Nvidia launches the GeForce GTX 980M and GTX 970M

On paper, the specs appear to be rather modest compared to the previous GeForce GTX 880M and 870M: Although the clock rates of the GTX 900M series are sightly higher, the activate shader units remain unchanged or even slightly lower in the case of the GTX 970M. Is this a bad omen for performance? No: The new GPUs should still benefit from their essentially revised architecture. For example, Nvidia restructured the streaming multi processors (SMM) in Maxwell to contain only 128 shader units and 8 TMUs (compared to 192 shader units and 16 TMUs in Kepler). Together with further optimizations, the GM204 should perform about 35 to 40 percent better than its Kepler GK104 predecessor with the same number of ALUs.

Changes in the memory interface have been primarily to increase energy efficiency. Nvidia increased neither the bus size nor the clock rate, but instead integrated larger L1 and L2 caches and better compression algorithms to increase effective bandwidth. The manufacturer promises double the performance of the GeForce GTX 680M in this regard. As both are processed in a 28 nm lithography, this is certainly impressive.

In principle, the Maxwell architecture has already been available for notebooks in the form of the GM107 or GeForce GTX 860M. However, the larger GM204 with its 5.2 billion transistors and 398 mm² die size is a higher-end model, which not only offers more processing units, but also some new features and tweaks. Certainly, the most important change is that it completely supports the current DirectX 11.2 standard. Previously, only a few AMD models have done so. It is likely that the GM204 is already compatible with DirectX 11.3 and DirectX 12, but  we cannot confirm just yet since the two versions are not currently finalized.

Also new is the implementation of DSR (Dynamic Super Resolution) and an innovative anti-aliasing mode called MFAA. DSR utilizes extremely high resolutions - higher than the native pixel count of the display - to improve the picture in a manner similar to superscaling. Furthermore, it should also set new standards in image quality thanks to its configurable Gauss filter. In contrast, MFAA improves performance: by changing the sample order between single frames, images should look subjectively similar between 4x MFAA and 4x MSAA, but with the performance penalty of 2x MSAA. Unfortunately, both features will only become available after a future driver update.

Nvidia's Battery Boost technology has been available since the launch of the GTX 800M series and continues to improve. In cooperation with notebook manufacturers, the available battery capacity can be dynamically distributed between the CPU and GPU, which should increase battery life. Otherwise, the basic procedure remains the same: Clock rates and voltages of both the CPU and GPU can be decreased via Nvidia settings and by limiting the frame rate to lower power consumption. Battery life can increase by up to 50 percent for older games with lower hardware requirements.