Nvidia Geforce Pascal with HBM2 will be launched during the second half of 2016

The standardization body JEDEC will get off to a flying start in 2016 with two memory technologies, which will primarily focus on graphics cards. One is GDDR5X and the other HBM2, both of which offer doubled bandwidth combined with larger memory capacities than current GDDR5 and HBM.

Both technologies are expected to be used in the next generation of graphics circuits from AMD and Nvidia, where GDDR5X is believed to be used in models for the entry-level segment and the popular middle class. The more expensive and in many respects more potent technology HBM2 is only expected to be useful in each company’s flagship.

Korean Digitaltimes now reports that within the next five to six months, Nvidia will complete its internal operational tests of graphics cards with HBM2, which will go into production in the same vein. The same report states that this will be followed up by a launch during the second half of the year.

The new architecture on which the HBM2-equipped graphics cards or circuit is based is Pascal, which will be manufactured on 16 nanometers with FinFET transistors at TSMC. According to rumors, it will be no less than 17 billion transistors for the top circuit, up from 8.0 billion for the Nvidia GM200 used by, among others, the Geforce Titan X and GTX 980 Ti.

The second generation HBM, HBM2, also means that the graphics circuit will probably have a real memory bandwidth of 1 TB / s. How much video memory it is about is not yet known, but probably the first graphics cards with the circuit will get 8 or 16 GB while computing cards in the Quadro and Tesla series can get a full 32 GB.