NVIDIA nForce Computex 2001 Preview

StreamThru

One feature of AMD's Hyper Transport bus is the ability to support multiple virtual channels. We believe that a particular implementation of this is what NVIDIA is calling StreamThru as a part of their nForce technology.

StreamThru basically guarantees that regardless of what other demands are placed on the chipset, the devices that are latency and bandwidth critical will receive the ultra fast access and bandwidth to main memory that they need in order to operate properly.

Devices that are latency and bandwidth dependent are known as isochronous devices, those that aren't latency dependent (meaning they can be interrupted for a bit in order to let another device occupy some time) are known as non-isochronous devices.

An example of an isochronous device is the integrated 10/100 network controller in the MCP. NVIDIA's internal architecture, made possible courtesy of Hyper Transport's support for virtual channels, will allow the 10/100 network controller to always get the bandwidth and low latency access to the graphics, CPU and main memory that it needs to operate without any interruption.

This is what NVIDIA is calling StreamThru but really the technology is made possible by Hyper Transport. It is useful nonetheless.

Bringing new meaning to the term "all in one"

NVIDIA is promising a single driver for all the components of the nForce chipset, including the integrated graphics, audio, Ethernet, HomePNA, busmaster IDE, chipset, etc. This of course makes setting up the drivers on an nForce system incredibly easy compared to any other chipset where the graphics, audio, chipset, busmaster, etc. drivers are not only independent from each other, but often come from different manufacturers.

One side benefit of this setup is that if you decide to upgrade with an NVIDIA-based graphics card, you won't have to install any new drivers if you already have the latest on your system.