NVIDIA has had great success the past four months with their nForce 4 chipsets, ushering a number of technology advances for the Athlon 64/FX desktop market. With the ability to run two identical NVIDIA GPUs in tandem, the SLI has proved to be a great success for those looking for the best framerates possible in supported games. Along with PCI Express, SLI, their improved FireWall, storage solutions and other advanced features, the nForce 4 Ultra and SLI garnered top spots in our year end listing as must have products.

That's all fine and dandy with the consumer desktop market, but how about the server and workstation? Businesses were left out in the cold when NVIDIA first offered the nForce 4 solution but that has all changed today with the announcement of the nForce Professional. While the business market may not be as large in terms of retail sales, the market itself is huge, generating billions in business overall. Can NVIDIA successfully offer a comparable solution for workstations and servers, and more importantly, should these businesses care?

The nForce Professional

For both servers and workstations, NVIDIA is offering two solutions for the market. Today, NVIDIA is announcing the nForce Professional 2200 and nForce Professional 2050 chipsets (the latter being an I/O companion chip). Designed specifically for the AMD Opteron platform, both chips are scalable PCI Express MCPs. Since they are based on a single-chip architecture, they are cost effective solutions for the business market. As mentioned earlier, they support NVIDIA's Scalable Link Interface (SLI), as well as the NVIDIA Quadro Multi-GPU. It's also somewhat future-proof as well, supporting the new AMD dual core processors without having to redesign the motherboard or PC once they are released later this year.

The Chipsets are based on the 5th-generation HyperTransport design and is integrated tightly with AMD's Opteron Direct Connect Architecture. Being as such, both 32-bit and 64-bit processes are supported, and is validated to work with all Opteron CPUs (single and dual configurations) including the dual core Opterons expected later in 2005. Thanks to NVIDIA's Unified Driver Architecture (UDA), drivers released will be backwards and forwards compatible making adoption easier, and is offered for both Windows and Linux.

The nForce Professional 2200 MCP is the flagship chipset for the new line and supports a 16x16 HyperTransport link. It features a 4 controller PCI Express platform with 20 lanes. A new technology NVIDIA is adopting for the 2200 MCP is a new Gigabit Ethernet with a TCP/IP Offload Engine (TOE). The purpose of TOE is to offload TCP/IP load from the host CPU at the latter phases of the communication and as a result free up resources which are vital in an enterprise environment. Storage-wise, NVIDIA added a new feature called MediaShield RAID which eases the setup of RAID via a wizard-based interface and supports RAID-0, 1, 0+1, JBOD, and RAID 5. RAID-5 is a nice feature since up until now, only high-end SCSI cards were capable of that. The chipset also supports Native Command Queuing (NCQ) as well as 10 USB 2.0 ports and PCI 2.3.

The nForce Professional 2050 operates with the nForce Professional 2200 MCP and provides additional I/O capability. Like the nForce Professional 2200, the 2050 has four PCI Express controllers with 20 lanes, Gigabit Ethernet with TOE and NVIDIA MediaShield RAID. One thing to note is that with multiple 2050s, you can span RAID across up to a total of 16 SATA and 16 PATA drives (using 3 2050s and 1 2200).

Note that memory support is still registered DDR400, which is more a limitation on AMD's part as the memory controller is on-die.

Application of the nForce Professional

Workstation Specifications

For workstations, there are a number of options available for system builders. For the high-end, we can expect to see boards with both the 2200 and 2050 MCPs combined, supporting dual Opterons and SLI capable Dual Quadro GPUs. This combination is capable of 40 PCI Express lanes, 16 SATA and 4GB Ethernet.

For the midrange, board makers will probably stick with a single 2200 MCP, but dual Opterons are still supported but with only a single Quadro GPU. For the entry-level, it'll be the same as the midrange except we'll be looking at single Opteron setups.

Server Specifications

On the server side of things, the new features of the nForce Professional become obvious. The architecture is designed to be highly scalable and is possible to combine the nForce Professional 2200 MCP with up to three 2050 MCPs. This can conceivably bring servers a total of 80 PCI Express lanes, 16 SATA and 4GB Ethernet. PCI Express will allow hot-plugging (with compatible devices) which will be a major boon for IT admins, Message Signal Interupts (MSI) and Advanced Error Reporting (AER).

While not all the names will be familiar with our readers, all of the above are well regarded from within the IT community and credible partners for the launch. Industry-wide adoption will occur with more announcements over the next couple of months as products are introduced. Tyan and Iwill are offering products as of today though.

Final Words

Compared to what was previously offered for the professional market in terms of the Opteron workstation and server side of things, the nForce Professional is a vast improvement. While the performance is in question until we see some professional level benchmark numbers, feature-wise NVIDIA is putting a lot on the table.

As an IT Manager by trade, personally I'm very excited about the networking and storage options available. Sure, these technologies aren't exactly new for me, but if NVIDIA can deliver at the price point we're expecting (we don't have any numbers, but we expect it to be at a fraction of the cost of other solutions), there are going to be a lot of happy corporate accountants.

If you have any comments, be sure to hit us up in our forums.

HOME