NVIDIA’s release of the GeForce GTX 560 Ti comes with a new chip (the GF114). The card is intended to replace the old GTX 460 (GF104), and fill the lowest high-end slot for the GeForce 500 family.
With the release of the newest high performance cards, like GTX 580 and 570, which utilize a redesigned and optimized GF110 chip, NVIDIA was able to take the crown for the production of the fastest and quietest single GPU cards in the world. Both the GTX 580 and 570 were able to excel in performance, and the GTX 480 with its GF100 chip became obsolete. Unlike the GF100 chip, which had 512 CUDA cores but was able to utilize only 480, the GF110 lives up to it full potential and uses all 512 CUDA cores. The GTX 560 Ti is also quipped with 8 PolyMorph Engines dedicated to tesselation and additional raster engines that covert polygonal shapes to pixel fragments to allow faster culling.
The successful release of GTX 580 and 570 has left many gaming ethusiasts wondering about the release date of lower-budget high performance GTX 560. Its predecessor, the GTX 460, was a tremendous success, providing excellent performance at a significantly lower price than GTX 470 or 480. We gave the GTX 460 in SLI our Golden Bear Award and Best Bang For the Buck Award. As of January 25, 2011 NVIDIA is officially releasing the GTX 560 Ti. In designing the GTX 560 Ti, NVIDIA tried to create a graphics card that would demonstrate devastating power and lighting fast speed packed in a light body. NVIDIA has reinstated the old Titanium designation to emphasize the light weight and strength of this card.
The GTX 560 Ti uses a GF114 chip, which is a redesigned and optimized version of the GF104 chip. Like all GeForce GTX 500 series cards, the GTX 560 Ti is designed with a greater performance to power consumption ratio than the GTX 460. As the result of these changes, the power leakage was minimized while the clocks were significantly increased. GTX 560 Ti has also been equipped with 384 CUDA cores and eight PolyMorph Engines in order to provide significantly higher performance than its predecessor GTX 460.
Certainly, with such modifications, the power usage has increased from 120W to 170W, but with this increase, the GTX 560 Ti is able to demonstrate 21% better performance per watt than the GTX 460. In DX11 games, NVIDIA estimates a 33% increase in performance in comparison to the GTX 460, and a 40-50% in performance in comparison to the HD 6870.
Though DirectX 11 was introduced in early 2010, only few games have actually used DX11 to its full potential: better tessellation, improved multi-core support, and support for multi-threaded applications. According to a survey, NVIDIA has determined that 84% of gamers have not yet updated to DirectX 11. However, DX11 is expected to gain ground among gamers in 2011, with the anticipated release of DX11-compatible games such as Crysis 2 and Battlefield 1943.
The GF114 Architecture
The GF104 Fermi architecture is based on parallel computing units (CUDA cores) that simultaneouly process different information. Therefore, the efficiency of the processor increases with the increase in amount of the CUDA cores present. The GTX 460′s GF104 chip contained 8 Streaming Multiprocessors (SM) with each linked to up 48 cores. However, only 7 out of the 8 SM’s were actually functional, meaning the number of usable CUDA cores was reduced to 336. Additionally, the GPU was equiped with 32 ROPs, 2 Raster engines, 64 textile units, and 8 geometry units.
NVIDIA completely redesigned the GTX 460.The new GPU GF114 now supports 384 CUDA cores. Additional Raster Engines and PolyMorph engines were added to provide better tesselation support and improved polygon-to-pixel conversion. The total memory bandwidth for the GeForce GTX 560 Ti has increased to 128.3 GB/s.
Due to the larger amount of active CUDA cores, the default clockspeed for a stock GTX 560 Ti was raised to 820 MHz, requiring a greater power consumption. Changes that occured in order to lower the temperature were performed on transistor level to ensure that the card has an excellent performance without excessive power loss. As compared to the GTX 460, the overall performance of GTX 560 Ti increased by 33%, and the performance per watt went up by 21%.
Not only is GTX 560 Ti faster than the GTX 460, it demonstrates a drastic increase in performance over the AMD HD 6870, which becomes markedly observed in DX11 games. NVIDIA claims that there is a 1.5-1.65x increase in performance in latest games like Civilization 5, Lost Planet 2, and HAWX 2. We’ll test this contention later on.
The GTX 560 was equipped with 8 PolyMorph engines in order to provide a great DirectX 11 experience. NVIDIA sees DirectX 11 as a future for the gaming industry and is gearing the products to serve the 84% that have not yet switched to Direct X11 in order to enjoy the revolutionary depth of field implemented in newer games. NVIDIA’s cards are one generation ahead in terms of DirectX 11, and with 8 PolyMorph engines in comparison to the AMD HD 6950′s 2 PolyMorph Engines, GTX 560 Ti shows 2.5x as much performance as the HD6950.
Due to the modifications made to accomodate 384 CUDA cores, the power consumption has been increased, but the card is still supplied by 2x 6-pin power connectors, and has an estimated consumption of 170 W. However, the card was designed to create a better performance per watt ratio. NVIDIA claims that in comparison to the AMD HD 5970, the GTX 560 Ti has up to 50% greater efficiency in power consumption.
With a clockspeed of 820 MHz, this card leaves plenty of room for any overclocker. It is also important to note that NVIDIA has implemented power monitoring features that monitor each 12 V rail. This features allows for the user to test the voltage with a lower risk of damaging the card. Once the threshold voltage is reached the card cuts power in order to avoid damage. This feature, common to the whole 500 series, differs from older cards, which would simply downclock the GPU when the card reached critical temperatures. An overclocked version of GTX 560 Ti will also be available. The power monitoring feature is not part of the reference design, and whether or not to include it will be the vendors’ decision.
What is CUDA?
CUDA is NVIDIA’s parallel computing architecture that enables dramatic increases in computing performance by harnessing the power of the GPU (graphics processing unit).
With millions of CUDA-enabled GPUs sold to date, software developers, scientists and researchers are finding broad-ranging uses for CUDA, including image and video processing, computational biology and chemistry, fluid dynamics simulation, CT image reconstruction, seismic analysis, ray tracing and much more.