Archive notice: This article was originally published on March 20, 2012. Links and embedded videos are preserved as part of the historical record.
NVidia will officially present the new Kepler graphics chip on March 22. What is certain is that the leading graphics card will be called GeForce GTX 680 and will be equipped with 2GB of GDDR-5 memory. Successor versions will also be available with 4 GB of GDDR memory. The 28nm GPU installed there is equipped with 8 compute blocks, providing a total of 1536 compute cores. Compared with the previous top model, the GeForce GTX 580, the number of cores has tripled.
Dynamic Clock Adjustment is also among the innovations and is comparable with Intel’s Turbo Boost technology. A turbo mode for the GPU, where NVidia sets a maximum power consumption(1006MHz). Common video games are normally far below this value and therefore the clock frequency can be raised automatically during operation up to the maximum clock frequency.
Because the GeForce GTX 680 has low maximum heat output, two six-pin power connectors are sufficient for connection to the power supply.
A further new feature is that four LCDs can now also be operated in parallel on the GTX 680.
Anti-aliasing is said to work with a new mode known by the abbreviation TXAA. Image quality in TXAA mode is said to be comparable with 16xMSAA(Multisampling Antialiasing), while the performance loss is at most equivalent to what is required for 2xMSAA. TXAA2, by contrast, delivers far better quality than 16xMSAA with the performance loss of no more than 4xMSAA. So it remains exciting to see what Nvidia will report, particularly with regard to the chip’s turbo potential.
