systems.”
Beach Solutions announced that the company has appointed Arthur Cook as CFO. Arthur Cook is a Fellow of the Institute of Chartered Accountants and has 25+ years' financial and general management experience in industries including transportation, trading, steel, and IT. He will report directly to Beach Solutions CEO Terry McCloskey.
Lanner Group has appointed Scott Dixon Smith as President of its U.S. operations. The company says Dixon Smith will also help lead an “aggressive expansion” into the rapidly growing business process management (BPM) market. Previously, Dixon Smith was one of the founders of Holosofx Inc., which was acquired in September 2002 by IBM. In 1989 through 1991, Dixon Smith served on the Malcolm Baldridge National Quality Award Committee and is a certified Malcolm Baldridge National Quality Award Examiner. He has a BA from Westminster College, an MS from the University of Alabama, and an MBA from Harvard.
Monterey Design Systems announced that Alan Feinberg has been named Vice President of North America sales, and will be responsible for all Monterey sales operations in the U.S. and Canada. Feinberg has 18+ years' sales and marketing experience in EDA. He was one of the earliest employees at Synopsys and opened the company's first field sales and became director of North American sales operations. He also held executive positions at ACEO Technology and Meta-Software prior to their respective acquisitions by Avanti Corp. Prior to joining Monterey, Feinberg was Founder and President of TEKSTART, a Silicon Valley based consulting firm. He has a BSEE and an MBA. Feinberg periodically
writes articles for the San Jose Mercury News as a member of the publications Board of Contributors.
ReShape, Inc. announced the appointment of Joe Mastroianni as Vice President of Engineering. Previously, Mastroianni was vice president of product development at Adaptive Silicon. Prior to ASI, he was with Cadence Design Systems for 10 years, most recently as vice president of design methodology engineering. Mastroianni also spent two years at Intel and seven years as a member of the technical staff at the RCA Microelectronics Center. Mastroianni has a BSEE and an MSEE from Rutgers University. He has published several science fiction short stories and his first novel will be published in the spring of 2005. A partner in Magee Scientific, he was formerly with an expeditionary team
funded by the National Science Foundation to research the use of alternative energy on Antarctica. While on Antarctica, he designed, constructed, and deployed an 802.11x network for harsh climates. David Gregory, President and CEO of ReShape, said, “We searched the world from top to bottom to find someone of Joe's caliber and we found him in Antarctica.”
In the category of ...
Issues at DAC - Revitalizing Silicon Valley and EDA
No matter that DAC will be in Anaheim, you're going to hear conversation in Southern California about what's going on in Northern California. To many observers, Silicon Valley is the epicenter of the EDA industry, but the bloom's off the rose, so to speak, in Silicon Valley these days. Jacques Benkoski, President and CEO at Monterey Design, expresses some strong opinions here on what needs to be done if EDA and Silicon Valley are going to rise and shine once again.
“We're seeing a switch, at this point, to a different growth pattern in the semiconductor industry. Carly Fiorina (CEO at Hewlett-Packard) was saying last week that although we're accustomed to have electronics revenue growth at 5x GNP [gross national product], going forward we're going to have to settle for 2x GNP.”
“Think about it. At 5x GNP growth, we could rush forward, concerned only about saving time and producing the latest, greatest product. There was no concern about the cost [of product development] and there was a ton of investment at the beginning of the product cycle.”
“But in a 2x GNP economy, the cost of development quickly becomes relatively more important than the time to market. In fact, you'll see today that most products are reaching their downward price slope even before they reach their [full market] volume. By the time anything's adopted or selling, the margins are not there - or the product's not being bought at all.”
“Look at 802.11. It will probably have as much impact on lifestyle as any technology over time, but is anybody making money out of it? No.”
“[Clearly] the semiconductor industry has to start looking at other industries now for our growth models - mature industries that grow at a more moderate pace. In the automobile industry for instance, they spend a lot more time planning and test-marketing a product before making the decision to manufacture it - and the overall product cycle is a lot slower. More importantly, car manufacturers, try to amortize R&D costs over multiple products.”
“Which is exactly comparable to IP reuse in the semiconductor industry and something we need to pay attention to. EDA has a key role to play in the IP integration mechanism and we can look at the car industry for guidance here as well. In fact, [similar to the way things are done in the automobile industry], we're seeing some of our customers today planning whole families of products before even launching the first one, based on IP reuse.”
“The ability to model products is even more important when it's so expensive to design and manufacture the products. FPGAs fit into the business picture here because contraction means the [overall] production volume is down. System manufacturers can use an FPGA as a step in the design process today, and as an important step in the early manufacturing phase of a product as well - [especially] when you don't want to commit to volume production until you know there's a market for the product. We're seeing systems guys creating boards with FPGAs and willing to go out into the market and lose money, [with the express purpose of] seeing if the product warrants the cost of the ASIC-based version.
But the only way to make money in volume is to use silicon integration, however, and our job is to enable that effectively.”
“Meanwhile, we're seeing a whole new way of doing business here in Silicon Valley. First of all, there are ten times as many meetings before a decision is made, and it's no longer easy to justify expending so much energy with so little return - which causes a trickle-down effect and more contraction in the industry.”
“But I'm optimistic that we'll return to full employment in Silicon Valley soon, and with roughly the same mix of employee types that we've seen in the past. In this market, only the best employees survive - if you want to hire good people, they're still not easy to find. Meanwhile, housing prices here continue to be almost at dot-com-era levels.”
“As we all push forward the benefits of globalization - opening economies and increasing the level of education - global markets are beginning to compete with us on talent. I expect, going forward, certain things will be done better in China or in other emerging markets. But we're not going to be weakened here in Silicon Valley just because some of the infrastructure has moved to India or China.”
“In fact, only the large players can afford to invest in India and China these days - the barriers of entry have become too high - so we've recently set up a subsidiary in Armenia. Countries like Armenia, and other former members of the Soviet block, have engineers who are well-educated and very skilled at writing computer code. They have the theoretical knowledge to contribute to our R&D, and there are excellent universities there as well.”
“Silicon Valley continues to be the center of the [technical] world and we'll bounce back. We have a unique blend here of everything one needs to be successful, a confluence of many things - technical capability, available capital, marketing savvy, and the requisite infrastructure, including legal counsel, accounting expertise, and an on-demand IT network.”
“I don't think history will repeat itself. I don't think there will be another boom cycle in Silicon Valley like we saw in the 1990's - a time when we certainly saw excesses. But we'll definitely see another upturn here. It may have some of the same characteristics of the last upturn, but the fact that we're more concerned today about cost rather than time to market - that will be with us for a long time.”
Issues at DAC - Structured ASIC versus Platform
Max Lloyd is CEO of ViASIC Inc., a young EDA company out of Research Triangle Park, NC, emphasizing physical design technology. Bill Cox is CTO for the company.
In a recent phone call, Lloyd and Cox sorted out definitions and explained the technology behind the term “structured ASIC.” It's probably good to master these definitions now before DAC. After all, LSI Logic is also talking about “structured ASICs” in their RapidChip product, so here's an opportunity to sort out what the thing really is.
Max Lloyd said, “The term, 'structured ASIC,' seems to be gaining the most acceptance. We switched over to using the term because every other phrase ended up getting someone upset. For instance, we were using the term, 'modular array,' and 'ASIC array' but that caused problems with the association with gate arrays. Now we've switched over and are using 'structured ASIC.' The term doesn't actually say a whole lot - it just refers to an ASIC with pre-defined layers and user-configurable layers, which means that you can go through and pre-manufacture certain layers of the chip.”
“Our ViASIC product is intended for the structured ASIC market, where there is only one user-configurable layer. The lower layers are fixed and the upper layers are fixed, so you only have one layer that needs a custom reticle. For instance, perhaps you have segmented wires on levels 4, 5, and 6. All of these wires have potential electrical paths down to layers 1,2, and 3. By having or not having certain vias between layers 3 and 4, the one-mask via layer, you can create your custom electronics. If you put a via there, you complete the electrical path. If the via's not there, it's an open circuit. The concept is like that of an FPGA, but you're using vias instead of a large programmable
electrical switch. Using a structured ASIC is a good way to implement a platform.”
Which begs the question - when is it a platform and when is it a structured ASIC?
Bill Cox said, “You can implement a platform in a structured ASIC. For instance, in LSI's RapidChip, each base is a platform. In fact, you might call what our customers make, a 'platform.' Typically they will have a high-value piece of silicon that they want to incorporate into several products. They'll customize each product with our technology. We call those types of customers, 'flexible standard product' guys. They might have a new networking widget, for example, and they're doing custom silicon at 0.18 or 0.13 microns. Now they want to sell it with 6 different applications. We'll do the logic that they need to differentiate those 6 different applications. Typically, we'll be used for just
a small piece of their die, where the rest of the die might be some difficult SoC type of thing. We allow customers to develop a product family of SoCs.”
Lloyd said, “Actually, platforms are defined differently by everybody. A 'platform' means that you're re-using a piece of the silicon - the part that's common across different product families. For instance, a flexible standard product is one of our concepts of what a platform could look like. If you wanted to make a basic platform for different customers - if you were selling chips to both Sony and Philips, for instance - you could customize the platform using our technology for each customer. Platforms are a great way to introduce new technology, but without the expense.”
Lloyd said, “Basically, you have four methods for producing products on silicon.”
“One is to trade initial engineering costs with high per-part prices - FPGAs “
“One is to use old-fashioned gate arrays - but with the number of metal layers common today, about half the masks still need to be customized.”
“One is to use a standard-cell ASIC at 130 nanometers and below - but that's killing people these days with the mask costs.”
“One is to use a structured ASIC - which is a new methodology. Big players are coming out with structured ASICs, but none of these companies can implement the products with fewer than 4 masks. We're offering tools to do it in one mask.”
Cox added, “Meanwhile, for structured ASICs, the only software available has been at the fab or at the manufacturing location. For example, one vendor has to have specialists who run the tool flow for a particular chip. The customers can't touch the tool. The vendor plans to fix that, but meanwhile we've allowed the physical implementation of structured ASICs back to the designer, the one who best understands how the chip should function.”
“Typically today, in a standard cell ASIC flow, the guys who know a particular market - say cell phones or consumer electronics - will throw the specifications over the wall and let the ASIC vendors design the chip. Most products at Circuit City are done with standard-cell ASICs, at costs of 50 cents per part. But it takes $750,000 to get the mask off the line.”
“By using structured ASICs, the parts costs may be up to 75 cents per package, but you have reduced mask costs - $40,000 to $50,000 - and improved performance. By using structured ASIC, we'll have a whole new market.”
The hard working folks in PR
The following was brazenly plagiarized from a Mother's Day greeting making the rounds in e-mail last week:
WANTED: People to work in PR
JOB DESCRIPTION: Team players needed for challenging work in an often chaotic environment. Candidates must possess excellent communication and organizational skills and be willing to work variable hours, including evenings, weekends, and frequent 24-hour shifts on call. Some travel required, including trips to conferences and trade shows as needed.
RESPONSIBILITIES: Must be impeccably polite to everyone at all times including clients, editors, and investors. Must be willing to bite tongue repeatedly. Must be willing to be hated, at least temporarily - especially when gently reprimanding a client for speaking out of school. Must possess the physical stamina of a pack mule and be able to go from zero to 60 mph in three seconds flat. Must be willing to face stimulating technical challenges, such as arranging conference calls, zipping/delivering PowerPoint presentations, telling time across multiple time zones, or understanding what in heck is being discussed around the table between technologists. Must screen phone calls, maintain calendars,
and coordinate production of multiple projects simultaneously. Must have ability to plan and organize social gatherings for clients of all ages and mental outlooks. Must assume final, complete accountability for the quality of the end product. Must be willing to be indispensable one minute, unnecessary the next. Your job is to remain steadfast, without complaining. Your compensation will be directly proportional to the state of the economy. Sometimes there will be no compensation at all. Must always hope for the best, but be prepared for the worst. Must appear to love work at all times.