NVIDIA will reportedly unveil its 28nm GPU architecture 2012 and 22/20nm sometime in 2014, respectively, one year later than originally planned, due to low yield rates of Taiwan Semiconductor Manufacturing Company's (TSMC) 28nm process.

NVIDIA shares showed some early weakness Friday following the report.

Nvidia was set to announce its new 28nm GPU architecture at the end of 2011 and 22/20nm in 2013. However, sources from graphics card makers in Taiwan believe that the company has adjusted its roadmap and delayed 28nm Kepler and 22/20nm Maxwell GPUs to 2012 and 2014, respectively.

Digitimes.com reported that the delay is related to unsatisfactory yield rates of Taiwan Semiconductor Manufacturing Company's (TSMC) 28nm process as well as lower-than-expected performance of Kepler.

TSMC had also met low yields at its 40nm process in 2009, causing delays to the introduction of Nvidia's Fermi-based GPUs as well as to AMD's 40nm GPU solutions.

TSMC originally expected its 28nm capacity at Fab15 to be available in the 2011 fourth quarter, with pilot production of its 20nm process slated to commence in the third quarter of 2012.

Nvidia's 22nm "Maxwell" chips will be the first chips that will combine an ARM CPU, with NVIDA GPUs (Project Denver) for super-computing.

Both the "Maxwell" and 28nm "Kepler" generation chips will consume "a little more" power than the current generation Fermi chips Fermi chips but they will also be faster, according to Nvidia chief executive Jen-Hsun Huang. The Kepler processor will be three to four times faster than Nvidia's current Fermi chip generation, while the Maxwell chips will be ten to 12 times the power of Fermi, Huang added.

In the meantime, AMD is expected to release its next-generation 28nm graphics chips Southern Island as well as Krishna and Wichita APUs in the first half of 2012.

The 28nm Southern Island family has reportedly seen its tapeout completed by TSMC, which is expected to start contract manufacturing of the chips for AMD by the end of this year.

"Krishna" and "Wichita" will replace AMD's current Ontario and Zacate chips. They will feature two and four-core 28nm APUs based on the next-generation sub-one watt "Bobcat" CPU cores and a DirectX 11-capable GPU, designed for the tablet, notebook, HD netbook, and desktop form-factors.