Intel is announcing today a trio of new low-power Atom microprocessors for enterprise hardware. The company made the announcements as part of its Intel Developer Forum event in Beijing today. The new chips show the company is serious about moving low-power processing technology into the heart of the data center.

The three new Atom chips are code-named Briarwood, Avoton, and Rangeley, Lisa Graff, VP and general manager of the Datacenter Marketing Group, said in a press call. The Briarwood chip, formally dubbed the Intel Atom Processor S12x9, will be used in storage applications. It can provide up to 40 lanes of input-output PCIe 2.0 capacity and other features for storage devices. That chip is available now.

The Avoton and Rangeley processors will be out in the second half of 2013. Avoton is a second-generation 64-bit system on a chip that uses Intel’s most advanced manufacturing process, the 22-nanometer technology with a microarchitecture dubbed Silvermont. Avoton will be targeted at micro servers, or low-power versions of enterprise servers that use lots of low-power cores rather than a small number of brainiac cores. That chip is available in samples now. Rangeley will be used in networking devices, including mid-range routers, switches, and security appliances.

Yesterday, Hewlett-Packard, which is perhaps Intel’s biggest customer, launched an Atom-based HP Moonshot server family. Those servers will use the Intel Atom S1200 processor family. HP will introduce that family of servers in the second half of the year.

In the long run, Intel’s goal is to enable technologies that will lead to the creation of smarter cities, healthier communities, and thriving businesses, according to Diane Bryant, senior vice president and general manager of Intel’s Datacenter and Connected Systems Group. Worldwide, global datacenter traffic will reach 6.6 zettabytes by the end of 2016, according to the Cisco Global Cloud Index. Global Internet protocol traffic will reach 554 exabytes per month. To put that in perspective, mankind created only 5 exabytes of data from the dawn of civilization until 2003.

Intel has also created its own software distribution for Hadoop enterprise software. With it, China Mobile has been able to deliver inquiry results 30 times faster and with great ability to scale capacity in the future.

Intel isn’t ignoring the high-end of the server chip market either. The company will launch its 22-nanometer Xeon processor E5 family in the third quarter. It also has some other chips, dubbed E3 and E3, coming this year. That means Intel will have a complete refresh of its enterprise technology during 2013.

Lastly, Graff said that Intel is launching a Cloud Innovation Center in China to allow developers to test their software on Intel cloud technology so they can get products to market quicker.

---

HealthBeat 2013 is a new conference showcasing how technology is transforming health care. We'll explore how IT is driving out inefficiencies on the hospital, practice, and patient levels. Check out full event details here, and register here.

---

Sept. 9 - 10, 2013
San Francisco, CA

Early Bird Tickets on Sale

Intel is poised to ship its first chips for “microservers,” a hot new category where it has been trailing its rival Advanced Micro Devices.

AMD says that Intel has been playing catch-up, particularly since it bought microserver pioneer SeaMicro for $334 million early this year. But Intel, the world’s biggest chip maker, says it has actually been thinking about microservers — which use Intel’s low-power Atom microprocessors — for a long time. In an interview with VentureBeat, Intel Fellow Matt Adiletta (pictured below) said that he had been contemplating low-power servers since 2006 and he was the point person in getting SeaMicro, an innovative pioneer in microservers, to use Intel Atom chips in its first machines. AMD has also allied with ARM in hopes of outflanking Intel.

In any case, microservers have become one of the most competitive battlegrounds in the multibillion-dollar server chip market. The whole idea is that it is more efficient to use small, highly efficient processors rather than big supercomputing beasts to process huge numbers of small workloads, like the tasks that data centers handle in serving web traffic. Intel is briefing reporters on Wednesday morning about “a new technology for data centers” in San Francisco. That’s where the company is expected to introduce its new code-named Centerton version of the Atom microprocessor for microservers. Another version, code-named Avoton, is coming next year.

AMD is still offering SeaMicro microservers with Intel chips in them. But at some point, it will have its own chips designed for the purpose. In that respect, while AMD can say its SeaMicro is a leader in microserver systems, Intel can rightly say that it is the only company shipping processors designed specifically for microservers today.

And those will likely be based on the low-power ARM architecture that AMD licensed. ARM chief executive Warren East sees microservers as its wedge into the huge data center server microprocessor market that has long been dominated by Intel.

Adiletta faced some skepticism as he built the case for increasing the density of processors so they could be used by the hundreds in a tightly packed server cabinet — without melting down a data center. His idea was the distribute a computing work load across a lot of low-cost servers, which would be better than having beefy servers working on light loads and becoming idle at some point.

Adiletta said he was inspired when former chief technology officer Pat Gelsinger asked him to explore the low-power blade server market, where thin, low-power server cards were used in servers to scale up big data centers.He had a conversation with Andy Bechtolsheim, co-founder of Sun Microsystems and chairman at Arista Networks. Adiletta also worked with researchers at Carnegie Mellon University in 2009 and started envisioning workloads for an Atom-based microserver.

He was the point man for Intel in winning a deal with SeaMicro. He saw the servers as a way to get rid of cabling, improve infrastructure, and get computing resources online more quickly so they can be used as needed. Microservers are turning out to be great for the era of Big Data and cloud computing. AMD views Intel’s latest efforts as late and its chronicle of Adiletta’s work as an attempt to rewrite history. SeaMicro executives, now at AMD, say that Intel fought SeaMicro on its choice of Intel’s Atom processors.

But Intel has to tread carefully. Intel executives have said that microservers might be 10 percent of the server chip market. But if they grow larger than that, they could cannibalize Intel’s sales of larger, more expensive server chips. And that could hurt Intel’s bottom line.

Sept. 9 - 10, 2013
San Francisco, CA

Early Bird Tickets on Sale

Hoping to disrupt the market for data center chips, Advanced Micro Devices announced today that it will offer ARM-based server chips in the future. This strategic move is akin to a major league baseball team switching to football.

“This is a historic day,” said Rory Read (pictured), chief executive of AMD, the No. 2-maker of Intel-compatible chips.

AMD’s mainstay is designing x86 chips for PCs and data center computers known as servers. Thanks to the internet, data center chips are growing in sales at a rate of 33 percent a year. But the problem is power consumption. Electricity is becoming the biggest cost in data centers, and chips that are too hot are in danger of melting down.

So AMD will now design chips based on the ARM architecture, which typically has lower computing performance but much better power consumption. ARM-based chips are the heart of low-power portable devices such as tablets and smartphones. Now AMD will create 64-bit ARM-based processors that can be used in data center servers.

Those server chips will use less power, which allows more chips to be put in the same space without overheating. That, in turn, allows for more densely packed server chips. AMD will also bring to bear a new networking technology, dubbed a fabric, that it acquired with its purchase of Sea Micro earlier in the year. Sea Micro’s fabric allows chips to communicate within servers at much faster rates, allowing the low-power server chips to work together in a more efficient manner.

“It’s all about power and efficiency,” said Jay Parikh, vice president of infrastructure engineering at Facebook, who spoke at the AMD event in a panel that included representatives from ARM, Red hat, Dell, and analyst Nathan Brookwood of Insight 64. The improved chips and server technology will make it easier for corporations to deploy huge clouds, or web-connected data centers that centralized a lot of processing and storage tasks.

“There is no doubt that the cloud changes everything,” Read said. “The cloud is the killer application that is driving industry forward. We are uniquely positioned to lead in the era of cloud processors.”

Sea Micro pioneered a new class of servers known as “micro servers.” They included Intel’s low-power Atom chips that had older processing technology but used significantly less power. Sea Micro launched its first micro servers in 2010, putting 512 processors in a small cabinet that was about a quarter the size of a usual server cabinet. They also used a quarter of the power.

“Why is it important?” Read said. “The ecosystems are going to continue to evolve. We’ll apply the technology to segments where it has the best fit. We think ARM will play an important role. The x86 won’t go away. It will be around a long time.”"

Patrick Moorhead, analyst at Moor Insights & Strategy, said, “AMD’s announcement gives a lift to the entire micro-server market, and launch contributions from Facebook, Amazon, RedHat and Dell, help, too.  AMD brings over a decade of server experience at the chip-level and can offer customers AMD, Intel and now ARM solution choices. Intel, who currently has over 95 percent market share in the entire server market, has a while to formulate a response given 64-bit ARM servers ship in 2014.”

Read said the micro-server segment of servers will be the fastest growing segment of the server industry over the next five years. Lisa Su, senior vice president of AMD, said the ARM license will change over time as AMD moves more deeply into the business. For now, AMD is licensing a 64-bit core from ARM.

“The market will develop over time,” Su said. “The engineering teams are well under way.”

At some point in the future, AMD could still go deeper with the relationship by designing its own cores and doing a lot more customization by securing a license that allows it access to the ARM instruction set. That is the kind of license that Apple has with ARM. But the licensing an ARM core is the first step in that process. The delivery of the first chips is set for 2014, since ARM still has to finish the core in question and then AMD will have to do work integrating that core into a full working, fully tested chip.

In a separate conversation, Intel fellow Matt Adiletta said in an interview with VentureBeat that his own company has been exploring micro server technology going back as far as 2007-2008, when the focus was on low-power “server blades.” Over time, Intel has also become interested in the micro server market, but it did allow the market leader, Sea Micro, to be acquired by AMD.

“With its planned 64-bit ARM solutions, AMD brings the experience of a proven enterprise CPU provider to the ARM ecosystem,” said Jimmy Pike, vice president and senior fellow of the Dell Data Center Solutions group. “ARM has the promise of being a serious player in areas like web front-end servers and as a worker node in a Hadoop environment.  AMD’s opportunity is to deliver serious value in performance-per-dollar and performance-per watt-where low-power server platforms running massively scale out workloads can shine. The availability of 64-bit ARM solutions is an essential milestone needed to accelerate enterprise adoption of this technology.”

Brookwood said that the strategy is a good one for AMD, since Intel is very unlikely to take a license from ARM, which is Intel’s arch competitor these days. And AMD, which is far smaller than Intel, has to do things that Intel can’t or won’t copy.

“If you are AMD, you have to do things that Intel doesn’t want to do,” Brookwood said. “This is one of those things.”

Intel announced today that it will create a low-cost, low-power “micro-server” platform dubbed Centerton, aimed to deliver energy-efficient computing.

The move is a reaction to Advanced Micro Device’s acquisition of micro-server startup Sea Micro for $334 million in February.

Sea Micro used Intel’s cheap, low-power, low-compute Atom chips to build dense servers with lots of energy-efficient chips. It consolidated a lot of functions into its chip sets so that it could make tiny server boards. Sea Micro’s computers used a third of the power and a third of the space of normal Intel Xeon chip servers, but they had a 150 percent improvement in compute density, or the amount of computing power packed in a given space.

That woke Intel up, and it began refashioning its Atom chip line-up to cater to micro-server platforms. Now its Centerton platform will focus on taking micro servers to a new level using Intel chips, while AMD will likely put its own chips into Sea Micro’s future platforms. The world’s biggest chip maker made the announcement at the Intel Developer Forum event in Beijing today.

Intel isn’t acknowledging that it is reacting to a rival, of course, and it continues to minimize the importance of micro servers. It says the micro-server market could be 10 percent of the whole server market by 2015.

Even so, it acknowledges that the workloads for servers are diverse, from web hosting (serving web pages) to mission-critical computing (running stock markets). The industry has responded by expanding the choices for data centers, with rack, tower, and blade servers. Since 2009, when Sea Micro appeared, the micro-server category has been growing fast in data centers where compute density and electricity savings are critical, Intel said. Today, Intel’s server chips range from 15 to 45 watts.

Diane Bryant, vice president of data center computing at Intel, said that in the second quarter, Intel will launch a new 22-nanometer Tri-gate family of Xeon processors dubbed the E3 product family. It will then launch the Centerton chips, which can operate on six watts, later this year. Centerton will have two 32-nanometer cores, or computing brains and will have error-correction code memory and 64-bit support. It also does hyper-visor-based virtualization, which ARM-based rival chips are not expected to be able to do, when they show up on the market.

Patrick Moorhead, analyst at Moor Insights & Strategy, said, “This is a clear indication that Intel is raising their game in micro servers, even though they still think it comprises 10 percent of all servers in 2015.” He thinks that Intel will win on raw compute performance per processor, compared to AMD’s chips and expects the overall Intel solution to be competitive, but a full comparison will only be possible once the servers can be tested side by side.

Intel

The world’s largest chip maker normally designs server chips to run as fast as they can within a given power level so that it can be used in a traditional board and rack configuration within a data center. But the micro servers — created by companies such as SeaMicro (server pictured right) — try to pack a lot more single-chip boards into a smaller space with much lower power levels and a
smaller overall footprint.If these micro servers take off, Intel may be able to stave off low-power rivals such as ARM from taking a chunk of the server business.

These micro servers are a new vector for innovation in servers and Intel wants to foster their growth.The question is whether a standard will slow the market down or broaden it so that lots of companies can make more money in the market.Intel expects that micro servers could become 10 percent of the total server market in the next four to five years.

Boyd Davis, general manager of Intel’s Data Center Group, said that Intel plans to create micro server chips that consume as little as 10 watts by 2012. Today, Intel’s Xeon server chips run at 45 or 30 watts.In the coming months, it will offer a 20-watt Xeon chip and a 10-watt Atom chip designed for servers.

The micro server vendors have created “density-optimized” servers, or those with many small one-socket servers that share a chassis, fans, and switching to get higher efficiency and density (or more processors in a given space).They can get four times as many servers per rack, run on fewer fans and power supplies, and can still run a full range of single-processor applications. The idea is that certain applications will work better on many low-power dense servers than on a handful of more powerful but power-hungry servers. SuperMicro and SeaMicro have introduced non-standard machines in the recent past. But Dell, Tyan, Quanta Computer and Intel are moving toward standard modules.

Intel said that the micro servers will be appealing to web site hosting companies that have lots of customers with relatively low-traffic web sites.The company says that developers will be able to analyze their softwrae running on a range of Intel-baesd micro servers in a new Intel Micro Server Evaluation Lab.

The micro servers are not unlike low-power blade servers launched years ago. But at the low end of the market, Intel said there is enough computing power to do the job right.

SeaMicro uses 32-bit and 64-bit Atom processors in its small servers today. It can cram 256 dual-core Atom processors in a box that is just a quarter the size of a traditional rack, and it uses a quarter of the typical energy.

SeaMicro feels that the opportunity for low-power, purpose-optimized servers is well above the 10 percent of the total server market in the four to five year timeframe that Intel has suggested, said John Turk, vice president of operations at SeaMicro.

But he said this announcement from Intel recognizes there is real opportunity. SeaMicro is not working with Intel on the standards, as it naturally feels it has an advantage now. But it is encouraged that Intel is pushing for innovation in the micro server market.

In the big picture, Intel probably loves the fact that SeaMicro is an ally against ARM. Andrew Feldman, chief executive of SeaMicro, said in an interview that the fact that Intel has put an Atom chip on its server roadmap is a huge recognition of the fact that micro servers are a good idea and that Intel is serious about heading off a threat from ARM. That threat from ARM isn’t imminent, but it will likely come.

“These are all good steps,” Feldman said.

Intel’s own interest is in seeing lots of server vendors selling the best possible micro servers. So Intel is essentially motivated to help SeaMicro’s server competitors. It’s a combination of competition and cooperation.

Gio Coglitore, director of Facebook Labs, said on the call, “We aren’t trying to cram a desktop processor into a server anymore.”

Coglitore said that Facebook has tested micro servers and believes that deployment will happen in the 2011 and 2012 time frame.