Nvidia Debuts the Very Efficient GeForce GTX 960

Today, Nvidia is launching its GeForce GTX 960 graphics card, which is a mid-tier card aimed at meeting the needs of value-oriented users.

This graphics card, like the GTX 970, GTX 980, and GTX 750 (Ti) that came before it, is based on the relatively young Maxwell architecture. It’s built on an all-new GM206 GPU, which is made using a 28-nanometer lithographic process. Inside this GPU you’ll find just what you’d find in a GTX 980, but halved – 1024 CUDA cores, 64 TMUs, 32 ROPs, and a 128-bit memory interface that addresses 2 GB of GDDR5 memory that’s placed elsewhere on the PCB.

This memory runs at an effective speed of 7010 MHz, meaning that it pumps out a bandwidth of about 112.2 GB/s. The GM206 GPU runs at 1126 MHz as a base frequency, but will boost up to 1178 MHz using Nvidia GPU Boost 2.0 when the thermal headroom is available. The 128-bit memory bus appears to be quite narrow, but the GTX 970 and GTX 980 have already proven that the Maxwell architecture is able to do more with less resources, so let’s hope the pattern continues here.

The graphics card has a TDP of just 120 W, which enables it to pull all the power it needs from just the PCI-Express slot and a single 6-pin PCI-Express power connector. The efficient architecture allows Nvidia’s board partners to opt for switching off the GPU’s fans when it is running idle or under a low load. Such a technology isn’t only helpful to reduce idle noise levels, but also to reduce dust buildup inside the card. According to Nvidia, on launch day Asus, Gainward, and Gigabyte are offering this feature, although we expect a handful of other manufacturers, including Evga, to offer it as well.

Nvidia also claims that the card is built for overclocking, boasting that you should be able to overclock the cards to speeds approaching or beyond 1500 MHz without too much effort. This is in part thanks to its incredibly efficient design. Nvidia managed to achieve 1450 MHz on its reference card without modifying the fan or voltage settings, so we expect that aftermarket cards will be able to do even better.

Nvidia’s aim for this card is to address the “sweet spot” in the GPU market, where for an acceptable price you can get a whole heap of performance. The card is intended to succeed the GTX 760, and it should provide impressive Full HD performance, with acceptable 1440p performance, at an appealing price point.

Naturally, it will come with all the features found on the other Maxwell based cards, including support for Nvidia G-Sync, DirectX 12, VR Direct Support, Dynamic Super Resolution and the new MFAA, among many more.

Various models should be hitting shelves sooner or later, with MSRP pricing set at $200. Expect pricing for factory-overclocked models to be not far above that, with premiums of around $10 in many cases.