LRDIMM video

Webcast of HP, Samsung, ANSYS, Intel and Inphi presentation at IDF 2011 for HPC applications
 

Q: What are the biggest challenges that the server market is faced with today?

A: Enterprise servers are the workhorses of today’s data centers, running enterprise applications such as DBMS, CRM, ERP and mail services; executing database queries; and powering search engine operations. In addition, virtualized servers are key to cloud computing. All these functions, of both physical and virtual servers, require massive amounts of memory, as well as bandwidth.

However, enterprise servers are challenged today, more than ever, to meet the growing demands for memory capacity and available bandwidth. Scaling server memory technology to higher capacity and bandwidth requires pushing beyond existing technological capabilities, forcing IT and data-center administrators to make trade-offs that prevent optimization.

Q: What is a LRDIMM?

A: Load reduced DIMM (LRDIMM) is a new class of memory module. Designed with a memory buffer chip to replace the register to help minimize loading, the LRDIMM is capable of delivering upward of four times the memory capacity and nearly double the bandwidth.

Q: How does a LRDIMM work?

A: On an LRDIMM, the Memory Buffer technology addresses the high-speed signal integrity challenges of the memory interface by isolating the CPU and the memory components and buffering the memory I/O, control and data signals. In addition to handling the pre-buffer interface between the LRDIMM and the memory controller, the Memory Buffer also assumes responsibility for interfacing with the memory elements on the memory controller’s post-buffer interface.

Memory buffer technology reduces the load on the memory controller while also increasing memory capacity by accessing more DRAM post-buffer. Using a new option called Rank Multiplication, the Memory Buffer technology enables more ranks of dynamic RAM (DRAM) to be populated on the memory module and seamlessly accessed by the CPU memory controller. The result is that each server’s memory capacity can be multiplied as much as fourfold, while server bandwidths can increase by up to 36 percent compared to RDIMM approaches.

Q: What are the overall benefits of LRDIMM?

A:

  • Maximize Memory and Bandwidth– LRDIMM is capable of delivering upward of four times the memory capacity and nearly double the bandwidth.
  • Lowers Total Cost of Ownership– LRDIMMs are able to scale to much higher capacities and frequencies using cost-effective mainstream DRAM technology, which enables servers to support more higher-speed CPUs without adding more memory modules, and with lower power consumption. This also allows for data center servers to support the same amount of memory with many fewer servers.


Q: What technology was used prior to a LRDIMM?

A: Most of the memory scaling schemes have been based on some modification to the widely used dual-inline memory module (DIMM) architecture. For instance, the registered DIMM (RDIMM) is an architectural implementation introduced in the last few years to facilitate higher memory capacities and bandwidth compared to an un-buffered DIMM (UDIMM). The RDIMM uses a logic component called a register to “buffer” some of the signals that traverse the DIMM.  Although the RDIMM improves the capacity and frequency scaling to some extent, its performance is reaching its limits.

 

Q: Which companies provide LRDIMM technology?

A:

  • Samsung
  • Micron
  • Hynix
  • Elpida
  • Nanya
  • Kingston
  • Apacer
  • Adata

Q: Who can I call for more information?

A:

  • Sameer Kuppahalli, Memory Buffer Product Marketing Manager at Inphi
  • Phone: (805) 405-8387
  • Email: skuppahalli@inphi.com









Return to top Return to top

LRDIMM video

Hear Intel talk about LRDIMM technology