ATI 90nm Graphics MPU Leverages Embedded Deterministic Test

ATI Technologies Inc. reported that it has implemented an embedded deterministic test (EDTTM) tool from Mentor Graphics Corp. in the testing of a new 90nm graphics processor.

The Modular TestKompress tool, a new block-level implementation of Mentor’s EDT methodology aimed at improving

With device complexity growing and process technologies shrinking, achieving cost-effective high test quality has become increasingly difficult, the companies noted.  At the same time, test sizes have grown to keep up with design sizes, and nanometer processes have created new challenges, requiring more tests to detect them.

Mentor says its Modular TestKompress tool implements an embedded compression scheme at the block level that reduces the size of the test sets required for complex designs, allowing comprehensive test coverage without sacrificing test throughput.

As well, Modular TestKompress provides designers with a block-based flow whereby each block, along with associated compression logic, can be implemented and verified independently to simplify the design flow for large block-based designs and reduces the routing area needed to connect the compression hardware.

“ATI is at the forefront of device complexity. Working together, Mentor Graphics adapted the proven TestKompress tool to create a new manufacturing test methodology that solves ATI’s test challenges,” said Robert Hum, VP and general manager of the Design Verification and Test division for Mentor, in a statement.

“Working closely with Mentor Graphics, we were able to use the TestKompress product to thoroughly test our new chip at no additional test cost,” said Michael Do, design for test (DFT) architect at ATI.

The TestKompress tool is meant to fit into most scan-based design flows and provide the same at-speed test capabilities as Mentor Graphics FastScan, an automatic test pattern generation (ATPG) tool. TestKompress contains patented compression capabilities to reduce the amount of test data required for the detection of the speed-related defects that are more prevalent at nanometer process technologies, Mentor concluded.