Samsung (Seoul) did not specify the foundry customer for the product. The company said its eMRAM module can easily be inserted at the back end of its 28FDS process, requiring less dependence on the front-end of the process for easy integration with existing logic technologies including bulk silicon, FinFET and FD-SOI transistor.

Samsung also said it plans to tape out a 1Gb eMRAM test chip later this year.

Embedded MRAM is a promising candidate to replace embedded flash, which is facing insurmountable scaling challenges. MRAM is a non-volatile memory that offers high endurance, fast read/write times and strong retention.

MRAM is also far more power efficient than flash since it runs at lower voltages and does not consume power when in the off mode. MRAM is considered particularly promising for Internet of Things applications.

At the International Electron Device Meeting in December, Samsung described its process for embedding spin-transfer torque (STT)-MRAM with magnetic tunnel junctions (MTJs) in its 28nm FDSOI platform. STT-MRAM is regarded as the best MRAM technology in terms of scalability, shape dependence, and magnetic scalability.

Samsung emram 2 Samsung describes its embedded MRAM process as a plug and play process that requires only three additional masks. (Image: Samsung)

At the International Solid-State Circuits Conference earlier this year, Samsung-rival Intel detailed its 22nm embedded MRAM technology, the first embedded MRAM to utilize a FinFET process. Intel said its 22nm eMRAM is ready for volume production, and some analysts it is already shipping in products Intel has made for foundry customers.

"By integrating eMRAM with existing proven logic technologies, Samsung Foundry continues to expand its eNVM process portfolio to provide distinct competitive advantages and excellent manufacturability to meet customers and market requirements," said Ryan Lee, vice president of foundry marketing at Samsung, in a press statement.