Articles: Graphics

Bookmark and Share


Table of Contents

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]

As you know from our news and reviews, AMD’s superiority in the sector of single-processor gaming graphics cards has been challenged by Nvidia releasing its GeForce GTX 480, yet the AMD Radeon HD 5970 is still unrivalled as the best dual-chip solution, carrying two RV870 Cypress processors on board. The GeForce GTX 480 fails to compete, largely due to its lower texture-mapping performance, but AMD’s solution is superior to its opponent in the rest of technical specifications, too. However, there is not doubt Nvidia wants to challenge AMD here as well, so the development and release of a dual-processor card with Fermi architecture is just a matter of time and technical ability.

Nvidia had serious technological obstacles until recently because the hypothetic dual-processor GeForce GTX 490/495 would turn out to be too hot and power-consuming even if based on a GeForce GTX 470 SLI configuration, let alone GTX 480 SLI. So, although a pair of GeForce GTX 470 cards working in SLI mode could deliver the desired performance, the GeForce GTX 490/495 was still not feasible. But now that Nvidia has its new, simpler and more economical GF104 core, the situation is completely different. This rather inexpensive GPU is a much better building block for a dual-chip gaming graphics card than the monstrous GF100. Our tests showed it to be competitive to the GeForce GTX 470 at increased clock rates, and a pair of GF104 chips in a SLI tandem looks impressive even at the reference frequencies:

Click to enlarge

The hypothetic GeForce GTX 490/495 is going to be slower if based on GF104 cores than if it were equivalent to the GeForce GTX 470 SLI configuration. On the other hand, it will be faster than the single GeForce GTX 480 while costing the same or less money and will also be good in comparison with the Radeon HD 5970, being only inferior to the latter in terms of texture-mapping performance, which is the common problem of all Fermi-based solutions. When developing its Fermi architecture, Nvidia enhanced the GPU’s geometry-processing resources at the expense of texture-mapping ones. Practice suggests that this doesn’t prevent GF100- and GF104-based graphics cards from doing well in today’s games and their advanced tessellation units are going to be most helpful in the next generation of games that will use a lot of tessellation.

A dual-GF104 card may make the GeForce GTX 480 unpopular if comes at a comparable price but it’s too early yet to make any predictions about that, especially as GF100-based solutions have a reserve of untapped resources and can get faster if all the 512 stream processors are enabled in them.

It is certain, though, that this variant is going to be cheaper than a dual-GF100 solution and AMD will have to cut its prices because the Radeon HD 5970 is still hard to find at its recommended price of $599. Most offers are in the range of $630-650, but sometimes this card may be priced at $700 and more. Of course, a $500 GeForce GTX 490/495 cannot bring in much profit, just as any other premium-class graphics card for that matter, but its release may boost Nvidia’s reputation and popularity among demanding gamers who don’t want to bother about building discrete SLI or CrossFire tandems. So, we can expect a graphics card with two GF104 cores working in their full configuration with 384 stream processors and 64 TMUs; it may be called GeForce GTX 495. The name of GeForce GTX 490 may be given to the same solution based on two cut-down GF104 cores.

It is the latter variant that we are going to test today using a couple of GeForce GTX 460 1GB cards in SLI mode. One of the components of this tandem is a Zotac GeForce GTX 460 1GB (ZT-40402-10P). Let’s have a closer look at it now.

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 ]


Comments currently: 18
Discussion started: 08/21/10 01:25:34 AM
Latest comment: 10/03/16 04:56:02 AM

View comments

Add your Comment