![]() |
|||
|
|||
| :
News : Reviews : Editorials : About : Contact : Advertising : Privacy : Links : Forums
|
|
The State of 3D |
|
|
NVIDIA and ATI, 1H 2005 |
|||
|
|
|||
|
by Josh Walrath |
|||
|
The R480 and R430 series of cards were supposed to be hitting the shelves by the end of December, 2004. This of course did not happen, and these cards were not sighted until well towards the end of January. As of February though, supplies of the X850 XT PE can be seen as “good”, but the price premium was well into the $600 range. The new X800 XL finally started to hit retail in good numbers at the beginning of March, but there was a hefty price premium for this part. Users initially were paying upwards of $430 for this $299 part. As of the end of March, these prices have finally reached the $299 point (and below). As with any new technology, there will always be a sharp drop in price after about a month post-introduction. Currently ATI’s position is still very solid, and with the recent release of the Rialto bridge chip, we will see the AGP side of their products gain a much needed refresh. While NVIDIA’s HSI chip may be slightly more advanced (it can bridge both ways) and was out nearly a year before Rialto, ATI’s part is very solid and it does the job it was intended to do. Until this point, ATI’s midrange has been addressed by older 9800 Pro boards in either 128 bit or 256 bit memory formats. It is not exactly cost effective to sell these parts at those price points (especially the 256 bit boards). Getting the X300, X600, and X700 to the AGP side is a much needed move for ATI to counter the products that NVIDIA has, as AGP is still a very important market. While ATI was the first to have quite a few native offerings for the PCI-E market, they have let the low end and midrange AGP market stagnate. Hopefully this will change soon, as many users (especially in the retail market) are not yet ready to make the jump to PCI-E. The future for ATI appears to be a bit more well known as compared to NVIDIA. The Xenon chip (X-box 2) is the R500, which appears to be a totally new architecture for ATI. It will definitely support SM 3.0+ rendering, and rumors have it that it supports unified vertex and pixel shaders, as well as encompassing many extra features that Microsoft wanted implemented in the X-box 2. This core is based on TSMC’s 90 nm Low-K process, and this chip has been seen in the recently released X-box 2 developer boxes. No one from the press has had a chance to look at this developer box, and apparently the developers that have received this technology have been keeping it under strict lock and key. Few real details are known about this product, but at E3 Microsoft is supposed to release more specifications and some possible working prototypes (or at least showing off one of the developer boxes running some demos). The R520, aka “Fudo”, is the next big jump for ATI on the desktop. While it will share some of the technology in the R500, it will still exhibit the traditional separate pixel and vertex pipelines. This appears to be a very massive chip, sporting well over 300 million transistors. It will be ATI’s first SM 3.0 part for the PC, and it is rumored to have at least 24 pixel pipelines and up to 10 vertex shaders. This will also be made on TSMC’s 90 nm process. This will be one pixel shading monster. Many expect that with the release of the R520, we will finally see developers work on solid SM 3.0 titles. While there are currently a few games that support SM 3.0 rendering, there has not been a mass migration to the standard due to only having NVIDIA support it. Also, with the overall horsepower that the R520 will bring to the table, we will see a greater use of SM 3.0 functions that previously were too performance intensive to implement (loops and branches, though major functions of the SM 3.0 specification, were rarely implemented in any SM 3.0 title due to the NV4x architecture’s overall SM 3.0 performance).Due to the complexity of this part, as well as TSMC’s relative newness to the 90 nm process, I do not expect to see large numbers of these products hit the shelves initially. The products that do will most likely be well over $600. Early summer is when most agree that we will see this part, and it will probably only come out in PCI-E form with 512 MB of GDDR-3. This product will be rare as hen’s teeth until about 5 months after its release. At that point, we will see the introduction of other R5x0 parts that will address the high end and midrange markets. Eventually ATI will transition their entire lineup to R5x0 parts, and will support SM 3.0 from top to bottom. From rumblings about the web, the R520 taped out in February, so ATI is very close to getting their first samples back (if they haven’t already). Modifications will be made, as a 300 million transistor chip will show issues at first (no matter who designs it). But with many of the new tools available to the IHV’s, the process is not nearly as painful as one would imagine. There is no word if Intrinsity’s Fast-14 technology was used to help design the R5x0 series of chips. ATI is also developing a technology to compete with NVIDIA’s SLI. From the very beginning of the R3x0 series, ATI’s chips have supported a functionality called Supertiling. This functionality allows the chip to break up a rendered scene into tiles, and when combined with multiple chips, each chip takes alternating tiles to render. This more evenly balances out rendering loads as compared to NVIDIA’s SLI which uses AFR (Alternate Frame Rendering) and SFR (Split Frame Rendering). This should be a very effective and efficient setup, but so far no product has been shown. So while the technology looks impressive, we haven’t seen an implementation as of yet. By the time ATI does release its competing product, NVIDIA will have had SLI on the market for around 8 months.
If you have found this article interesting or a great help, please donate to this site.
Copyright 1999-2005 PenStar Systems, LLC. |
|||