Mature GeForce GTX 780 Ti-based graphics cards are only just being released yet it appears that NVIDIA are already stepping up their plans for 2014. According to Scandinavian enthusiast site SweClockers their next generation may come sooner than thought in the form of a GeForce GTX 750 Ti built on the Maxwell architecture.
Maxwell is NVIDIA's successor to the well-regarded Kepler architecture central to the prior and current GeForce GTX 600 and 700-series. Ostensibly designed for 20nm rather than the 28nm of the current generation in order to provide a significant boost in performance per watt, said process clearly isn't quite ready yet. This mid-range Maxwell GPU therefore screams 'proof of concept', bringing with it a number of features new to NVIDIA GPUs before it's shrunk down to the more energy-efficient fab process.
Most notably the GTX 750 Ti will include both Unified Memory Architecture and a 64-bit 'Denver' ARM CPU core, which between them are expected to greatly improve GPGPU performance for a part in the mid-range. As well as capturing a small install base and independent feature testers, the GTX 750 Ti would also allow NVIDIA to feel out the market for improved GPGPU support in the mid-range.
Taiwanese technology forum Coolaler.com today released some early synthetic benchmark results on the GTX 750 Ti which muddy the water somewhat. Likely based on pre-release drivers and firmware, the new GPU undershoots a GTX 660 by as much as 20%. Maturing drivers have been known to provide great performance improvements so the cards potential is likely to be more robust than it might at first appear, but early indications are that the new card won't exactly set the world alight.
The new card is due for release on February 18th, whilst high-performance 20nm parts will likely not see the light of day for some while.
Sources: SweClockers, Coolaler.com via TechPowerUp