ChipCenter Questlink
SEARCH CHIPCENTER
Search Type:
Search for:




Knowledge Centers
Product Reviews
Data Sheets
Guides & Experts
News
International
Ask Us
Circuit Cellar Online
App Notes
NetSeminars
Careers
Resources
FAQ
EE Times Network
Electronics Group Sites

  ASIC News

    Editor's Choice

Archives | Feedback

Magma Introduces Blast Fusion APX

The manufacturer says . . . Murray Disman says . . .

Magma Further Accelerates RTL-to-Silicon Flow With New Blast Fusion APX

Provides 2× more capacity, runs twice as fast, and includes advanced power and timing-analysis capabilities for nanometer design.

CUPERTINO, Calif.--Dec. 16, 2002--Magma Design Automation, Inc., a provider of chip design solutions, introduced Blast Fusion APX Advanced Physical Design System for very high-performance, high-complexity nanometer designs. The new product provides a complete netlist-to-GDSII flow, including placement, routing, timing, and extraction, with twice the capacity, half the runtime, and a 50 percent smaller memory footprint than the previous release of Blast Fusion. Blast Fusion APX also includes low-power design and advanced timing-analysis capabilities that allow it to address the requirements of nanometer processes during the implementation flow, reducing time-consuming and error-inducing iterations resulting from nanometer-process effects.

"With IC providers having to cope with smaller budgets, fewer engineers, and shorter design cycles, an integrated-circuit (IC) implementation system's ability to handle larger designs, reduce runtime, and minimize costs is critical to their success today," said Rochelle Perry Drenan, senior product marketing director of Magma Design Automation. "Magma has made significant improvements in these areas, and our customers are leveraging these new capabilities to minimize the design cycle of their complex, high-performance designs.

Proven Higher Capacity and Smaller Memory Footprint

Blast Fusion APX has been benchmarked against existing point-tool flows and version 3.1 of Blast Fusion, and has been tested on customer designs. A recent benchmark of a 1.6-million-gate design shows Blast Fusion APX with a memory footprint of less than half its competition. Throughout the entire physical design flow, peak memory of only 2GB was required, while competitive flows required as much as 5.5GB.

Memory usage is even more efficient for very large designs. For example, when Magma took an existing 2.6-million-gate customer design through Blast Fusion APX, only 2.3GB of peak memory was required. When Magma took a 5.2-million-gate design through Blast Fusion APX, only 3.5GB of peak memory was required. Because it has an extremely small memory footprint, Magma users typically run Blast Fusion on single-CPU, 32-bit Linux platforms with 4GB RAM. For very large designs, existing flows typically use multi-CPU, 64-bit platforms with 10GB or more RAM.

Integrated Architecture and Unified Data Model Key to Improvements and New Nanometer Design Capabilities

Magma's unique integrated architecture and single data model were key to enabling the improvements and to building in support for the new low-power design and advanced timing-analysis capabilities. To enable the capacity, runtime, and memory-footprint improvements, Magma implemented new algorithms and optimized the structure of the data model. Since Blast Fusion APX is a fully integrated netlist-to-GDSII flow, the entire IC design flow can run faster and use less memory.

Magma was also able to extend its signoff-quality timing engine to include on-chip variation (OCV) effects throughout the Blast Fusion APX flow. The effects of pessimism induced by common paths are removed to increase the accuracy of OCV analysis. Unique to Magma, OCV analysis is performed during the IC implementation flow rather than during a post-layout timing-verification phase where timing failures resulting from OCV are detected, and iterations are required to correct it. This approach provides a correct-by-design timing methodology for nanometer technologies.

Magma is the first to deliver a series of new capabilities to support 90 nm technology within a fully unified netlist-to-GDSII environment. For example, multiple-threshold-voltage libraries (multi-Vt) will become mainstream at 90 nm, allowing designs to have reduced leakage current through the selection of high-Vt (low-leakage-power) cells for noncritical paths, and low-Vt (high-leakage-power) cells for the critical paths. During its FixedTiming optimization phase, Blast Fusion APX automatically selects the best cell such that timing is met while power is minimized.

Pricing and Availability

Blast Fusion APX is currently shipping, and has a list price of $600K per year for a 3-year time-based license.

Magma, with its release of Blast Fusion APX Advanced Physical Design System, has substantially improved its netlist-to-GDSII tool by doubling the performance and decreasing the memory requirements by a factor of two. A complete RTL-to-GDSII flow is provided when Blast Fusion APX is used in conjunction with Magma's Blast RTL synthesis tool.

The company says that its users typically run Blast Fusion on single-CPU, 32-bit Linux platforms with 4GB RAM. For example, Magma was able to run a 5.2-million-gate design through Blast Fusion APX and only use 3.5GB of memory.

Magma claims that its FixedTiming methodology and single-data-model architecture are key to reducing the number of iterations required between the logic and physical processes. According to the company, the FixedTiming method sets the delays before physical design, rather than attempting to fix cell sizes first. The final cell sizes will be determined during a sizing-driven placement process, such that the fixed delays can be met based on the real wire load. The result is predictable timing without the iteration loops.

The FixedTiming flow starts with a structural netlist that is read, flattened, and unmapped. Target libraries are then read and timing constraints are set. The design is then re-optimized and mapped onto SuperCells. All delays in the circuit are determined and frozen.

Next, the cells are placed and sized simultaneously to meet the timing based on the actual loads. Buffers and a clock are inserted at this point. Detailed wiring is then performed, while maintaining the delay budgets by tuning the wire widths and spacings.

The price Blast Fusion APX is higher than that for Blast Fusion—$600K per year for a 3-year time-based license vs. $400K for Blast Fusion.

Home    Product of the Week    App Notes    Tech Notes    Newsletters   

Click here to get your listing up.

Copyright © 2003 ChipCenter-QuestLink
About ChipCenter-Questlink  Contact Us  Privacy Statement   Advertising Information  FAQ