The nVidia GeForce GTX 580 takes the crown of "World’s Fastest Graphics Card", sliding ahead of tough competition.

nVidia and AMD's GPU arms race continues unabated, and the "world's fastest" moniker hangs in the balance. The latest salvo comes in the form of nVidia's latest flagship graphics card, the nVidia GeForce GTX 580. Priced at £400, the card sits right at the upper echelons of the graphics card market, aimed squarely at the enthusiasts who will suffer nothing but the best.

And they'll find speed in spades here: the nVidia GeForce GTX 580 is the fastest single-GPU graphics card on the market today, sailing through our battery of tests with aplomb.

nVidia GeForce GTX 580: A Tale of Two Cards

A scant six months ago nVidia released its then flagship graphics card, the nVidia GeForce GTX 480. Built on nVidia's fledgling Fermi architecture, it was their first DirectX 11 graphics card. And it was quite a beast: a confluence of blistering performance and temperatures that generally topped the performance charts, but ultimately stumbled in light of AMD's well-established DirectX 11 competition.

With the nVidia GeForce GTX 580, nVidia quite literally went back to the drawing board. The GPU inside the GTX 480 was re-tooled from the transistor level, as nVidia worked to mitigate some of the power and temperature issues that plagued the card. As a result, the GTX 580 is quieter, cooler, and relatively more power efficient than its predecessor.

nVidia GeForce GTX 580: Putting Fermi to the Test

We pitted the nVidia GeForce GTX 580 against its predecessor, the GTX 480, and AMD's ATI Radeon HD 5870 and AMD ATI Radeon HD 5970. A note about the Radeon HD 5970: while it is a single graphics card, it's actually a pair of GPUs, sandwiched together. We've chosen to include it as it represents the absolute best that AMD has to offer.

Synthetic benchmarks aren't necessarily a barometer of real-world performance, but they're handy for getting a solid idea of how cards stack up. This is particularly true for the Unigine Heaven benchmark. This strenuous DirectX 11 test is laden with geometry-heavy scenes and tessellation - the Fermi architecture's bread and butter. The GTX 580 takes a decisive lead, ousting even the HD 5970's ample muscle.

While the Radeon HD 5970 clambers back into the lead on the 3DMark Vantage benchmark, it isn't ahead by very much - 7 percent, on the Extreme setting. That's not a bad showing from the GTX 580's single-GPU.

Our gaming tests saw a bit more of the same. The dual-GPU Radeon HD 5970 maintains a near constant lead, with the nVidia GeForce GTX 580 nipping at its heels. The Dirt 2 Demo's results were the outlier here, with nVidia's GTX 580 pulling ahead by as much as 80% over the Radeon HD 5870. It maintained a steady lead over the Radeon HD 5970 by up to 40%.

nVidia GeForce GTX 580: With Great Power, Comes Great Efficiency

The GTX 480s chief faults lay in its unwieldy power demands, and nVidia has gone a long way to rectifying that issue. AMD's Radeon HD 5870 is (unsurprisingly) the least demanding of the cards we looked at, but the nVidia GeForce GTX 580 managed to shave 20W off of its predecessor's idle and full load power ratings.

The improvement shows once we take performance per watt into account. The Radeon HD 5970 takes the lead as the most efficient card, but the nVidia GeForce GTX 580 makes dramatic strides over its predecessor, overtaking even the venerable Radeon HD 5870.

See also:  Group test: what's the best graphics card?

NEXT: our expert verdict >>

See also: Group test: what's the best budget graphics card?

nVidia GeForce GTX 580: Specs

  • CUDA Cores: 512
  • Graphics Clock (MHz): 772MHz
  • Processor Clock (MHz): 1544MHz
  • Texture Fill Rate (billion/sec): 49.4
  • Memory Clock (MHz): 2004
  • Standard Memory Config: 1536MB GDDR5
  • Memory Interface Width: 384-bit
  • Memory Bandwidth: (GB/sec) 192.4
  • nVidia SLI-ready, 3-way, nVidia 3D Vision Ready
  • nVidia 3D Vision Surround Ready
  • nVidia PureVideo Technology, HD
  • nVidia PhysX-ready
  • nVidia CUDA Technology
  • Microsoft DirectX 11
  • OpenGL 4.1
  • Bus Support: PCI-E 2.0 x 16
  • Certified for Windows 7
  • CUDA Cores: 512
  • Graphics Clock (MHz): 772MHz
  • Processor Clock (MHz): 1544MHz
  • Texture Fill Rate (billion/sec): 49.4
  • Memory Clock (MHz): 2004
  • Standard Memory Config: 1536MB GDDR5
  • Memory Interface Width: 384-bit
  • Memory Bandwidth: (GB/sec) 192.4
  • nVidia SLI-ready, 3-way, nVidia 3D Vision Ready
  • nVidia 3D Vision Surround Ready
  • nVidia PureVideo Technology, HD
  • nVidia PhysX-ready
  • nVidia CUDA Technology
  • Microsoft DirectX 11
  • OpenGL 4.1
  • Bus Support: PCI-E 2.0 x 16
  • Certified for Windows 7

OUR VERDICT

nVidia was late to the DirectX 11 game, and the GTX 480 ultimately failed to make much of a splash. But an extra six months in the proverbial tool shed has resulted in quite a showpiece. The nVidia GeForce GTX 580 is everything the GTX 480 should've been, but it's still saddled with some of the same problems. There's still no answer for AMD's Eyefinity technology - if you want to run three (or more) monitors, you'll still need to pick up a second GTX 580. The GTX 580 also lacks a DisplayPort connector, offering up a pair of DVI connectors and a mini-HDMI port instead - AMD's cards offer the full gamut. DisplayPort concerns aside, nVidia has managed an impressive feat. An unwieldy titan was reforged, delivering a component that outpaces the competition, and earns the vaunted "world's fastest" moniker.

Find the best price