NVIDIA 2200 MCP:

NVIDIA’s 2200 is a highly scalable chipset for servers and workstations alike. At a quick glance, it has support for 10 USB 2.0 compliant ports for more external peripheral connectivity options. Then it supports both SATA and PATA connections for those clients who may be interested in using a bit slower technology for the time being, but later upgrade to a much faster solution. That being said, we know PATA is going to lose support on consumer end boards by the end of this year, as SATA is quickly moving up the ladder and gaining support from enthusiasts. In essence, however, this feature seems more or less redundant for the time being. But for companies who don’t mind having a slower server, support for PATA drives may just be the option they might look for in 2200-based boards.

The SATA support is spanned out to four ports while four PATA drives (ATA133) can be connected via two IDE connectors. The 2200 MCP also features the SATA II spec. for up to 3Gb/s data throughput, which is becoming the next standard in storage rather quickly. Both PATA and SATA drives can co-exist in a RAID 0, 1 or 0+1 configurations. The hard drives are hot plugged, which is a great option, as it will system administrators to backup files and whatnot with an effortless step. In addition to the aforementioned features, NVIDIA’s RAID allows you add an additional drive to the system, so if a drive was to fail while the system is in operation, the system will recognize the failed drive and use the backup drive to keep functioning. To make the 2200 even more appealing, NVIDIA supports drives with NCQ and TCQ technologies for more data throughput and faster data transfers, so companies using the NVIDIA based chipsets can take advantage of the latest in storage technologies in full strength.

The chipset has two full 16x PCIe lanes for dual graphics cards, namely NVIDIA’s PCIe Quadro GPUs. In reality, though, the 2200 MCP supports 20 configurable PCIe lanes via four controllers. This will allow system vendors to support PCIe slots in x8, x4, x2 and x1 configurations. In addition, the PCIe lanes have virtual channels enabled, which theoretically guarantees full bandwidth per lane and low latencies.

2200 MCP Feature-Set

The 2200 MCP uses the fifth generation of HyperTransport link to connect to the microprocessor. The I/O bus should deliver maximum of 8GB/s data between the microprocessor and NVIDIA’s MCP. Not to mention that 2200 also features 32-bit slots to support current and older generation of connections. NVIDIA has done this to make the transition more smooth, since PCI Express still hasn’t caught up to its older counterparts, but you can be rest assured that PCI Express is the future and manufacturers should deliver PCI Express based products in the very near future.

NVIDIA’s 2200 has built-in Native Gigabit Ethernet and TCP/IP Hardware Offload support as well. The Native Gigabit Ethernet option is implemented on the MCP via HyperTransport link that directly connects to the microprocessor, in return reducing network latencies for more data flow. The TCP/IP Hardware Offload is network traffic manage controller that uses less CPU cycles to process network traffic.

Hardware aside, NVIDIA seems to have made the transition as effortless as possible for companies with a good number of server farms and whatnot. Not only is the hardware backwards compatible and has support for what seem to be outdated technologies, NVIDIA’s Unified Driver Architecture (UDA) is a welcome addition for software developers. Thanks to UDA, software developers will not have to worry about compatibility issues, as the aforementioned software architecture works with a single driver solution across all NVIDIA products. This guarantees easy installation, manageable maintenance and effortless upgrades.

With the amount of features that NVIDIA’s has implemented in their 2200 MCP, it is without a doubt that the aforementioned chipset is targeted at companies who can afford the cost of a having such a system. Undoubtedly, boards based solely on the 2200 MCP will be expensive and companies can expect to pay a fortune for future-proofing their servers.

Next