Nvidia Corp. is working on a yet another consumer-oriented graphics card based on code-named GK110 graphics processing unit (GPU). The product will be less powerful than the GeForce GTX Titan and will fill the gap between the GeForce GTX 680 and the most powerful single-chip graphics board in the world. The novelty will likely be available in higher quantities than the GTX Titan.

Given the vast amount of transistors (7.1 billion) and execution units inside the GK110, it is barely possible for Nvidia and its manufacturing partner Taiwan Semiconductor Manufacturing Co. to make a chip with all 15 SMX clusters and 2880 stream processors working correctly, which is why all commercial products powered by GK110 feature lower amount of SMX clusters/stream processors than the chip physically has for the purpose of higher yields.

At present, Nvidia uses GK110 chips with 14 SMX (2688 stream processors) and 384-bit memory bus for its GeForce GTX Titan and Tesla K20X products, whereas the configuration with 13 SMX (2496 stream processors) and 320-bit memory interface is used for Tesla K20 compute accelerator boards.

In a bid to offer a product that fills the gap between GeForce GTX 680 ($449) and GeForce GTX Titan ($999), Nvidia plans to use a cut-down version of the GK110, potentially with 13 SMX clusters 208 texture units, 40 raster operating units and 320-bit memory bus, to create a brand new graphics card, reports SweClockers web-site. The product will feature 5GB of GDDR5 memory. The name of the novelty has not been decided yet, but GeForce GTX 685 sounds rather logical.

Nvidia did not comment on the news-story.

Tags: Nvidia, Geforce, GK110, 28nm, Kepler, TSMC, Tesla