by Anton Shilov
09/01/2010 | 10:42 PM
Nvidia Corp. is reportedly working on a PCI Express-based graphics processing unit that also integrates input/output functionality typically found in core-logic set. The chips will allow the company to partly return to the market of core-logic devices. But the success of the move may be undermined by technical details.
Intel Corp.'s upcoming client platforms based on the Sandy Bridge, Tunnel Creek and other microprocessors will come with PCI Express controllers that will allow connecting additional controllers directly to the central processing units (CPUs). Technically, this allows Nvidia to plug-in one of its graphics chips with I/O functionality and substitute Intel's own core-logic that connects to processors using DMI bus, a technology that Nvidia does not have a license to use. In fact, theoretically, Nvidia will also be able to hook up its rumoured chip to Advanced Micro Devices' code-named Llano processor with integrated graphics engine via PCIe bus as well.
Intel officially claims that Tunnel Creek and Queens Bay embedded platform will rely on third-party PCI Express controllers for additional functionality. Nonetheless, when it comes to Sandy Bridge processors the company plans to sell its own Intel 6-series (Cougar Point) chipset. Moreover, Intel 6-series core-logic sets come with integrated base clock-speed generator and it is highly likely that it will be technically impossible to start up an SNB chip without an Intel core-logic.
It is not completely clear whether Intel's next-generation Sandy Bridge platform actually needs low-end graphics processors considering the fact that Intel's next-gen integrated graphics promises tangible performance improvements compared to current Intel Graphics and Media Accelerators. The same can be said about AMD Llano, which is projected to offer graphics performance comparable to ATI Radeon HD 5500/5600-series graphics processors.
Neither Intel not Nvidia commented on the news-story.