Intel DDR2 FB-DIMM Performance

Written by Michael Larabel in Processors on 16 June 2006 at 01:00 PM EDT. Page 12 of 12. Add A Comment.

Will FB-DIMM have the same fate of RAMBUS RD-RAM? Likely not. Fully Buffered Dual Inline Memory Modules with its high-speed serial interface should prove to be an invaluable asset for mission-critical environments that require maximum performance and a low error rate, while it will likely be only a matter of time before measures are taken to deliver this same technology to desktop users. Servers using Fully Buffered DIMMs can now dramatically increase the number of modules in the configuration while also increasing the speed and reliability of the system memory. With the FB-DIMMs still in their infancy, require a greater number of PCB layers, and use an AMB controller chip, serialized memory is about 30% more expensive than standard DDR2. At the time of writing, the available selection of FB-DIMMs at the popular Internet retailers is also very limited; we would anticipate that additional manufacturers will begin offering their wares shortly.

Looking over the results from our tests performed, FB-DIMM memory is certainly capable of being the Speedy Gonzales of system memory. In fact, in many of the RAMspeed benchmarks, the FB-DIMM DDR2-533 was pretty much performing the same as non-ECC non-Registered DDR2-667. The higher memory bandwidth from FB-DIMMs will largely come when using a greater number of modules -- as we began to demonstrate in our second set of benchmarks. Meanwhile, the ECC Registered DDR2-400 and DDR-400 lagged behind the faster alternatives, but they remain good comparison values. When it came to the second set of tests that looked at the two versus four FB-DIMM memory occupancy and correlating these numbers to real-world performance, in Enemy Territory there was an evident gain while the performance delta in the other benchmarks were minimal.

To reiterate some of the features for FB-DIMMs is support for up to eight memory channels each with six modules, Cyclic Redundancy Check, bi-directional serial interface, and greater bandwidth. The process of upgrading memory has also been simplified with Fully Buffered Dual Inline Memory Modules, especially when it comes to being compatible with DDR3 FB-DIMMs. The downsides for this new memory adaptation are the increased heat output due to Advanced Memory Buffered (AMB) chip, and elevated costs. The heat output is approximately a few Watts per module, which has improved since the original FB-DIMM design. The cost of these modules should hopefully decrease with time and once higher volumes are shipped.

Though enthusiasts and desktop users should not look for this technology reaching their systems anytime soon, it may likely come down the road within the next couple of years. For now, Intel has been the only manufacturer to adopt this memory with their new Bensley platform, and according to a trusted and reliable source, it is not expected that Advanced Micro Devices will make the FB-DIMM jump anytime in the near future. Good-bye stub bus!

Feel free to discuss this article in our Phoronix Forums.

Look for additional articles on FB-DIMM Linux performance to come soon at Phoronix. A special thanks goes out to Tyan, Intel, and Kingston for their support with this article.

If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.


Related Articles
About The Author
Michael Larabel

Michael Larabel is the principal author of Phoronix.com and founded the site in 2004 with a focus on enriching the Linux hardware experience. Michael has written more than 20,000 articles covering the state of Linux hardware support, Linux performance, graphics drivers, and other topics. Michael is also the lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org automated benchmarking software. He can be followed via Twitter, LinkedIn, or contacted via MichaelLarabel.com.