NVIDIA Kepler GK104 - back to games?

So far, the consensus has been that in the long run, the overall computing power of the GPU represents a higher value than the speed shown in three-dimensional games. This trend has also had a significant impact on NVIDIA Fermi graphics cards.

Although the GeForce GTX 680 has extremely extensive API support (CUDA C, CUDA C ++, CUDA Fortran, OpenCL, DirectCompute, and Microsoft C ++ AMP), which are indeed key to general-purpose programmability, something is very much missing. With some brazenness, we could even say that a day after its release, we are still looking for more complex GPGPU applications where the GK104 chip performs really well. We can already see that we are not lying to him.

Double-precision computing power is an important factor in the GPGPU market for tasks rich in performance-intensive, floating-point computing, but also in general. At this point, the new Kepler tile can be attacked quite seriously. The chip is theoretically capable of twenty-quarters of single-precision computing power with double-precision, which immediately strikes in the SiSoftware Sandra 2012 measurement program. The GeForce GTX 680 can't even beat its predecessor, and the AMD Radeon HD 7900 series moves in a completely different category. For further optimization, engineers did not implement ECC and virtual memory support, and the size of the various caches was limited to what was still acceptable, so the GeForce GTX 580 had one and a half times more secondary storage than the newcomer.

The situation is no better under Luxmark 2.0, which is recognized in ray-tracing testing, so to speak, the above image is repeated. A new driver may still improve, but AMD Radeon HD 7900 series cards still seem out of reach. ComputeMark can be used to discover the computing capacity of DirectCompute. The interface released as part of DirectX 11 is used by Battlefield 3 and Dirt 3 for the ambient occlusion effect. The situation is better here, but the GeForce GTX 680 is still only around the Radeon HD 7870. Last but not least, SPECviewperf 11, which is a "workstation test program" written for OpenGL. The application performs extensive simulations and produces an extremely complex load, which is due to the large models and textures. The testing also covers applications such as Siemens NX 7, Siemens Teamcenter Visualization, Pro/ENGINEER Wildfire™ 5.0 and CATIA V6 R2009. It is clear that the controllers based on the GCN architecture are really in their element here. In our opinion, this is due to particularly high loads, where it can easily be seen that the HD 7970 has four times more register space than the GeForce GTX 680. From the above data, we definitely conclude that NVIDIA’s real joker card will not be the GeForce GTX 680, but a GK110-based controller. The chip numbering also confirms that the GK104 represents the gamer line, which the GF104 and GF114 have already touched quite well. Accordingly, the first Kepler-based chip is not brilliant under GPGPU applications, but it is unbeatable in games, and chip size and power consumption have been distinctly appealing. This also reflects the fact that, through the GCN, AMD has very strongly targeted professional needs, the wider use of the GPU. Due to this, Tahiti-based cards offer less performance in most games than the GeForce GTX 104, which uses the GK680.

Somewhat surprisingly, in the above applications, very often significant differences developed between the two cannons. In conclusion, we would definitely note that some of them are professional or synthetic programs. Closer user solutions such as members of the CyberLink product family may benefit more from the new GeForce. CUDA is still considered a large-caliber weapon in the backyard of the Greens, however, the first Kepler was not as expected in all areas.

Source: vr-zone.com