PROS / Breathtaking visual effects rendering thanks to Fermi and DirectX 11 technology.
CONS / Runs hotter and louder than the competition.
VERDICT / This is a great card with a decent air flow and cooling in your box.
Editor’s Note: This product has been removed from viewing as part of this site because it has been replaced by a new generation of graphics card. You can still read our original review below, but TopTenREVIEWS is no longer updating this product’s information.
Nvidia has finally brought out its DirectX 11 series of graphics cards--an entire refresh cycle late--the GTX 400 series. Better late than never, we say. The flagship of the lineup is the Nvidia GeForce GTX 480. Although the newest Radeon model has the current claim on the most powerful graphics card, it is a dual-GPU card, which puts it in a realm of its own. Nvidia's previous flagship was also a dual-GPU card but in no way can it stand up to 5000 series. With the current series they have gone back to the single GPU for now, though that could change in the future. Why Nvidia has chosen not to go toe-to-toe with AMD's newest model is, for the time being, a mystery. We rated this card so high because of its superior tessellation abilities which greatly improve many aspects of this amazing card. It did have some heating and noise issues when it first came out, which is why it was quickly replaced by the GTX 500 series.
As far as single-GPU graphics cards go, the GTX 480 is the most powerful. Benchmarks on many graphics-intense games have shown it to have an overall lead against other single-GPU cards. The dual-GPU ATI Radeon card is still unrivaled, but knowing the competition between the two companies, Nvidia will come out with something soon enough.
The GTX 480 has an overall average set of GPU stats, when compared to other high-end graphics cards. The exception here is its impressive number of transistors per die. The highest end Radeon card has 334 million transistors on a 2154 mm2 die, whereas the GTX 480 has 3 billion transistors on a 529 mm2 die. That is a huge difference.
Compared to single GPUs, video memory is one place where the GTX 400 shines. It has the highest clocked memory, 1848MHz, as well as the most memory per GPU. Most cards sit at or around 1GB per GPU but the GTX 480 gives us an extra half gigabyte on top of that for a total of 1536MB.
It is of course GDDR5 memory, which is the high-end standard for now. The bus width is larger than most, at 384-bit which helps its overall performance greatly. The 177.4 bandwidth is far greater than most other graphics cards but is technically out-done by dual-GPU cards that have two separate sets of memory at their disposal.
Much like the other high end cards for Nvidia the 480 has 2 dual channel DVI slots and 1 Mini-HDMI 1.3 slot which is the standard for graphics cards today. You should have no problem finding these type of connections for any system on the market today. We would like to see a few other options for this card but in the long run having at least two will make it possible for anyone to use this card.
NVIDIA finally gives us DirectX 11 on their cards. They cut it a little close but still managed to jump in before DirectX 11 becomes standard in games and video rendering. The new Fermi architecture is based on CUDA, but also has the integration of DirectX elements. DirectX 11 supports things like greater tessellation, DirectCompute (which allows the GPU to act as a second CPU) and superior multi-threading. DirectX 11 is a huge step for GPUs, and it's the future of graphics rendering.
Although the Nvidia GeForce GTX 480 is great for gaming, the focus seems to be super-computing for developers. Nvidia is known for its CUDA technology, which has been the company's GPU computing architecture for quite a while now. With the release of the GeForce GTX 400 series cards came the upgrade version code named "Fermi". Fermi is the biggest reason for the long wait for this card line. It takes everything that CUDA was, and added many of the developments that DirectX demands. The combination of DirectX and Fermi makes for some exciting potential for developers, which will hopefully convert to some intense gaming experiences soon enough.
The Fermi technology could probably be best described in Nvidia's own words. They say, "Fermi makes GPU and CPU co-processing pervasive by addressing the full-spectrum of computing applications. Designed for C++ and available with a Visual Studio development environment, it makes parallel programming easier and accelerates performance on a wider array of applications than ever before – including dramatic performance acceleration in ray tracing, physics, finite element analysis, high-precision scientific computing, sparse linear algebra, sorting, and search algorithms."
With the big focus on rendering technology, as long as a large group of developers jump on board, Nvidia could open some serious doors for future video development, which could land them in a very good position.
The biggest push they're making is for tessellation to become a major implementation. Tessellation, when it comes to computer graphics, is used to manage and divide polygons so they can be rendered faster. The more "tessellated" they get, the more detailed and realistic they can become. The example most commonly used is with hair. Hair is quite difficult to render realistically because there are thousands of fine strands that have to act independently. For complete realism it needs to flow with the physics of natural-looking movement. That can be a person turning their head or ducking. In cases like this, developers have to think, "What should the hair do to look natural?" Tessellation implements the natural effects of nature to give us realism. That includes things like the wind blowing the hair around. In older games, hair is usually just a couple of chunks, and the movements are not graceful at all.
In-game content is limited generally to the cinematic "pre" and "post" play scenes, but with the astounding tessellation performance that the Fermi provides, you may start seeing a lot more of it in actual game play. Although it isn't a new concept, tessellation has yet to be integrated on a wide scale into anything but movies and those cinematic scenes throughout games. We've all seen it before. Good examples of realistic tessellation are Manfred the mammoth in the movie Ice Age and also Sullivan in Monsters Inc. Watch one of those movies again and pay close attention to their fur and its movement. In-game content may be limited now, but with the astounding tessellation performance that the Fermi technology provides, hopefully we'll start seeing a lot more of it. With the implementation of intense tessellation, you may find yourself wondering if it's real or rendered. And we look forward to that.
One of the really cooled features we liked about this card was the fact that it used heat pipes which are not seem quite as much on recent cards. These are simply used to keep the card cooler when running high intense games; the problem is that somewhere along the line the heating was a big issue for this card. It is the same size as its older brother at 4.4 inches high and 10.5 inches in length making it quite a large card.
In short, the Nvidia GeForce GTX 480 is an amazing single-GPU card. SLI dual cards would compete with the overall performance and FPS (frames per second) of the dual-GPU 5970 that ATI has out, but not in its singular state. Two cards are probably too expensive for the PC of your average gamer, but it's definitely a great option for the enthusiasts and developers. On the other hand, if your main concern isn't those extra FPS and you're more interested in the Fermi architecture that the GTX 400 series has, than the less expensive option is a single 480 rather than ATI's 5970.