So good it's Classified
NVIDIA's GeForce GTX 680 graphics card became an instant hit with deep-pocketed enthusiasts as soon as it was launched back in March. Built as a gamers' card and packing in the enviable qualities of performance, features and energy efficiency into one tidy package, it remains to this day our choice for a high-end build.
AMD's hit back with the Radeon HD 7970 GHz Edition and has encouraged add-in board partner Sapphire to take the fight straight to NVIDIA with the HD 7970 TOXIC 6GB, but the green team's own partners have parried the AMD thrust with pre-overclocked GTX 680s from the likes of KFA², Gigabyte and ASUS, amongst others.
Now muscling into this single-GPU, £500 territory is EVGA. Its best air-cooled GTX 680 is the Classified 4GB, designed from the grounds-up to beat the NVIDIA reference card in every way.
In a category where triple-slot-taking cards and outlandish cooling isn't rare, EVGA's Classified still manages to turn heads. The card is significantly larger than any other GTX 680 we've seen before - it makes the PCI bracket look small in comparison - and the only meaningful contribution NVIDIA has made to it is to supply the underlying GPU; everything else is EVGA's work.
Measuring about an inch longer and taller than a standard GTX 680, EVGA doesn't bother with the popular open-sided, triple-fan coolers we've seen on many competing cards. Rather, an 8cm radial fan - yup, it really is 8cm - pushes air over an enclosed area and out of the back, and the extra-large board and shroud is needed to house the fan without encumbrance.
Extra airflow is required because EVGA clocks this beastie in at 1,111MHz core, with GPU Boost up to 1,176MHz, compared with the 1,006/1,058MHz clocking of a standard GTX 680. The GPU Boost speed, which is the average frequency the core will run at most of the time, is competitive against other premium GTX 680s.
Turning it around confirms that, while obviously rather large, it still retains the two-slot profile of the reference card; EVGA's gone longer and wider rather than taller. Two 8-pin power connectors are de rigueur for any aftermarket top-line card, enabling up to 375W to course through its silicon veins. Sitting next to the connectors is the header for what the firm calls EVBot, which is a device that's able to properly over-volt the card. Bear in mind that it's an optional extra, retailing for around £50, and more on this later.
Hard to see but present nonetheless, to the right, is a three-position BIOS switch. Before enthusiasts begin salivating with untapped frequency promise, there's no obvious difference between the three states. The firm says that the latter two positions enable 'extreme overclocking mode,' designed for water- and LN2-cooling, where NVIDIA's overclocking-orientated Power Target control is disabled. This statement is important, readers, as it means the base PCB is built for enhanced cooling in mind, though, interestingly, the GTX 680 Hydro Copper uses a smaller PCB.
Last but not in a line of EVGA-specific goodies, also carried on over from the GTX 580 Classified, are voltage-monitoring points, and the firm is to release a small adapter for easy attachment of voltage probes.
If you paid attention to the first part of this review you'll know the card ships with 4GB (4,096MB) of memory, or double that of a standard GTX 680. We're not aware of any shipping 4Gbit devices clocked in at 6,000MHz-plus, so EVGA needs to use 16 2Gbit chips - eight on each side - to constitute the 4GB buffer. The eight Hynix chips on the backside are identical to those found on the reference card. Speeds, too, are the same, as EVGA keeps memory chugging along at an effective 6,008MHz - we'd really have liked to see it higher on this best-in-class card, though, perhaps, the lower-than-expected clocks are an indication that a double-sized buffer doesn't scale so well on this GPU.