HardwareIBM's 2nm Breakthrough: Implications for Chip and Server Makers

IBM’s 2nm Breakthrough: Implications for Chip and Server Makers

A picture of a semiconductor technician holding a wafer of the newest innovation in transistors, IBM's 2nm.
A 2 nm wafer fabricated at IBM Research’s Albany facility. The wafer contains hundreds of individual chips. Courtesy of IBM

IBM last month unveiled its proof of concept for the world’s first 2-nanometer transistor technology. Initial projections show the breakthrough could lead to 45% higher performance and 75% lower energy use than today’s 7nm processors.

While 2nm technology is a few years away from coming to devices, IBM’s success could benefit much of the chip industry. While AMD and TSMC’s advancements in recent years put them in a competitive market position, the implementation of 2nm will pull Samsung, GlobalFoundries, and Intel into a new decade of chip manufacturing. For buyers, the forecast of strengthened competition and new capabilities could mean big gains in processing power and energy and space savings.

Also Read: Cloud Giants Turn to Custom Chips

Here’s how IBM’s most recent innovation will lead to new features and capabilities for the entire semiconductor industry, and in the end, for buyers of everything from smartphones to supercomputers.

IBM’s latest innovation: Nanosheets and GAAFET

From their Albany Nanotech Complex in New York, IBM’s semiconductor development efforts continue to offer a pipeline for innovation. IBM’s contributions to chip technology can’t be understated, from crafting the first 7nm and 5nm node technologies to single-cell DRAM and 3D chip stacking.

Also Read: IBM Reaches 7nm Milestone for Chips

Unlike transistors of the last fifteen years, IBM’s 2nm design notably does not use a FinFET architecture. Instead, the innovator opted for the world’s first stack of ultrathin nanosheets that combine to produce a gate dubbed GAAFET. For IBM products, the 2nm transistor will be center stage in their POWER series for server processors and zArchitecture for mainframes. With Samsung and GlobalFoundries as expected manufacturing partners, the newest chip design could be available by 2024.

Projected features of 2nm

Using IBM’s nanosheet technology, the 2nm design will fit up to 50 billion transistors into a space roughly the size of a fingernail, up from 30 billion offered by the 5nm in 2017. With more transistors packed in, IBM has also infused more capabilities relevant to AI, cloud computing, and security. Expected benefits outlined by IBM include:

  • 4x increase in cell phone battery life
  • Decreased data center carbon footprints
  • Increased laptop functioning speed
  • Faster object detection

For the announcement, SVP and Director of IBM Research Darío Gil stated:

The IBM innovation reflected in this new 2nm chip is essential to the entire semiconductor and IT industry. It is the product of IBM’s approach of taking on hard tech challenges and a demonstration of how breakthroughs can result from sustained investments and a collaborative R&D ecosystem approach.

Also Read: Server Security Best Practices

Comparing Current Chips

VendorForm FactorDensity
IBM2nm333 million transistors per mm2
TSMC5nm173 million transistors per mm2
Samsung5nm127 million transistors per mm2

Coming soon: POWER10

The announcement comes as IBM prepares to launch the POWER10 CPU family. In a 7nm form factor, the newest iteration has three times greater energy efficiency, workload capacity, and container density relative to the POWER9. Other features include:

  • Memory inception: Support for multi-petabyte memory clusters and intensive workloads
  • Hardware-enabled security: 4x the AES encryption capabilities of its predecessor
  • Matrix math accelerator: 20x faster AI inference for FP32, BFloat16, and INT8 calculations

Fit for the hybrid cloud environment, IBM’s POWER10 is co-optimized with Red Hat OpenShift for container orchestration. For the time being, manufacturing for IBM’s newest chip belongs to industry-leading Samsung.

Also Read: Top Rack Servers of 2021

Will IBM’s breakthrough benefit Intel?

Intel and IBM have a partnership going back nearly 40 years, so IBM’s 2nm achievement is good news for Intel. While Intel’s delays developing 10nm and 7nm chips have left analysts wondering what happened, AMD and TSMC jumped ahead with their next-generation 7nm offerings.

Read More: AMD & Intel: A Tale of Two 3rd Gen Processors

While Intel isn’t IBM’s primary manufacturer, some analysts predict this innovation will give Intel the resources to skip over a Moore’s Law generation. From investing $20 billion in developing their 7nm for a release in 2022, Intel could be jumping to 2nm development by 2024. A few years before that happens, AMD and TSMC hold the competitive edge, and Intel has an opportunity to rise to the challenge later this decade.

Industry impact and the IBM Joint Development Alliance

The successful implementation of the GAAFET chip design and IBM 2nm is big news for the industry. While the concept isn’t new, IBM is the first to make it work, and the rest of the semiconductor market will adapt.

Progress is likely for the entire industry in part due to an on and off again culture of “radical collaboration.” To stay competitive, each vendor puts billions of dollars towards R&D and capital investments. With a window of communication and information sharing, the IBM Joint Development Alliance is a case study of how commercial innovation can push an entire industry forward. From Samsung and Intel to AMD, TSMC, Nvidia, and more, IBM enables many of the industry vendors.

With 2nm expected to be the dominant processor by the mid-2020s, the operating devices we all use will see faster processing, reduced energy consumption, and a drastic increase in battery life.

Also Read: Best Server Virtualization Software for 2021

Latest Posts

Related Stories