Both NVIDIA and ATI released new graphic chips target to low cost PCs based on the new PCI Express bus using a new technology that “steals” part of the PC RAM memory to be used as video memory. This technology is called TurboCache (or simply TC) by NVIDIA while ATI call this technology HyperMemory. By now ATI has released Radeon X300 and NVIDIA has released GeForce 6200 using this technology.
Since the video card uses part of the system RAM memory as video memory, the board can be manufactured with less memory, making it cheaper. On the other hand, your PC will have less RAM memory available.
The problem is that video card manufacturers are announcing their produts as if they had more memory than thet actually have, in the majority of cases telling that the VGA “supports 256 MB memory”, which can lead users to think that the video card has 256 MB memory, which is not true.
For example, the GV-NX62TC256D card from Gigabyte has only 64 MB of video memory but is announced as having “256 MB video memory support”. In fact, the difference between 256 MB and 64 MB it “steals” from the PC main RAM. The same thing happens on ATI arena. Radeon X300 SE from PowerColor, which has 128 MB video memory is sold as “with 256 MB HyperMemory”.
Brian Burke, from NVIDIA, sent us an e-mail clarifying some points and I’d like to share it with you:
“With TurboCache, the graphics driver uses system memory to augment the local graphics memory as needed by the application. As such, a GeForce 6200 GPU w/ TurboCache can deliver the same experience that a non-TurboCache GPU can. In fact a TurboCache GPU will outperform a non-TurboCache GPU at the same price point. Thus, it shouldn’t matter to an end-user where the actual memory is located.
We have packaging guidelines that we expect our add-in card partners to follow to be sure that their TurboCache products are clearly labeled. They include specifying the amount of local frame buffer memory on the card, and noting that a 512MB of system memory is required in order to see the full of support of 128MB. When it is brought to our attention that an add-in card partner is not following these guidelines, we work to have it fixed as quickly as possible.”
Brian also pointed out that GeForce 6200 with TurboCache can be available in several different memory configurations: 32 MB or 64 MB, with 64-bit interface or 128-bit interface, and with 275 MHz or 350 MHz memory clocks. So you should pay a lot of attention when buying a VGA with this feature, since a GeForce 6200 VGA from one manufacturer can be a lot faster than a GeForce 6200 VGA from another manufacturer, due to the memory configuration.
ATI has the same issue 64-bit/128-bit issue, but at least they change the product name: Radeon X300 goes for the 128-bit model and Radeon X300 SE goes for the 64-bit model.
As you can see, the idea behind this technology is the same as the on-board video – called UMA, Unified Memory Architecture –, which also “steals” part of the RAM memory to be used as video memory, but in the case of TurboCache and HyperMemory, these technologies at least come with some memory on the board (64 MB or 128 MB), using the main RAM only if the game needs more memory that the VGA has.
For those who doesn’t lnow, this idea is not new. It comes from the AGP bus, which also has this feature but was little used. By the time, the VGA manufacturers concluded that was better to put more memory on the video board than “stealing” RAM from the system, for performance reasons. But at that time the AGP bus ran at 528 MB/s and a typical PC had 16 MB of RAM, if that much. The new PCI Express x16 bus runs at 4 GB/s and a typical PC nowadays has 256 MB RAM.
The target market for this new low cost graphics chip generation is the users that want the cheapest PC available but want a 3D performance better than the one provided by on-board video solutions. But since these new video cards are based on the PCI Express bus, which was just released, they need a brand new motherboard, which may lead to a solution that is not that cheap – at least for now.
Leave a Reply