High-end cards are all well and good for the privileged few, but the real money is in the mid-range. It’s an area in which Nvidia has enjoyed significant success, thanks to its GeForce GTX 460 and GTX 560 chipsets, and its latest effort, the GTX 660 Ti, aims to continue this dominance.
Nvidia hasn’t strayed from its successful Kepler architecture to make its latest card, with the GK104 core used in the GTX 690, 680 and 670 once again pressed into service. This time, it’s the lesser of those three cards, the GTX 670, that forms the blueprint of the GTX 660 Ti.
Very little of the GTX 670’s architecture has been changed to form the GTX 660 Ti. Both cards make do with seven of the eight Graphics Processing Clusters used in the GTX 680; this means that both include 1,344 stream processors – a small step down from the 1,536 utilised in Nvidia’s top-end chip. The differences between the two cards are minor: both include 112 texture units but, while the GTX 670 used 32 Render Output Units, the GTX 660 Ti has to make do with 24 – a change that could impact on the new card’s ability to cope with demanding anti-aliasing.
Both cards include a 915MHz core clock with a maximum boost clock of 980MHz, and the 294mm[sup]2[/sup] die still includes 3.5 billion transistors. There’s 2GB of GDDR5 RAM included in the GTX 660 Ti and, like the GTX 570, it’s clocked at 6,008MHz – the most significant difference is that the GTX 660 Ti utilises a 192-bit bus rather than the 256-bit interface of the GTX 670.
The tweaked specification results in reasonable theoretical throughput figures: the GTX 660 Ti’s total memory bandwidth of 144.2GB/sec is lower than the 192.2GB/sec of the GTX 670, but its texture filtering rate of 102.5GT/sec is equal to that of Nvidia’s higher-end card.
Those figures would have trounced its main rival, too, but AMD has been busy tweaking its Radeon HD 7950 in the run-up to the GTX 660 Ti’s launch; a new BIOS update sees its clock speed boosted from 800MHz to 925MHz. That bumps its theoretical throughput figures up to 240GB/sec and 103.6GT/sec respectively.
There was precious little to choose between AMD and Nvidia in our Very High quality Crysis test, with the GTX 660 Ti scoring 66fps and the AMD HD 7950 falling two frames behind. The gap widened in the same test at 2,560 x 1,440, with the GTX 660 Ti’s 48fps score leaping six frames ahead of its rival. The roles were reversed when we hooked up three screens – at 5,760 x 1,080, the recent BIOS update helped the Radeon to score 35fps compared to the GTX 660 Ti’s 26fps.
Core Specifications | |
---|---|
Graphics card interface | PCI Express 3.0 |
Cooling type | Active |
Graphics chipset | Nvidia GeForce GTX 660 Ti |
Core GPU frequency | 915MHz |
RAM capacity | 2.00GB |
Memory type | GDDR5 |
Standards and compatibility | |
DirectX version support | 11.0 |
Shader model support | 5.0 |
Connectors | |
DVI-I outputs | 2 |
DVI-D outputs | 0 |
VGA (D-SUB) outputs | 0 |
S-Video outputs | 0 |
HDMI outputs | 1 |
7-pin TV outputs | 0 |
Graphics card power connectors | 2 x 6-pin |
Benchmarks | |
3D performance (crysis) low settings | 187fps |
3D performance (crysis), medium settings | 128fps |
3D performance (crysis) high settings | 101fps |
Disclaimer: Some pages on this site may include an affiliate link. This does not effect our editorial in any way.