In line with my latest news item, I encourage those of you out there who are interested to check out this excellent article on NVIDIA's new "Fermi" GPU architecture by AnandTech:
http://www.anandtech.com/video/showdoc.aspx?i=3721&p=1
It sheds a lot of light on what NVIDIA is trying to do to compete with ATI right now. They seem to be banking heavily on tessellation performance, but whether or not game developers will actually take full advantage of it remains to be seen. More articles (including a great article that compares ATI and NVIDIA architectures) as well as my take past the break.
Lets look at the facts: the ATI 5xxx series (supporting DX11) has been out for the better part of a year now. The ATI HD 5870, ATI's flagship single-core CPU, has been out for 6 months now. The dual-GPU version, the HD 5970, has been out since November. Just now NVIDIA is releasing their answer to the 5870 with the single-GPU GeForce GTX 480, boasting 5-10% improved performance over the ATI card, at a price point about $100 higher and with a TDP almost equal to ATI's dual-GPU card. Lets also point out that this will be the first DX11 card that NVIDIA brings to market.
As expected, tessellation performance is significantly better than ATI can manage, but it's going to be a moot point unless the game developers really begin to use the technology (included in DX 11) in a way that will actually give NVIDA an edge. ATI cards can perform tessellation (as mentioned, it's mandatory to be supported in DX 11) but developers are likely to only include textures in their games at a low enough resolution that any edge NVIDIA has in tessellation performance will be practically negligible.
NVIDIA would only truely shine if very high-resolution textures were dropped into DX11 games (of which only 6 exist currently, but of course, many more are on the way). The only issue is that high-resolution texture files eat up huge amounts of space, and software developers are already struggling to cram their games onto 1 or 2 DVDs. NVIDIA will have to make a very good case for this to get developers on board.
If you're looking to learn more about the GPU architecture of the 5870 (to compare to the GTX 480 / Fermi article above), check out this article: http://www.anandtech.com/video/showdoc.aspx?i=3643
Lastly, for a good comparison of how the ATI and NVIDIA architectures fundamentally differ, check out this article (I've started you at page 3): http://www.anandtech.com/video/showdoc.aspx?i=3341&p=3
The article is 2 years old but the core architectures used by ATI remain fundamentally the same (though for NVIDIA, the Fermi architecture changes several key things), and it's a great reference on how both companies tackle the same problems in different ways.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment