Asus GTX 470 review

With its Radeon HD 5000-series graphics cards released six months ago, ATI would seem to have the graphics card market sewn up. Nvidia hasn’t been sitting idle though, and has been working on its own DirectX 11 cards, code-named Fermi. Here we’re looking at the cheaper GeForce GTX 470. Its price to puts it into competition with ATI’s high-end Radeon HD 5870 and HD 5850 (see below) graphics cards. A more powerful GTX 480 card is also available, but with prices starting around £480 inc VAT, it’s beyond the budget of the vast majority of gamers

As you’d expect from a modern graphics card, the GTX 470 is DirectX 11-compliant. This evolution of DirectX 10 introduces some big improvements. The main ones are tessellation, DirectCompute and multithreaded rendering.

Tessellation allows the complexity of an object to increase as the player moves towards it. It means that less computing power is required for objects in the distance, but boosts the realism of a game by making objects closer to the camera more detailed. The way tessellation works means that it’s easier for programmers to deal with the transition to more complicated objects. ATI also has tessellation hardware, but nVidia’s card architecture is better set up for this kind of work.

DirectCompute is designed to allow the GPU to be used for other jobs, such as video encoding. GPU’s are highly efficient at dealing with parallel computing tasks, and can massively reduce computation times in certain cases – like video encoding. It’s something that Nvidia has been pushing with its CUDA branding, and you’ll find numerous big-name software vendors, like Adobe, supporting the standard.

Finally, multithreaded rendering means that multi-core processors can be used to the full by queuing up work for the graphics card ahead of time, avoiding the limitation of traditional games where only a single processor core could be used. This should help boost efficiency, but you’re unlikely to see that transferred into real frame rate improvements in games, as developers will use the improved efficiency to introduce new special effects.

It’s hard to directly compare the graphics architectures used by ATI and Nvidia. Differences in the number of processors and MHz ratings aren’t useful comparisons, as it’s the core architecture that makes the real difference. For those interested, the GTX 470 has four graphics processing clusters, 14 streaming multiprocessors, 448 CUDA cores, 56 texture units and 40 ROP units. The graphics clock runs at 607MHz and the CUDA cores at 1,215MHz. The 1,280MB of memory has a 320-bit memory interface with a data rate of 3,348MHz, giving a total memory bandwidth of 133.9GB/sec.

This dual-slot card requires two six-pin PCI-E power connectors. This may be the same as the HD 5850, but comparatively the GTX 470 is a bit of a power hog. It has a maximum board power of 215W, compared to the HD 5850’s 151W. Then there’s heat – while the 5850 runs relatively cool, the GTX 470 was incredibly hot to the touch.

It has three outputs: two dual-link DVI and one mini HDMI – although you can only run two displays at the same time (unlike the triple-display capable Eyefinity cards from ATI). Nvidia remains the only graphics card manufacturer with support for 3D monitors. With the new 3D Vision Surround you can run three displays side-by-side in full 3D; but remember that you’ll need two graphics cards running in SLI to support this many displays, making it a very expensive configuration.