NVIDIA Corporation, a true expert in creating extremely complex and powerful semiconductors, showcased numerous its chipsets for mainboards during Computex Taipei 2003 trade-show. At the same presentation Mr. Jen-Hsun Huang demonstrated the NV38 aka GeForce FX 5950 Ultra the company unveiled four new chipsets: NVIDIA nForce3 150, NVIDIA nForce3 250, NVIDIA nForce3 250Gb and NVIDIA nForce3 Go 150.

Let us talk about those core-logic products each consisting of only one chip in a more detailed way.

The nForce3 150 and nForce3 Go150 are similar to the nForce3 Pro 150 with the exception that the latter supports both Socket 940 and Socket 754 CPUs, while the former two only boast with support for Socket 754 chips. It means that only nForce3 Pro 150 can be used with AMD Athlon 64 FX microprocessors.

The nForce3 250-series will not divide into families for desktops and workstations/servers since none of them actually support Socket 940 Opteron or Athlon 64 FX CPUs. In addition to nForce3 150’s basic capabilities that include AGP 8x, PCI, 2 ATA-33/66/100/133 channels with RAID 0,1 and 0+1 support, 10/100Mb/s Ethernet, USB 2.0, audio and so on the new nForce3 250 provides 2 Serial ATA-150 ports as well as external 2 Serial ATA-150 PHY support. The nForce3 250Gb also adds Gigabit Ethernet support to all those capabilities motioned above as well.

Mass production of NVIDIA’s nForce3 250-series will start in the fourth quarter of this year. Mainboards based on the new NVIDIA nForce3 products are expected to appear in November.

It is also very interesting to note that NVIDIA now positions its nForce3 Pro 150 solution for 2P servers and high-end workstations. The company suggests using AMD’s 8131 controller for PCI-X support as well as its own nForce3 Pro 150 for the rest capabilities.

As you see, NVIDIA promotes its single-chip core-logic products for various market segments. Given that features provided by single-chip solutions are enough for most end-users, we may expect a major shift towards single-chip core-logic sets for AMD64 platforms in future. Everything is not that simple with chipsets for Intel processors, keeping in mind that Intel’s MCH contains a memory controller, while the K8 processors boast with integrated PC3200 controller. Nevertheless, it will all be about manufacturing technology and ability to integrate more features into a single chip. NVIDIA already has this kind of experience from its graphics processors, moreover, the company has what is needed to increase yields and lower costs of such kind of solutions. One chip is better and more efficient than a number chips – NVIDIA and ATI Technologies already proved it earlier in the graphics field. Now its time to start bringing the same trend into the core-logic market!