Nvidia Geforce GTX780 Ti unveiled – “fastest ever built”

Nvidia has officially launched the Geforce GTX780 Ti, scheduling it for a late November release to coincide with the launch of next-gen games featured in Nvidia’s Battlebox program.

Not much is known about the GTX780 Ti at this point. Jen-Hsun Huang, on the stage at Nvidia’s Montreal event on 17 October, described the GTX780 Ti as “the fastest gaming graphics card we’ve ever built.”

This may be a version that was going to be called “Titan Ultra” featuring the full implementation of Nvidia’s Geforce GK110 design based on the Kepler architecture. It is known that Nvidia is currently working on a new card in the Tesla and Quadro family with everything enabled in the silicon.

The current implementations of the GK110 design is not opening up the floodgates entirely. According to Nvidia’s whitepaper on the Kepler architecture, most graphics cards based on it will only have 13 or 14 out of 15 possible shader groups enabled for yield purposes.

A fully unlocked design would feature 2880 CUDA cores, 240 texture units, 48 Raster Operators (ROPs), a 384-bit bus, and up to 6GB of video memory. Nvidia may not want to devalue the purchase of a GTX Titan, and so may elect to only raise the specifications slightly.

In context of how the GTX780 Ti would fare against the current GTX780 and the GTX Titan, it may be much faster than the GTX780 and Nvidia may introduce some price drops to combat AMD’s Radeon R9 290X.

Below is a table displaying a comparison of the full implementation of GK110 against where I think the GTX780 Ti should sit. Performance should be in between a GTX780 and a Titan.

Specification	Kepler GK110	Geforce GTX 780 Ti	Geforce GTX Titan	Geforce GTX780
CUDA Cores	2880	2496	2688	2304
Raster Operators	48	48	48	48
Texture units	240	208	224	192
Clock speed	—	—	837MHz	863MHz
Boost speed	—	—	876MHz	900MHz
Memory speed	—	—	1500MHz	1500MHz
VRAM buffer	6GB	6GB	6GB	3GB
Bus width	384-bit	384-bit	384-bit	384-bit
TDP	250W	250W	250W	220W
Launch price	—	—	$999	$649