Skip to Content

The Social-Network Chip

The growth of social media, search, and shopping could help chip startups get a foot in the data-center door.
October 14, 2011

Looking at friends’ pictures on Facebook or searching résumés on LinkedIn are relatively simple computing tasks in which information is called up, retrieved, and then shipped to a user’s screen from a distant data center. Yet such tasks are handled mostly by powerful microprocessors designed for more complex jobs like number crunching and running operating systems.

Cool compute: Low-energy processors are etched on a silicon wafer.

That means a waste of electrical power, says Ihab Bishara, director of cloud computing products at Tilera, a chip startup in San Jose, California. Microprocessors serving the cloud are too powerful, he says; in the future, he believes, many tasks carried out in data centers will be handled by cheaper, low-power chips like those his company makes.

Currently, the chips inside data-center servers are nearly all manufactured by Intel, which commands roughly 90 percent of the server market with its family of Xeon microprocessors. Xeon chips have up to 10 processing centers, known as cores, that work in parallel to do hefty computational lifting. In contrast, Tilera’s chips contain up to 100 smaller, lower-power cores. When networked together, the cores are capable of handling common cloud applications like retrieving user data while consuming about half as much electrical power, Bishara claims.

Electrical power use is an increasing economic concern for companies such as Facebook, Salesforce.com, and Google. Data centers now consume about 1.5 percent of the world’s electrical power. Electric bills currently account for one-third of the cost of running a data center, according to recent estimates from Amazon, and that percentage is expected to rise steadily as the price of computer equipment falls.

Some cloud operators are already starting to put computationally intensive jobs on servers that can handle them while shifting simpler tasks to low-power servers, says Reuben Miller, a senior research analyst with IDC. “Large companies [need] processors that are more power efficient,” he says. “It’s creating opportunities.”

Low-power contenders include Tilera as well as SeaMicro, which makes servers using Intel’s Atom processors (and sells them to buyers like France Telecom and Mozilla), and Calxeda, a company that builds low-power servers using mobile-phone chips from ARM Holdings.

Hungry for energy: Data centers now consume as much as 1.5 percent of the world’s electricity. Electricity accounts for about 30 percent of the cost of running a data center, leading to demand for low-energy processors.

Intel is likely to remain dominant, not least because of the large amount of software that’s already designed to run on the company’s chips. Intel executives also say that performance still matters more than power consumption for many cloud applications, such as data mining and financial services. “It’s about the most useful work done per watt per dollar,” says Raejeanne Skillern, director of cloud computing marketing for Intel.

However, IDC’s Miller says that as simple cloud computing tasks proliferate, the market for other chip designs will expand. In the next few years, he says, “I think Intel has the potential to see its market share come down.”

Bishara believes that changes in the market for servers could speed the adoption of new chip designs. Ten years ago, he says, no company bought more than 10,000 servers annually, but today companies like Amazon, Google, Apple, and Baidu collectively buy hundreds of thousands every year. “You’re getting a little bit of a Walmart effect in the supply chain,” he says. Today big buyers can demand new types of less expensive chips custom-designed for the cloud. “Before, the supply chain was controlled by Intel,” Bishara says. ��Now companies can make a choice.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.