Intel Makes 3-D Transistors Reality
A new design promises better performance and lower power consumption for core chips.
Intel on Wednesday announced what it calls 3-D, or tri-gate, transistors. The new transistor design will help Intel improve performance on three fronts, all of which are critical to the continued march of improved performance with lower power consumption.
Devices built on the tri-gate transistors will have lower active state power consumption, lower off-state power consumption, and higher performance. Intel can design devices to maximize any of these three parameters, giving the new process applicability from the fastest server chips to the most power-conscious personal device chips.
More Hardware Insights
- The Critical Importance of High Performance Data Integration for Big Data Analytics
- Simplicity & Usability - New Principles in Workload Automation
- Top 10 Considerations for Getting Started with Virtualization
- Beyond Cost Savings: Four Compelling Reasons to Expand Virtualization of Your IT Environment
Intel is applying this new design to all devices manufactured on its new 22-nm fabrication process. Currently, Intel primarily uses 32- and 45-nm fabrication processes for most chips it produces. Intel competitors are also moving to 22 nm and are talking about their own 3-D transistor designs, but those types of chips are not planned for production until the competitors reach a 14-nm manufacturing process sometime in 2013 or 2014. Intel effectively has a two- to three-year lead with the new design, which will give it a much-needed competitive advantage against British chipmaker ARM, which currently dominates the market for mobile devices. The new design should also provide performance gains over similar chips from rival AMD.
While the details of Wednesday's announcement were closely guarded, the concept 3-D of transistors is over a decade old, originally bearing the name finFET, referring to the fin-like vertical design component of classic field effect transistors. What's important about Intel's announcement is that it has made the manufacturing of these transistors commercially viable, which Intel claims adds just 2% to 3% to the cost of manufacturing 22-nm devices while resulting in performance improvements as great as 37% or active power savings of up to 50% versus the company's current 32-nm devices.
At a press conference in San Francisco, Intel showed devices ranging from laptops to desktops to servers running on chips built using the new process. These chips are the successor to Intel's "Sandy Bridge" chips, which are currently used in state of the art laptops, desktops, and servers.
Dubbed "Ivy Bridge," the new chips will be plug compatible with existing Xeon processors. As production ramps up, end users can expect to see systems based on the chip appearing late in the year or early next year.
Intel's Atom processor will also be built on the new process, but will trail the Xeon chips to the market. Moving Atom to the new process will be critical for Intel if it wants to grab market share from ARM in the mobile processor market. The battle lines are being drawn there as ARM has announced partnerships with IBM to move to its own 14-nm processor, and currently works with IBM for production of 32- and 28-nm chips.
While moving to a 22-nm process will allow ARM to fit more transistors on a chip, the power savings and performance improvements will not be as great as for Intel with its new transistor design. The immediate and likely long-term advantage for Atom may give device manufacturers enough reason to consider either switching to Atom or, more likely, making devices with both chips.
Virtualization can hide performance issues from traditional tools. We provide guidelines to help you understand the hurdles that virtualization puts in front of APM, and how to get over them. Download it now. (Free registration required.)