Powering a New Era: Extreme Low Energy Servers Process Tasks at Hyperscale

Nothing in business is more important to me than spending quality time listening to our customers.  To enable the cloud-based experiences that we all enjoy, bleeding-edge cloud data center operators taught us early on that deploying applications at large scale required new techniques.  Many of these cloud service providers have over 100,000 servers.  By focusing on increasing performance, lowering non-value added costs (such as power consumption) and enabling a standards-based approach, the Intel Xeon processor family now powers 9 out of every 10 cloud servers.

Back in 2009, we began to notice that a certain sub-set of cloud workloads were far less dependent on high performance and more dependent on just performing a simple task at cloud scale.  Examples of these less processor-intensive workloads are serving up web pages and streaming content. These tasks, while not requiring high compute requirements, have to be executed at high scale, servicing potentially millions of customers.  For these special cases, we realized that we had to create new extreme low energy server designs that reduced the number of components used, shared resources among hundreds if not thousands of daughter cards/cartridges, and were integrated into a cluster design that contained many smaller, power-efficient processor cores.  By doing this could potentially produce an additive effect in power savings that would be directly proportional to the number of servers they deploy.

We called them “microservers” and we pioneered the segment.  We initially addressed this segment by lowering the power Intel Xeon processors from 45W in 2010 to 17W this year due to our 22nm process technology.   Meanwhile, we began validating Atom for specific servers and increasing the feature-set and quietly in the background we were working on our first server-focused SoCs.

Back in 2008 Intel outlined a plan to use our chip design expertise, factory capacity, advanced manufacturing techniques and the economics of Moore’s Law to usher in a new category of highly integrated, purpose-built and Web-savvy System on Chip (SoC) designs and products.  Think of this as the close integration of system components for a specific purpose, such as memory, input/output (I/O), graphics & video engines, and networking on one integrated circuit.  Mobile devices benefited from this integration of many customized silicon features that could deliver features far more efficiently than just the processor alone.  Servers could benefit too.  At that time we already had over 15 SoC projects underway internally and we added a new project for the data center.  By designing more complex systems onto smaller chips, we could scale the performance, functionality and software compatibility of Intel Architecture while controlling the overall power, cost and size requirements to better meet respective market needs.

Today we launched our 3rd generation of microserver processors with the Intel Atom S1200 product family. There are more than 20 low-power servers, storage, and networking systems being designed from Dell, HP, Huawei, Inspur, Quanta, Wiwynn, CETC, Supermicro, Accusys, Microsan, Qsan, and Qnap. This is the first low-power 64-bit dual-core SoC for these data center systems that is in production and shipping now to customers, drawing only 6W of power and delivering four threads with Intel Hyper-threading.  Our SoC also includes key server features like 64-bit support for optimal memory use, Error-correcting Code (ECC) memory support for higher reliability and Intel Virtualization technology for enhanced workload management.  We put in these features because it was important to help Cloud providers not only reduce space and power for specific workloads to cut costs, but we also wanted to help them maintain the reliability and software compatibility associated with server class products that they’ve become accustom to.  When cloud providers deploy Intel Atom processors based microservers – it just runs – no software porting required.

And we’re committed to continue to lead in the microserver platform.  In 2013 our next generation Intel Atom processors codenamed “Avoton” will extend our SoC capabilities using 3D Tri-gate 22nm transistors and lead in power consumption and performance.  Starting from 2014 onwards we will leap further ahead in energy efficiency and performance leadership with 14nm manufacturing technology for low-power Intel Xeon products followed by Intel Atom product lines after that.  Our roadmap for this segment has Intel very well positioned.

Nothing in business is more important than spending quality time listening to customers, because they help you see new trends.  New Intel Microservers.  Now Served.

To learn more, visit us at our newsroom.

2 Responses to Powering a New Era: Extreme Low Energy Servers Process Tasks at Hyperscale