When reading the Silicon One paper from Magma Design Automation, one is struck by the style of the document. In the EDA industry we are accustomed to technical papers that assume strong technical knowledge of the semiconductors industry. We talk to ourselves expecting that those outside our industry will somehow appreciate both the complex problems and the elegance of the solutions. Magma does no such thing. Its audience, which includes financial and business professionals, is given an accurate, yet understandable description of the requirements of our industry and the possible solutions.
Magma is quite forward in stating that it has chosen to address only some portion of the design flow, and that it welcomes the cooperation of other companies to provide an entire flow of tools from architectural conception to manufacturing. Its expertise resides in supporting the creation of electronic design descriptions that will support the realization and manufacturing of integrated circuits, be they digital, mixed signal, or analog.
The clear message of Silicon One is that Magma’s mission is to enable its customers to develop profitable products, in contrast with most pronouncements that stress performance as the primary objective. It is clear that an underperforming product is not likely to be profitable, so a profitable product will have all of the physical characteristics required including competitive execution speed, minimal power consumption, and acceptable functionality.
I asked Magma for an example of Silicon One in action. We talked about how it applies to system-on-a chip (SoC) design.
Silicon One Solutions
Although by my own count Magma offers 34 products, Silicon One focuses on the five it considers the foundation of its technology: Talus, FineSim, Titan, Tekton, and Excalibur.
The Talus implementation system provides a fully integrated RTL-to-GDSII flow for high-performance, high-complexity, low-power nanometer designs. Talus includes synthesis, optimization, placement, routing, clock skew generation, floorplanning, power planning, incremental RC extraction, and a single incremental timing analysis engine.
FineSim enables the simulation of analog/mixed-signal SoCs with SPICE accuracy and unprecedented performance. According to John Cooley’s DeepChip report, FineSim is winning market share. It allows designers to functionally verify mixed-signal SoC designs seamlessly using a single engine. In addition to enabling designers to work with detailed parasitic information, FineSim allows complete control of accuracy-versus-performance tradeoffs with a few simple control statements.
The Titan Mixed-Signal Design Platform is a comprehensive, state-of-the-art platform specifically tuned to meet the current and future needs of analog/mixed-signal designers. Titan comprises full-custom schematic and layout editors, an analog simulation environment with an option to be integrated with the FineSim simulator, correct-by-design schematic-driven layout, and integration with Magma’s tools for physical verification and digital implementation. By fully embedding Talus for digital design, Titan combines full-custom analog design tools with a digital flow for high-level mixed-signal chip design efficiency.
Tekton has been architected to be the next-generation timing analysis platform, offering the most accurate static timing analysis (STA) engine in the industry with significant runtime advantages. Tekton has been well received by the industry and is also gaining market share. Tekton is a next-generation STA tool. It offers very competitive multi-mode, multi-corner (MMMC) performance on single CPU machines. It provides full support for crosstalk analysis and advanced on-chip variation (AOCV) while offering an embedded SPICE engine when extremely high accuracy is required.
Magma is confident that Tekton provides an order-of-magnitude performance improvement on a single-CPU machine over traditional STA tools. The tool is fully multi-threaded to deliver even faster performance on multi-CPU machines. Using Tekton, design teams can run very large designs with multiple scenarios on a single-CPU machine in minutes. The near-linear performance of timing and extraction engines scaling across multiple CPUs enables Tekton to scale well into the future for designs with hundreds of millions of gates and beyond.
According to Magma, Excalibur is the industry's premier defect and yield management system for semiconductor wafer fabs. Excalibur automatically collects a wide range of data into a single unified database for quick access. Excalibur’s open architecture and powerful extraction and engineering analysis and data correlation tools make it easy to find the sources contributing to yield problems. Automated reports, inline monitoring and alarming are rapidly customizable and can be distributed across the enterprise. Additionally, various automation and defect signature analysis options are available providing for an expanding base of yield and defect analysis solutions.
Magma’s Unified Data Model Still a Key Differentiator
From its inception Magma has always based all of its products on a unified data model architecture. All of Magma’s products use it and this creates a consistent and predictable user environment, as well as fosters tool execution speed and semantics coherence. (A unique name for this key technology would certainly make editors lives like mine easier.) The unified data model allows each of the tools in Magma’s inventory to “mold” the common data for the specific purpose it serves, while keeping a continuous coherence of information that can be then molded again in subsequent algorithms.
To find examples of how the unified data model makes Magma’s tools different than the competition I talked with Behrooz Zahiri, Vice President of Business Development at Magma Design Automation. Here is what he told me:
“There are a lot of EDA tools that interface and exchange data with other EDA tools, that is not the point of the integration. How open any of these tools are can be questioned but there is no lack of standards of how tools can talk to each other. Magma has gone about this differently. Take FineSim and our QuickCap field solver, they are highly accurate, foundry qualified, and built to run extremely fast using multi-CPU computing, but that is not where it ends. Magma has fully integrated FineSim its SiliconSmart characterization solution to provide a seamless, fast and accurate characterization of standard cells, memory and analog IP. Furthermore, the same SPICE engine is integrated in Tekton, along with QuickCap to accurately analyze timing-critical paths during STA, removing pessimism for the best performance. Later, the same STA and extraction engine is integrated into the Talus digital implementation platform to remove the source of timing miscorrelations between implementation and sign-off, reducing the cycle time to design closure. That is not all, the Quartz sign-off quality physical verification engine is integrated in Talus to check and remove DRC violations early in the flow, taking into account timing impacts of DRC fixes before sign-off. Also, SiliconSmart is embedded in the Talus flow to remove unnecessary margins for added performance and predictability. And finally, the entire Talus platform is available inside the Titan mixed-signal implementation environment whereby a full-chip mixed-signal design is easily assembled, verified, and finished without ever leaving the ONE environment. That is the unique value of the Silicon One solution. This unique integration creates increased productivity and profitability for our customers.”
To 28-nm SoCs and Beyond
Magma’s solutions address all areas of SoC design, including yield improvement during manufacturing. Today’s state-of-the-art SoC designs integrate everything – digital (including multiple processing cores and hardware accelerators), analog, mixed-signal and memory – together on a single chip. In fact from an architectural point of view, every SoC can be divided into three parts: the processing cores, the memory, and the Application Specific Logic (ASL) containing both digital and analog circuitry.
High-performance cores are the building blocks of any application or graphic processing chip today. The entire system performance is often limited by how fast these processing engines can execute. Also, the system’s total power consumption depends significantly on how low power these cores can be. Multi-core processors with 2, 4, 8, or 16 cores processors – each with multiple voltage domains – are required to meet the throughput demands of data-intensive advanced applications while still consuming the same power as a single-core design. Designing these low-power multi-core devices is becoming increasingly difficult and implementation mistakes are costly.
Magma’s Talus and Tekton products allow designers to produce optimum digital logic and to optimize the speed of execution through an accurate timing analysis that also indicates ways to minimize power consumption. As computing, information management, and communication become mobile, the ability to minimize power consumption becomes a key element in products serving those markets.
Modern SoC devices need a very large amount of memory to store large amounts of executable code and data. SRAM, DRAM, Flash and image sensors are all memory types that smartphones, cameras, tablets, notebooks and other portable devices need in abundance. This has created one of the most competitive markets in terms of cost and time to market for memory vendors. The key concerns of engineers are high reliability memory chips (every bit of memory must work) at the lowest possible cost. Although one can think of a memory as being a digital device, it is either on or it is off, its cells, or bits if you will, are analog circuits. Magma’s solution to design competitive memories consists of Titan and FineSim that together provide highly accurate but fast simulation combined with an implementation and characterization platform. With this solution engineers are able to design regular structured memory chips in less time. For example, the task of routing top-level memory designs is constrained in terms of shapes, balance loading and uniformity. Titan’s shape-based routing is able to automate these tasks for a significant saving in productivity and cost.
Application Specific Logic
Network processing, data storage chips and consumer multimedia chips such as ones in digital televisions (DTVs) are among the biggest silicon chips in the world. These chips also contain both analog and digital circuitry that must operate as a system, not as a collection of separately designed parts interfaced in a follow-on step. The massive sizes of these devices have made the cost and time associated with developing them nearly impractical. Magma's solution allows project teams to simulate these designs flat with fewer people in less time; that is, there’s no need to break the designs into smaller pieces and run them hierarchically. Designs of this size were impossible to run flat prior to this breakthrough technology. Furthermore, with designs of this size comes a lack of predictability among different designers and different design teams. Front-end designers lack predictability to back-end physical design resulting in numerous time-consuming iterations with suboptimal results. The fact that Titan, Talus, and FineSim all share the same data model is critical to support an integrated design flow. Since Tekton is also engineered to support very large designs and uses the same data model as its front end companions, using Magma’s integrated front-end to back-end solution avoids this problem altogether – what you see (in the front-end) is what you get (in the back-end).
Clearly Magma Design Automation does not aspire to become another Cadence or Synopsys who aim to serve all sectors of the EDA industry. Magma’s expertise rests in transforming logic and circuit designs into efficient and cost-effective IC products. The company stresses the importance of using competitive technological characteristics and manufacturability to generate financial gains for its customers. They are working with leading foundries, such as TSMC and GlobalFoundries, to develop and qualify predictable and efficient design flows for both 28 and 20 nanometer processes.
Their strategy seems to be paying off. The company has just announced its tenth straight quarter of positive cash flow, and all indications point to a significant increase in its market share.