SANTA CLARA, Calif. An IBM Corp. executive questioned the new 28-nm process rollout from rival Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC), saying TSMC's strategy is risky.
As reported, TSMC is rolling out its 28-nm process and has revealed a major surprise: It has pushed out or delayed its initial high-k/metal-gate offering until 28-nm, putting it behind rivals Chartered, IBM and Samsung. TSMC was originally supposed to offer its high-k/metal-gate process at 32 nm.
At 28 nm, TSMC plans to offer two options for the gate stack: conventional silicon oxynitride (SiON) and a new high-k/metal-gate technology. But at 32 nm, the company will only offer a SiON for the gate stack.
TSMC's foundry rivals are scratching their heads about the announcement. Gary Patton, vice president of IBM's Semiconductor Research and Development Center, raised questions here about TSMC's strategy to extend and scale SiON to 28 nm or even at 32 nm.
"I was confused about [TSMC's] announcement,'' Patton said. Even at 40 nm, silicon dioxide--or its SiON variant--for the logic gate stack ''is running out of gas," Patton told EE Times.
TSMC's strategy to extend SiON is ''a risky proposition," he said. The problem with SiON at 40 nm and below is ''short channel control'' and ''VT variability,'' he said.
In contrast, IBM and its partners plan to offer what they claim is a better gate-stack solution at 32 nm. As previously reported, IBM, Chartered and Samsung plan to offer a high-k and metal-gate solution for 32 nm. The companies in IBM's ''fab club'' will not offer a SiON option at 32 nm and beyond, Patton said.