GeForce GTX 470 SLI against Radeon HD 5970

@ 2010/06/22
If you keep track of what is going on in the graphics card industry, you should know that AMD has met but little competition since the release of the Radeon HD 5000 architecture. Even the single-processor Radeon HD 5870 could often beat the dual-processor GeForce GTX 295 whereas the dual-processor Radeon HD 5970, released somewhat later, left no chance to Nvidia products. Now that Nvidia has introduced graphics cards based on the Fermi (GF100) architecture, the situation begins to change. At least Nvidia has got solutions that are functionally as good as AMD’s ones.

However, our tests showed that even the flagship model of the new series, GeForce GTX 480, cannot beat the Radeon HD 5970 singlehandedly because the latter has got the combined power of two RV870 graphics cores. Such a feat would require more computing muscle. In other words, there is only one way for Nvidia to throw the long-time king of the 3D world off its throne. It is by releasing a dual-processor GF100-based graphics card. But it would be very difficult to build a card with two GF100 cores, highly sophisticated and power-consuming as they are, if both were to work in full configuration. As a matter of fact, even the GeForce GTX 295 was a pair of GeForce GTX 275, but not GTX 285, cards! ASUS did develop the latter version, but its Mars GTX 295 turned out a highly expensive product and was released as a limited edition only. It couldn’t change anything in the market situation.

If Nvidia indeed has a dual-processor GF100-based card on its plans (it might be called something like GeForce GTX 490 or 495), the specs of that solution would be identical to the specs of a GeForce GTX 470 SLI configuration. That is, the card would have 968 ALUs, 112 TMUs, and 80 RBEs. Considering the scalar Fermi architecture, this should suffice to compete with the Radeon HD 5970 even if the GPU clock rates are lower than those of the GeForce GTX 470.

No comments available.