IBM's 2nm Chip Will Remake IT Industry
Page 1 of 1
IBM’s announcement last week that it has developed the world’s first 2-nanometer process for developing chips promises to reverberate throughout the IT industry in the coming years, from the largest enterprise computers down to the tiniest Internet of Things (IoT) devices.
Processors built from the process will deliver as much as 45 percent higher performance or 75 percent lower energy use than current 7nm node chips, which will have significant impact on the power consumption of cell phones, the capabilities of self-driving vehicles, the ability to run such advanced workloads as artificial intelligence (AI), cloud computing and the power efficiency of the largest data centers.
“This process increases performance and efficiency dramatically, allowing for smaller personal devices and more dense central computers,” Rob Enderle, principal analyst with The Enderle Group, told InternetNews. “The related performance boost is expected to be significant, allowing the devices that use the chips this will create to do more with far less energy and heat.”
Proof of Concept
IBM’s announcement essentially amounts to a proof-of-concept for the 2nm process, and there are still other steps that need to be taken, from adapting it to existing architectures and building a production fabrication facility – known as a fab or foundry – so it could be another year or two before actual 2nm chips become available on the market, Enderle said.
“But whichever architecture gets this process should also receive a significant competitive advantage,” he said. “This process will make for more powerful AIs that will function in more places. Think opening the door for intelligent robots... With hybrid cloud, this will improve performance across the board, but it will do the same thing for pure on-premises and pure-cloud, so its impact on that market is more of a ‘tide that raises all boats’ thing.”
Time of Change
The chip breakthrough comes at a time of change and strain in the global semiconductor industry. Chip makers like AMD and Nvidia currently have 7nm chips on the market, and chip manufacturing companies like Taiwan Semiconductor Manufacturing Corp. (TSMC) and Samsung can produce 5nm processors now. According to AnandTech, TSMC is scheduled to begin production of its 4nm chip process by the end of this year, with mass production coming in 2022, followed by 3nm nodes starting in the second half of next year. As for 2nm, that is still in early development.
Intel – the world’s largest chip maker – has had difficulties over the past several years hitting deadlines, such as with delays in the early development of its 10nm chips. It is still trying to make the jump to 7nm, which is now expected to happen early next year. New Intel CEO Pat Gelsinger in March surprised some industry observers when he announced the company was doubling down on manufacturing, from building new fabs in the United States (at a $20 billion cost) to leveraging third-party foundries to ramping up its own foundry business (see Intel's Bold New Chip Manufacturing Plan).
There also is ongoing consolidation within the semiconductor space as chip makers look to add to their capabilities. AMD, which over the past three-plus years has seen a resurgence in both the data center and consumer system markets with chips based on its Zen microarchitecture, is looking to buy Xilinx, which makes programmable processors – field-programmable gate arrays, or FPGAs – that are becoming more popular as accelerators for servers. The price tag on the proposed deal is $35 billion.
Meanwhile, Nvidia is proposing to spend $40 billion to buy processor designer Arm, whose architecture powers most mobile devices and is getting some traction in the server space.
All of this comes as an ongoing worldwide semiconductor shortage is squeezing a broad array of industries, from technology to automotive, and is a driving force behind an effort by the Biden Administration and U.S.-based chips makers to increase processor production in the United States and lessen the reliance on foundries in Taiwan and China.
The semiconductor shortage has forced some car makers to change or suspend production while manufacturers of consumer devices like smartphones and game consoles have said they may have to delay production and release of new products.
In a recent statement, Richard Barnett, chief marketing officer at Supplyframe, an intelligence platform for the global electronics value chain, noted the impact the shortage was having in even some of the more obscure industries.
“The current state of the supply chain and semiconductor shortages are impacting industries worldwide, hitting big-name organizations and smaller businesses alike,” Barnett said. “Surprisingly enough, the semiconductor shortage has even begun to affect our furry friends. CCSI International, a manufacturer that produces electronic dog-washing booths, had to change its circuit boards as a result of chips being out of stock. The ripple effect of these shortages and their cross-industry effects demonstrate the need to build resiliency into product designs, and the severity of the impact when organizations do not.”
Nanosheet Tech to Benefit All
IBM’s 2nm announcement is good news not only for IBM but for the chip industry as a whole, Patrick Moorhead, principal analyst with Moor Insights and Strategy, told InternetNews.
“This gives me better confidence that Samsung, Intel and TSMC will be able to have access to 2nm,” Moorhead said. “2nm products should provide significantly lower power, higher performance or a combination of the two. IBM will continue to monetize the IP that it creates and sells to companies like Samsung and Intel. This same 2nm technology will make it.”
The 2nm work is being done in a research lab at IBM’s Albany Nanotech Complex in Albany, NY, where it also has developed its 5nm and 7nm process technologies. IBM’s commercial 7nm chips will appear later this year in its Power10-based systems. With 2nm, the company is leveraging its nanosheet technology, which company officials said will enable them to put up to 50 billion transistors on a chip the size of a fingernail.
Better Performance and Power Efficiency
An AnandTech report said the new 2nm chips will include about 333 million transistors per square millimeter, compared with about 173 million transistors per square millimeter for TSMC’s 5nm process products. The performance and power efficiency gains in the 2nm process can lead to everything from a quadrupling of the battery life of cell phones – requiring that users charge their devices every four days – to the reduction of the carbon footprint of data centers, which currently consume about 1 percent of energy worldwide, according to IBM officials.
“By increasing overall performance and reducing latency due to vastly higher system density,” enterprise IT from the cloud to on-premises data centers will see a boost, Enderle said. “The benefit is jobs finish far faster and parts can more easily work on solutions that currently don’t allow for the power and heat of current technology.”
For laptops, it could mean vastly improved performance in application processing and internet access, while speeding up object detection and reaction time in autonomous vehicles.
It also fits in with IBM’s growth strategy, which is centered around AI and hybrid clouds and uses Red Hat open technologies like the OpenShift Kubernetes platform as its foundation. Moor’s Moorhead said that the IBM’s new Power systems and Z mainframes “are all about Linux and Red Hat cloud workloads and will take advantage of 2nm.”