Bridging the gap between the GeForce gaming cards and the Tesla Workstation cards, the new GeForce GTX Titan takes the reign of the newest $1000 videocard.
Featuring 2688 CUDA cores compared to 1536 in the GTX 680 and the Titan can make the most of its hardware. With 6 GB of GDDR5 RAM and a 384-bit bus compared to 2 or 4 GB of GDDR5 on the GTX 680 on a 256-bit bus, the memory bandwidth is much higher as well. Even though the clock speed is reduced to 837 mhz compared to 1006 mhz in the GTX 680, the sheer number of cores more than makes up for the deficit. Finally, the double precision floating point (DPFP) calculations are effectively 8 times higher per clock in the Titan which should incredibly enhance some GPU-heavy software such as that used by computational biologists.
You may think that the GTX Titan is more comparable to the GTX 690 card being that they are both around the same $1000 price, but the GTX 690 is essentially two GTX 680 cards running at a slightly reduced speed and the bottleneck of two separate processors each with its own RAM makes it an odd comparison.
What I am the most impressed with is the significantly increased DPFP capabilities of this card. While most games won’t benefit as much, CUDA accelerated software such as Adobe Premiere could. In addition, because the Titan is its own card and not a hybrid of 2 cards as the GTX 690 is, true 3-way SLI can be used with three GTX Titans giving you control of over 8000 CUDA cores and allowing your system to take on any game out there at any resolution you want (even across multiple monitors) without so much as a drop of sweat. On Thursday the NDA for benchmarks will drop so keep your eyes open for performance analysis across the tech news world.