Support|documentation

  Xcell Journal Online
  Xcell Journal Archives
   
  Writing for Xcell
  Advertising in Xcell
  FREE Subscription
   
  Partner Yellow Pages
  Reference Pages
  Contact Us

    

Home : Documentation : Xcell Journal Online : Article
Optimize Memory Subsystem Performance with Network FCRAM



by Scott Beekman, Business Development Manager, Toshiba America Electronic Components, Inc.
scott.beekman@taec.toshiba.com (1/15/05)


Toshiba’s Network FCRAM often provides the best cost/performance by combining DRAM densities with random cycle performances that approach SRAM speeds.
article link to PDF
Article PDF 478 KB


Among the many cost/performance tradeoffs system designers face, one of the critical decisions in network systems, communications equipment, and high-performance consumer electronics is the type of memory to use to ensure that performance can keep pace with the processor.

Traditionally, network system designers had to choose between dynamic random access memory (DRAM), available at a lower cost-per-bit because of the high volumes used in personal computers, or higher performance static random access memory (SRAM), available only in low densities and at a much higher cost. A combination of the two is typically used with DRAM for buffer memory and SRAM for look-up table (LUT) memory.

More recently, high-performance, lowlatency DRAM solutions developed specifically for high-bandwidth applications, including Toshiba’s™ Network FCRAM™ (fast cycle random access memory), provide another alternative. Which type of memory is right for your particular system? What additional requirements for memory controllers are associated with each choice?

Generally, you can choose the option that provides the highest performance within the system’s specified cost constraints, and in the time available to bring the system to market. In many cases, Network FCRAM provides the best cost/performance for networking and communications customers by combining DRAM densities with random cycle performances that approach SRAM speeds. This allows equipment manufacturers to develop higher performance, lower cost, and lower power communications systems than they could with double-data-rate synchronous dynamic RAM (DDR SDRAM) and high-speed static RAM (HSSRAM).

In this article, we provide an overview of Network FCRAM and the advantages it offers in comparison to standard DDR SDRAM or high-speed SRAM, and discuss the alternatives available for memory controllers supporting Network FCRAM.

Network FCRAM
Toshiba Network FCRAM is a high-performance, low-cost replacement to DDR SDRAM and high-speed SRAM targeted primarily for buffer memory and LUT memory in networking/telecom applications. Network FCRAM incorporates enhanced DRAM technology optimized for the high-bandwidth, lowlatency requirements of network and communication systems. Narrowing the active memory area achieves low power consumption and random cycle time performances almost triple that of standard DRAM.

Network FCRAM devices offer the following advantages:

  • Fast random cycle time (tRC) of 20 ns to 25 ns
  • Fast data transfer rate of 666 Mbps+ (For purposes of measuring data transfer rate in this context, megabit per second and/or Mbps = 1,000,000 bits per second.)
  • Large density up to 512 Mb (When used in relation to memory density, megabit and/or Mb means 1,024 x 1,024 = 1,048,576 bits. Usable capacity may be less. For details, please refer to specifications.)
  • Simplified command input
  • Low power consumption
  • Multiple sources
Network FCRAM technology excels in applications where you need DRAM densities and random cycle performance approaching SRAM-like speeds. Its high bandwidth and low latency makes Network FCRAM suitable for network applications, cache applications, and high-performance consumer applications. Typical network equipment applications include packet buffer memory, table look-up memory, and external cache memory in servers. Network FCRAM is also being used in digital consumer and supercomputer applications.

Performance Comparison
Network FCRAM and the specification-compatible, dual-source Samsung™ Network DRAM™ feature one of the shortest cycle times and latency among existing DRAM. As a result, Network FCRAM can improve system performance approximately 20 to 25 percent in comparison to DDR SDRAM. This is achieved as a result of higher data transfer rates, as shown in Figure 1, and an approximately threefold faster random cycle time (tRC), as shown in Figure 2.

As an alternative to HSSRAM, Network FCRAM costs approximately 1/16th as much per bit, and offers much higher densities (up to 512 Mb) compared to maximum densities of 36 Mb or 72 Mb for HSSRAM. Network FCRAM offers not only performance improvement alternatives but also lower-cost solutions, as shown in Figure 3.

Customers today are taking advantage of these features to boost performance and bring down their system’s cost by replacing DDR SDRAM with Network FCRAM, thus reducing chip count and board space because of Network FCRAM’s higher performance, and/or by replacing HSSRAM.

Selecting the Right FCRAM
Network FCRAM is available with a selection of interfaces, speeds, and organizations to meet various requirements:

  • 256 Mb (x8/ x 16) Network FCRAM1 (up to 400 Mbps with tRC = 25 ns)
  • 288 Mb (x18) Network FCRAM2 (up to 666 Mbps with tRC = 20 ns)
  • 288 Mb (x36) Network FCRAM2 (up to 666 Mbps with tRC = 20 ns)
  • 512 Mb (x8/ x 16) Network FCRAM1 (up to 533 Mbps with tRC = 22.5 ns)
Network FCRAM1 supports non-ECC bit densities (such as 256 Mb and 512 Mb as a single component), while Network FCRAM2 supports ECC bit densities (such as 288 Mb with roadmaps to higher densities).

Memory Controllers
Once you have selected Network FCRAM as the memory of choice for a design, the next step is to determine the best source of a memory controller for your system. For large-volume applications, some customers develop custom ASICs that include the memory controller; in addition, many network processors (NPUs) now support Network FCRAM. However, for many smaller volume applications, FPGAs offer lower cost and faster time to market.

Xilinx® Virtex-II™, Virtex-II Pro™, and Virtex-4™ FPGAs interface to Network FCRAM.

When evaluating memory alternatives for network systems, consider the performance advantages of Network FCRAM and the time-to-market advantages of an FPGA-based memory controller.

Development Tools
Toshiba offers several design guides to help customers and systems architects identify the key advantages of incorporating Network FCRAM technology into their high-performance applications. Network FCRAM devices are also supported by advanced simulation models to facilitate and accelerate design-in activity. Models supported include Verilog™, HSPICE™ and IBIS models, and SOMA models jointly developed by Toshiba and Denali™ Software Inc. For more information, visit www.fcram.toshiba.com.

Conclusion
As a result of Network FCRAM’s cost-performance advantages, today it is designed into more than 100 network solutions at more than 70 companies. Toshiba first introduced Network FCRAM working samples in 1999 and has continued to expand its product offering and build momentum in the network/telecom market.

Today, Network FCRAM is in production with data transfer rates as high as 666 Mbps and random cycle time performance as low as 20 ns. Toshiba now supports three densities in mass production, with higher density, higher bandwidth, and faster devices planned for 2005.

The official Network FCRAM/DRAM website can be found at www.networkfcram.com.

Printable PDF version of this article with graphics. PDF logo (1/15/05) 478 KB

 
/csi/footer.htm