Seven GeForce GTX 660 Ti Cards: Exploring Memory Bandwidth

Benchmark Results: 3D Performance

GeForce GTX 660 Ti Review: Nvidia's Trickle-Down Keplernomics included a ton of detailed performance benchmarks. Today, we're focusing on comparing the cards themselves. The largest differences between them are their factory GPU and memory overclocks. When it comes to quantifying the effects of those subtle changes, 3DMark 11 is pretty effective at normalizing variables and giving us good results.

Gigabyte's GeForce GTX 660 Ti Windforce 2X OC Edition at 1033 MHz and MSI's N660 Ti PE 2GD5/OC at 1019 MHz clearly pull ahead of their competition at 3DMark's Entry and Performance presets. But what about Zotac’s GeForce GTX 660 Ti AMP! Edition at 1033 MHz, the card with the highest memory clock rate? And why do the Palit and Gainward cards, which are almost identical, demonstrate different results?

The answer is both simple and surprising. But first, the benchmark:

The GPU Boost clock rates of the cards behave differently under load. The Asus and Zotac boards don’t increase at all. The Gainward card's GPU is apparently more scalable than Palit's, granting it somewhat better performance in spite of the almost identical technical specifications and build. This is interesting in light of the fact that Gainward is Palit’s flagship brand, and the Gainward GTX 660 Ti Phantom is more expensive than the Palit GTX 660 Ti Jetstream.

The Entry and Performance presets in 3DMark 11 aren't very memory bandwidth-dependent, but they do push each GPU to its limits. The following GPGPU-oriented benchmarks apply to 3DMark 11’s Extreme preset as well.

The two cards based on Nvidia's GeForce GTX 670 PCB clearly win under the synthetic's Extreme preset. This shows that a factory-overclocked GPU isn't everything; memory interface bandwidth matters as well.

This is a phenomenon we expected to see when Nvidia first announced GPU Boost, and we're starting to see it mess with the value of add-in board partner offerings. It makes less sense to pay more for a higher base clock when the effects of GPU Boost affect each piece of silicon differently (and there's nothing you can do about it).

DirectX 11 vs. DirectX 9 and Resolution

We benchmarked all seven cards using two game presets and two resolutions. Under DirectX 9, Gigabyte has the fastest GeForce GTX 660 Ti, followed closely by Zotac. On paper, this should have been the other way around, given Zotac's 100 MHz-higher memory clock rate. Our best guess is that the GPU on Zotac's card simply wasn't as good. Zotac does manage to outmaneuver the Gigabyte board under DirectX 11, though.

The other GeForce GTX 660 Tis perform on par with what we were expecting, based on their specs.

Although the GeForce GTX 670s behave as we'd expect, based on previous benchmarks, the GeForce GTX 660 Ti cards are slower under DirectX 9 than 11. Moreover, the clock rates of the most aggressively overclocked 660 Tis weren't high enough to contend with the slowest GeForce GTX 670. Of course, this could be by Nvidia's design after watching GTX 670s overtake pricier 680s.