Anyone involved in high-performance computing understands that among the most crucial elements of a successful cloud implementation is storage and memory technology that complements all ingredients of the computer system. Memory, in particular, must provide speed, capacity and scalability. Each of those characteristics is critical, especially with the introduction of Intel’s Xeon Processor E5 family. Servers based on this powerful new processor technology are poised to both redefine high-performance servers, as well as place enormous pressure on component vendors to develop devices that will optimize cloud implementations.
At Inphi, we’ve accelerated this optimization by developing memory buffer products for load-reduced dual-inline memory module (LRDIMM) technology. We’ve earned the leadership position in productizing our memory buffer component as well as helping to enable the entire ecosystem. LRDIMM does more than just add speed to the memory in cloud servers; it enables those servers to optimize memory capacity, as well as speed. By replacing a memory module’s register with an isolation memory buffer, the capacitive load is reduced on the signal drivers, thus enabling the servers’ memory to easily scale upward to higher capacity and speed. As an example, LRDIMM allows a Xeon E5 2600 to scale to 768GB running at 1333 MT/s. This is a 50 percent increase in capacity and a 66 percent improvement in speed, over the previous technology. Additionally, LRDIMM uses cost effective mainstream DRAM components, which helps keep overall system costs down.
Inphi will continue to build on its LRDIMM technology leadership, develop advanced memory-buffering solutions and provide them to the fast-growing cloud infrastructure.