PCB Design and Specifications
The new GeForce GTX 560 Ti is about 2 centimeters longer than its predecessor GeForce GTX 460 1GB (we mean the latter’s reference version). It is 23 centimeters long and should fit into most system cases without much difficulty.
You may only have some problems installing it into a short system case because the power connectors are placed at the shorter side of the PCB. This was not a problem with the reference GeForce GTX 460 but the extra couple of centimeters may get in the way here. So, you may want to make sure the GeForce GTX 560 Ti will fit into your system case prior to purchasing it.
The cooling system is fastened with ordinary crosshead screws rather than with Torx T2 ones as in the GeForce GTX 570. Hopefully, Nvidia will keep on using the more popular type of screws.
The PCB layout of the new card hasn’t changed much compared to its predecessor. One of the memory chips is still situated to the left of the GPU whereas the other seven form a letter L on the right. The new PCB seems to have borrowed some parts of the reference PCB of the GeForce GTX 460. The power system follows a 4+1 design. The increased clock rates of the GF114 make the stronger GPU voltage regulator justifiable.
The GPU is still based on an NCP5388 controller from ON Semiconductor. The memory voltage regulator is based on a Richtek RT8101 chip located below the chokes of the main power system. Like its predecessor, the GeForce GTX 560 Ti has two 6-pin PCIe 1.0 connectors with a max load capacity of 75 watts. The reference PCB design does not provide for an 8-pin PCIe 2.0 power plug.
The new card is equipped with Samsung’s popular K4G10325FE chips. You can see them on board the GeForce GTX 580 as well as many other cards. This is GDDR5 memory type. The chips have a capacity of 1 Gb (32 Mb x 32), the HC04 suffix denoting the rated frequency of 1250 (5000) MHz. The card can lower its memory frequency in its two power-saving modes to 324 (1296) MHz or to 135 (540) MHz. The memory chips are connected to the GPU with a 256-bit bus and has a default clock rate of 1002 (4008) MHz, delivering a peak bandwidth of 128.3 GBps.
The new chip is externally no different from the GF104 and has the same dimensions. According to Nvidia, the die is 360 sq. mm but we can’t check this out because it is covered with a protective heat-spreader. The marking on the GPU indicates that this sample is revision A1. It was manufactured on the 46th week of 2010, in late November, when the GF114 was being mass-produced already. The middle number 400 suggests high frequency potential just as expected since the core of the GeForce GTX 560 Ti must be able to work at frequencies above 800/1600 MHz.
The last version of GPU-Z doesn’t yet identify all the features of the GeForce GTX 560 Ti. It doesn’t report such parameters as tech process, die size, transistor count, and announcement date. The DirectX Support and Texture Fillrate fields are empty, too. The bottom line erroneously says that the GPU doesn’t support PhysX. On the other hand, the key features like the number of ALUs and clock rates are identified correctly. The 1MHz difference from the official specs is unimportant.
We can also add that the GF114 has 64 texture processors capable of filtering FP16 textures because there are four filter units per a texture address unit in each TMU. The GF104 had the same texture-processing capabilities, though. Like other solutions from Nvidia, the new card supports two power-saving modes. When decoding HD video, it lowers the GPU frequencies to 405/810 MHz. When the GPU is idle, the frequencies drop to 51/101 MHz. Thus, the GeForce GTX 560 Ti should be as economical as the GeForce GTX 460 in these modes.
The reference design of the GeForce GTX 560 Ti doesn’t provide for a DisplayPort although the GPU itself supports that interface. The card has the same connectors as the senior models of the series: two DVI-I ports and one HDMI 1.4a connector. The top of the card's mounting bracket is a vent grid for exhausting the hot air from the cooler. There is one MIO connector on board, so you can’t build a SLI configuration with more than two GeForce GTX 560 Ti cards. More advanced SLI systems can only be assembled out of senior GeForce 500 series products.
This selection of interfaces is not as gorgeous as you get with the Radeon HD 6000 series, but only few users really need to connect six monitors simultaneously or link displays via DisplayPort in a daisy chain. Most gamers use one, occasionally two, monitors with DVI interface. Large panels are connected via HDMI. Thus, the GeForce GTX 560 Ti offers the reasonable minimum in this respect. Its GPU supports DisplayPort 1.1, so some custom-designed GeForce GTX 560 may come with an appropriate connector.
The reference GeForce GTX 560 Ti doesn’t use an evaporation chamber in its cooling system although it has done well for the GeForce GTX 580 and 570. The new card uses a more classic cooler with a round central heatsink that resembles the boxed coolers from Intel. The central piece is connected to the two additional curved heatsinks with three heat pipes. The axial blower is driving the air downwards as well as sideways to cool all the heatsinks. Some of the hot air will be exhausted to the right, into the interior of the system case.
The cooler’s base is quite standard. The aluminum frame serves as a heat-spreader for memory chips and power circuit elements, taking the heat off them via elastic thermal pads. A layer of dense gray thermal grease is applied between the main heatsink and the GPU.
We guess this cooler is rather odd, especially as Nvidia has designed very efficient coolers for the senior GeForce 500 series products. The only reason why Nvidia didn't use an evaporation chamber here is that the cost of such a cooler would be too high. But perhaps this solution is quite efficient, too. Let's check this out.