Not to be confused with our earlier review of the GTX 660 Ti, this is the slower Nvidia GTX 660 based on the less exciting Kepler GK106 die. Is it still any good, especially in the face of the aggressively priced AMD "Pitcairn" HD 7800 series?
This week, NVIDIA completes its top to bottom line up with the launch of the GK106 based GTX 650 Ti (not to be confused with the much, much slower GTX 650, difference is shit and chocolate really). Frankly this naming conundrum is bound to cause a lot of unnecessary confusion (some might call it a marketing masterstroke) among consumers about the product's actual capabilities and performance. Not that we haven't historically seen such behaviour before from both the green and red camps anyway…
The card that we are dealing with today is ASUS's GTX 660 (non-Ti) Direct CU II TOP edition, which debuted last month to plug the rather sizable hole in the mainstream segment that the more powerful GK104 based cards like the GTX 660Ti/670/680 were priced out of. The local ASUS office has always been very supportive of our publication and has once again provided us with an OC edition of virtually every GPU release this year to review.
As with all of ASUS's recent non-reference offerings, they come loaded with the DirectCU II cooler for silence and performance, high quality digital VRMs and the enthusiast oriented GPU Tweak software for overclocking.
The trademark I/O panel configuration from the rest of the higher end Kepler offerings also makes an appearance here, with a pair of DVI ports augmented by full-sized DisplayPort and HDMI options.
Although we have seen shorter PCB designs from the other AIB vendors, ASUS chose to extend the length of the PCB to 10-inches, which could deter some owners of space constrained HTPC cases.
Only a single SLI connector is provided, which means only up to 2-way multi-GPU action. Board power is supploed by a single 6-Pin PCIe connector (up to 150W of power draw from external + slot).
The GK106 die is made up of five SMX units, providing 960 CUDA Cores for pixel/vertex/geometry shading/PhysX calculations and 80 texture units. Like the 660 Ti (which btw has 1344 CUDA cores), it also comes with a 192-bit GDDR5 memory interface and features GPU Boost technology to throttle the effective clock speed according to TDP and other parameters.