Nvidia’s planning goes awry – Maxwell is delayed by one year

Already at the launch of the original Fermi, it was said that not everything was right with the semiconductor manufacturer TSMC, which produces graphics circuits for both AMD and Nvidia. Problems with the transition to new manufacturing technology caused constant postponement of launch dates, something that also affected the later generation Kepler.

VR-Zone now reports that this gives Nvidia gravel in the machinery. The next generation Maxwell will not appear until 2014 at the earliest, which means that the new architecture will be launched at least one year later than planned.

Maxwell will be manufactured with the 20-nanometer technology Gate-Last HKMG, which is not only used by TSMC but also IBM, Samsung and Globalfoundries in the Common Platform Alliance. This means that Nvidia has a straightforward opportunity to let other semiconductor manufacturers produce the circuit, should it turn out that TSMC cannot live up to its commitments.

The new architecture is also the first to be used for all of Nvidia’s graphics parts, from Tegra to Geforce, Quadro and Tesla. In addition, it is said that Maxwell includes an integrated 64-bit ARM processor known by the code name “Project Denver”. With this in place, the graphics card can become independent of the traditional CPU and in practice be used to run its own operating system.

More Kepler approaching

Before Maxwell, however, more graphics parts are launched with the Kepler architecture. After GK104 and GK107 comes GK106 for the wide middle class, more specifically Geforce GTX 660 and 650, which will be launched in September.

The mythical GK110, known as the “full-featured Kepler”, will also appear on the Tesla family’s computing card towards the end of the year. Then there is talk that the GK110 can also enter the consumer market in the form of the Geforce GTX 780.