14 DAY TRIAL // The GeForce GTX 680 video card from NVIDIA is already the strongest single-GPU model out there, but Galaxy wants to shake things up even further.
Most of the OEM GTX 680 card versions stick to the reference specifications, but NVIDIA's partners are definitely creating means of differentiating themselves from each other.
Currently, we are looking at Galaxy, whose upcoming video card will have double the memory capacity, among other things.
That's right, as if 2 GB of GDDR5 weren't more than enough, Galaxy is making a GTX 680 4 GB.
Sixteen GDDR5 memory chips with a capacity of 2 Gbit each make that amount possible, though this means that some big changes had to be made to the PCB.
Then again, it would be more accurate to say that the standard printed circuit board was scrapped in favor of a totally new one.
As such, being built from the ground up, the PCB was able to integrate a 5+2 phase VRM that gets energy from two PCI Express power connectors, a 6-pin and an 8-pin one to be specific.
Furthermore, the cooler gets a large fin array, heatpipe(s) and two 90mm fans. It is one of Galaxy's own inventions.
There is another important factor to consider in all this, namely that the Kepler graphics processing unit itself is also taken beyond the reference settings.
Instead of running at 1,006 MHz, it will work at 1,100 MHz. Alas, the GPU Boost frequency was not specified.
Based on the leaked photos of the Galaxy GTX 680 4 GB, the connectivity of the GTX 680, at least, doesn't seem to have changed.
That means there will be two DVI outputs, an HDMI port and a DisplayPort ready to run single or multi-display setups (unlike previous-generation cards, Kepler can handle four monitors instead of just two).
UPDATE (Mar 28, 2012): Removed erroneous shader clock. The CUDA cores work at the same speed as the GPU.
