"If you are looking for a reliable nevertheless highly performing memory expansion for your servers and workstations, Fully Buffered DIMM is just what you need. This new technology provides an unprecedented signal quality, thus significantly increasing system stability. FB-DIMM technology is a safe investment into the future and nothing less than a must for enterprise-critical applications", says Gerald Diercks, Managing Director of Memorysolution GmbH. „There is another unique advantage of this technology designed especially to meet the requirements of the demanding market of enterprise platforms. Resulting from recent advancements in processor performance the need for higher memory frequencies arises to fully benefit from the increase in performance. Up to now, the maximum memory equipment of a system was the limit: total capacity per channel had stayed flat at best if not had to be decreased in certain instances. With FB-DIMM, this is no more the case; you may even increase the total amount of RAM to a new high of 192 GB. This perfectly satisfies the ever-increasing appetite for RAM of both current and future operating systems and enterprise-level applications. It also perfectly fits the latest speed and I/O improvements of high-speed processors and server architectures.“
Technical Background:
FB-DIMM is the new memory subsystem bus architecture technology standard specified by JEDEC. Its design allows for the increased requirements of the most demanding mission-critical fields of application. Central element of this architecture is the Advanced Memory Buffer (AMB). In the prevailing bus architecture, the memory modules connect to the system bus by a shared parallel interface. In contrast, you would now find a point-to-point serial interface between memory controller, buffer (AMB) and the first chip and among the single chips themselves.
The serial interface technology of the AMB results in a boost in performance and optimal scalability. Opposed to old-style parallel Registered DIMM, memory capacity is significantly increased (approx. by factor 24) as is the available memory bandwidth (approx by factor 4) up to an impressive maximum transfer rate of 4,8 Gbit/s per I/O-Pin. Simultaneous read- and write operations further increase the effective bandwidth per memory channel. They come in the same DIMM form factor of today's commodity DRAM chips and with the same number of pins, however sport a changed pin-out.
On every FB-DIMM, the so-called AMB chip takes over the distribution of data to the single chips on the DIMM. The speed of bus access no longer depends on the I/O speed of the DRAM itself. The AMB chip buffers the data and transfers them to the next DIMM or the memory controller. Similar to serial PCI Express™ technology, FB-DIMM serial communication uses 24 pairs of wires between the chipset and the memory sockets. FB-DIMM supports eight modules (of 1 gigabyte DRAM at maximum) per channel (6 channels) in two ranks per DIMM.
Substantial data security features perfectly complete the concept of built-in reliability. However, the bundle of advantages results in an increased energy consumption and thermal output mainly caused by the AMB. Due to the slightly different dimensions (the necessary cooling elements results in considerably thicker modules) these special DIMMs may only be operated on designated mainboards. The upside is their unrivalled performance and reliability.
According to information released by manufacturers the Silent Data Corruption Failure in Time (SDC-FIT-Rate) compared to the complete system is decreased by the factor 1000. This unprecedented low failure rate is the result of a whole bundle of new technologies. A so called Transient-Bit-Error Function extends current ECC (error-correcting code) and CRC methods of checking the integrity of data in DRAM to include protection of commands and address data. Additionally, FB-DIMM technology automatically retries when detecting transient bit errors. Reliable availability of data is increased by the so-called Path-through-path logic subsystem. Bit Lane Fail Correction detects defective data wires and corrects this error by internally rerouting the signal path over a fully functional wire instead. All corrupt data memory areas are logged in an Error Registry inside the AMB. In excess of a certain threshold, diagnostical software triggers a message to the system administrator, who can exchange the module before any loss of crucial data occurs.
The Hot-Add Function allows for an exchange of memory modules during operation of the system without deactivating the memory bank prior to this - thus ensuring uninterrupted operation, increasing system availability and stability.
The new modules are on stock now. Please request actual prices, contact person at Memorysolution is Ralf Thoma.