XGI’s New Low-End Part to Sport Memory Sharing Tech

XGI's XG47 to Copy TurboCache, HyperMemory Approaches

by Anton Shilov
12/28/2004 | 11:06 PM

XGI’s forthcoming graphics processors aimed at entry-level market segment are projected to support a technology that allows the company’s visual processing units to use system memory as frame-buffer. Similar capability is also available from market leaders ATI Technologies and NVIDIA Corp..



“XGI’s [XG47's] new memory sharing system will be very similar to NVIDIA’s recently launched TurboCache technology, which uses a cache and allows a GPU to render directly to system memory,” according to a report from VolariGamers web-site.

Officials from the company did not comment on the news-story.

NVIDIA’s recently announced TurboCache technology allows graphics cards makers to equip their products with small amount of memory, down to 16MB, but still deliver relatively high performance, as the VPUs that sport the tech are able to store textures and other data in the system memory. This allows graphics cards makers to save on graphics memory, but requires some additional system memory.

Given that PCI Express bus has about 8GB/s transfer rate, while some modern entry-level graphics cards are equipped with memory providing only about 3GB/s of bandwidth, graphics cards with reduced amount of onboard memory may easily compensate the lack of high-speed onboard storage by using system memory with up to 8.5GB/s bandwidth and PCI Express bus with up to 8GB/s of transfer rate.

Similar capability called HyperMemory was also introduced by ATI Technologies.

XGI Confirms ‘Shader 3.0’ Capability

Earlier this year an XGI rep confirmed the company’s plan to support Shader Model 3.0 feature-set by the next-generation of the company’s VPUs.

“Next year we are pushing out XG45/XG47/XP10. XG45 supports Pixel Shaders 3.0 and native PCI Express, and the XG47 is the lower end. XP10 is for mobile [market]. There will be some special technologies [announced] at the end of the year/beginning of next year,” said Danny Lee, Corporate Marketing Manager for XGI, in an interview.

Earlier this year XGI was planning to unveil the new graphics processors already this year, but then redesigned its plans and said it would roll-out innovative visual processing units in early 2005. The new processors were expected to feature “enhanced capability that will allow XGI to compete head-to-head with its market rivals”. Some sources close to the company suggested that the new products for XGI would support Shader Model 3.0.

The first company to support Shader Model 3.0 – including pixel and vertex shader 3.0 – was NVIDIA Corp., which latest graphics chips all boast this feature. ATI Technologies has been downplaying the importance of Shader Model 3.0, but is expected to offer a range of VPUs with their support at some point in future.

No More Multi-GPU from XGI?

XGI’s products were the first that sported DirectX 9.0 capability along with a multi-GPU technology that allowed graphics card designers to install two such chips on a single graphics card to gain performance. However, it is unlikely that the company will go this route again, as Mr. Lee expressed pretty pessimistic attitude towards NVIDIA-driven SLI technology.

“SLI is not very cost effective and the market is very small. Thus, we do not plan to target that market. We want to focus our marketing effort on most of the gamers' market and not on specialized markets,” said Mr. Lee.

XGI’s new chips are expected to be released in early 2005.