Print
Every GPU generation has its flagship videocards: the ones with the top-of-the-line GPU with all cores enabled, loaded for bear. In this generation, those cards are Nvidia’s GTX 680 (with a full GK104 GPU inside) and AMD’s Radeon HD 7970 (with a full Tahiti GPU). These cards are monstrously fast, but they’re also expensive and tricky to manufacture. Not all parts come off the line fully functional. So a few months after each flagship GPU launch, the vendors come out with a slightly stripped-down version that uses binned top-end GPUs with a few parts disabled, or lower clock speeds. AMD’s Radeon HD 7950, for example, uses the same GPU as the 7970, but with 28 GCN units instead of 32, and with an 800MHz reference clock instead of 925MHz. The cheaper, lower-powered video cards appeal both to gamers with shallower pockets and also to vendors, who clock those stripped-down, less expensive GPUs right back up to within spitting distance of their full-powered peers. Thus we arrive at the Asus GeForce GTX 670 DirectCU II TOP, a factory-overclocked GTX 670 with a custom cooling solution.
Nvidia’s reference GeForce GTX 670 uses the Kepler GK104 GPU with one SMX disabled, reducing the total number of CUDA cores to 1,344 from 1,536 and the number of texture units to 112 from 128, and lowering the base and boost clock speeds to 915MHz and 980MHz, respectively.
The DirectCU II cooler’s three direct-contact heat pipes keep the GPU cool.
The reference GTX 670 is a dual-slot card just over 9.5 inches long, although the PCB is less than 7 inches long—the rest is fan shroud. Asus, however, uses an entirely custom PCB and fan shroud. The GeForce GTX 670 DirectCU II TOP, like other DirectCU II cards, uses direct-contact copper heat pipes that feed into a stack of fins under two 8cm fans, all in a spiffy black-and-red shroud. Like the reference card, the Asus board takes two 6-pin PCIe power adapters and has 2GB of GDDR5 frame buffer at 3,004MHz on a 256-bit bus.
To ensure the best overclocks, Asus cherry-picks the best GTX 670 GPUs for the DirectCU II TOP card. The whole assemblage makes the Asus card much heavier and longer than the reference design—the PCB alone is over 9.5 inches long, and the fan shroud makes the card 10.5 inches long, nearly as long as a dual-GPU card.
The extra size and weight pay off, though. The larger fans and greater heat dissipation area mean Asus can push the hand-picked GPU to dizzying speeds. Where the reference GTX 670 has a base clock of 915MHz and a boost clock of 980MHz, Asus’s DirectCU II TOP version has a base clock of 1,058MHz (the same speed as the reference GTX 680’s boost clock) and a boost clock of 1,137MHz. Asus includes its GPU Tweak software to allow further user overclocking.
We tested the $430 Asus GTX 670 against a $400 reference-design GTX 670, a reference GTX 680 ($500), and a factory-overclocked Sapphire Radeon HD 7950 ($400, reviewed May 2012), its AMD equivalent. As you can see in the benchmark chart, the results are impressive.
At 2560x1600 with all settings maxed and 4x MSAA, the stock GTX 670 outpaces a factory-overclocked Radeon HD 7950 in all but two benchmarks, and gives playable (30-plus) frame rates at these settings in every game except Metro 2033 and Shogun 2. But the impressive performance doesn’t stop there. Thanks to those extraordinary factory overclocks, the Asus GTX 670 DirectCU II TOP actually outperforms a reference GTX 680 across the board while being substantially cooler and quieter.
The Asus GTX 670 DirectCU II TOP is large and heavy, but it’s quiet, and thanks to the cherry-picked and overclocked processor, it’s cooler and faster and has a lower TDP than a stock-clocked GTX 680. If you want top-end performance for less than top-end price, this is your card.
Faster than a GTX 680;cool and quiet; good-looking.
Longer and heavier than the reference card; doesn’t ship with any cables or connectors.
| Asus GTX 670 DirectCU II TOP | Nvidia GTX 670 Reference | EVGA GTX 680 | Sapphire Radeon HD 7950 OC | |
|---|---|---|---|---|
| Price | $430 | $400 | $500 | $400 |
| 3DMark 11 Performance | P9,664 | P8,706 | P9,555 | P7,683 |
| 3DMark 11 Extreme | X3,300 | X2,957 | X3,249 | X2,562 |
| 3DMark Vantage Performance | P34,700 | P32,568 | P34,339 | P31,752 |
| Unigine Heaven 2.5 (fps) | 30.7 | 28.6 | 31.2 | 26.7 |
| Shogun 2 (fps) | 19.4 | 18.5 | 18.7 | 25.8 |
| Far Cry 2 / Long (fps) | 110.5 | 103.4 | 107.3 | 89.6 |
| Dirt 3 (fps) | 73.3 | 65.5 | 73 | 53.8 |
| HAWX 2 (fps) | 134 | 123 | 131 | 106 |
| Metro 2033 (fps) | 16.7 | 16 | 16.3 | 18.7 |
| STALKER: CoP SunShade(fps) | 35.4 | 32.4 | 34.3 | 34.1 |
| Just Cause 2 (fps) | 57.4 | 52.7 | 54.7 | 45.3 |
| Batman: Arkham City (fps) | 60 | 57 | 58 | 54 |
| Base Clock | 1,058MHz | 915MHz | 1,006MHz | 900MHz |
| Boost Clock | 1,137MHz | 980MHz | 1,058MHz | N/A |
| Memory Clock | 1,502MHz | 1,502MHz | 1,502MHz | 1,250MH |
Best scores bolded. Our GPU test bed consists of a stock-clocked Intel Core i7-396X on an Asus P9X79 Deluxe board with 16GB DDR3/1600,a 256GB Samsung 830 Series boot SSD,and a 1050W Thermaltake Toughpower Grand PSU, in a Cosmos II chassis.All tests performed at 2560x1600 with all settings maxed and 4x MSAA except where noted.Power use measured with a Watts Up Pro.
