A Step Forward in Workload Optimized Capability for the Data Center

The role of Intel architecture within the data center has grown tremendously since the introduction of the first IA server processor, the Pentium Pro, in 1995.  Within the last two decades our data center product lineup has grown from two processor configurations to over a hundred iterations addressing an equally expansive workload array found in the data center.  Almost as rapidly as we can develop new variants of processors we are seeing creative developers ask for another special optimization for their workload.  The move to mobility, and the massive apps ecosystem mentality it has spawned, have caused companies to rethink how they deliver services and thus how they are going to construct their data centers to meet ever growing demands.  Our mainstream Xeon processor line is fine tuned for what we believe is >90% of data center workloads with special attention paid to compute, storage and network requirements all delivered with a common instruction set architecture.  This foundation has enabled incredible industry innovation that has in turn enabled the next generation of supercomputing, the birth of the cloud, and the transformation of enterprise IT from supporting the business to becoming a strategic business asset.

In recent years we've identified adjacent market opportunities for data center workloads and have enhanced our product portfolio accordingly.  The introduction of the Xeon Phi and Atom product lines provide us the ability to meet emerging needs in highly parallel technical computing and extreme low power workloads respectively.  This reflects our goal of delivering the best silicon products available for every data center workload that customers require.

With advancements in silicon manufacturing and design, our ability to augment our roadmap of product offerings and be more nimble to market changes has grown.  With this nimble design capability we’re better able to deliver products that are specifically designed to unique workload requirements when we see a potential improvement in solution capability that merits a unique design.  We have over a hundred projects underway to deliver specialized SKUs today, most notably projects we've discussed with customers such as Facebook and eBay.

We have taken this approach even further with the planned delivery of an FPGA that has been integrated into the same package as the Intel® Xeon® processor.  The FPGA can be programmed to accelerate specific functions (e.g. video codecs or search algorithms) or re-programmed as workload patterns change.  Based on our internal estimates we believe packaging the FPGA with Intel Xeon processors using both coherent and non-coherent links will allow us to deliver up to >10x performance efficiency versus discreet FPGAs.  We see the FPGA integrated with Xeon solution as a major step forward in delivery of customer value initially targeted at hyperscale and telco environments.  While we don’t see FPGA based solutions as optimal for broad data center requirements in the near term, we see this as an excellent potential for a unique set of customers…and that’s pretty much what our customization efforts are all about. k