by Anton Shilov
07/30/2009 | 10:01 PM
Micron Technology, on Thursday announced that it had produced the industry’s first DDR3 load-reduced, dual-inline memory module (LRDIMM) and will begin sampling 16GB versions this fall. By reducing load on the server memory bus, Micron’s LRDIMMs provide the option to support higher data frequencies and significantly increase memory capacity on next-generation servers running twelve-core AMD Opteron 6000 or eight-core Intel Nehalem EX.
Most midrange enterprise servers today utilize approximately 32GB of memory per system, but this is expected to more than triple by 2012, according to a recent report from Gartner. With server manufacturers continuing to take advantage of multi-core processors and data centers opting for efficient virtualization technology, memory requirements are being driven ever higher. By increasing the available memory a server system has, it is able to run more programs concurrently, handle larger data files more efficiently, and generally exhibit better overall system performance.
The new LRDIMMs will be manufactured using Micron’s 2Gb DDR3 memory chips made using 50nm fabrication process with 1.35V voltage, allowing the company to easily and cost-effectively increase server module capacity because of the chips’ high-density and small die size. Micron’s
Micron’s LRDIMMs currently use Inphi’s recently announced isolation memory buffer (iMB) chip in place of a register to reduce the bus load when transferring data between the memory and processor. Micron’s new LRDIMMs reduce this load by 50% for a dual-rank module and 75% for a quad-rank module, when compared to today’s standard DDR3 server modules – registered DIMMs (RDIMMs). By reducing the load on the bus, Micron’s LRDIMMs enable servers to handle higher frequencies of data to improve overall system performance and support increased number of modules for greater system memory capacity.
“With the rise in virtualization, our new 16GB modules allow customers to easily expand their memory capacity. While traditional RDIMMs limit the amount of memory that can be accommodated due to their loading profile, LRDIMMs eliminate that problem by reducing the module load,” said Robert Feurle, vice president of DRAM marketing at Micron.
Today, using RDIMMs, a typical server system can accommodate up to three quad-rank 16GB RDIMMS per processor. However, that same system can support up to nine quad-rank 16GB LRDIMMS per processor, pushing the memory capacity from 48GB to 144GB. Measuring performance levels, Micron’s 16GB LRDIMM offers an increase of 57% in system memory bandwidth, when compared to an RDIMM. As server power consumption continues to be a top concern for customers, Micron’s LRDIMMs will also operate at the industry’s lowest 1.35V.
Micron is currently sampling an 8GB LRDIMM with select enablers. Mass production of its 16GB LRDIMMs is expected to begin in 2010 in time with the launch of next-generation enterprise-class server platforms by AMD and Intel.