Nvidia may Skip 20nm and go straight to 16nm finFET
GK104 may retire soon. 800 Series Incoming?
Published: 9th July 2014 | Source: videocardz |
800 Series Incoming?
The exact plans of AMD and Nvidia are a mystery at present, there have been many problems when it came to the 20nm process technology that has undoubtedly delayed the next generation of graphics cards.
Rumors until now hinted at the next generation was delayed until at least the end of next year and will be using the TSMC 20nm process node. Now Vidiocardz has reported to the contrary, citing semiaccurate as a source (which sadly requires a paid subscription to access). They say that new 28nm based Nvidia GPUs will be made available starting in October of this year and will be using the GM204 GPU core with between 1920 to 2560 stream processing units. These GPUs are to be names as follows.
- GeForce GTX 880 Ti
- GeForce GTX 880
- GeForce GTX 870
- GeForce GTX 860
One thing to not here is that the Flagship GPU here has 2560 GPU cores, which is less than that of the current Nvidia Flagship GPUs. This suggests that the Next Generation Nvidia GPUs will not be much more powerful that the current generation or Nvidia has greatly increased their GPUs efficiency with Maxwell.
Nvidia to skip 20nm to use 16nm finFETs
TSMSs 20nm process has been a strange one, at present only one version of this process is available which can be used to make low cost mobile chips to even high end processors (like desktop CPUs and GPUs). This process has been considered too expensive by both AMD and Nvidia, as process yields are reported to be very low and production capacities are very limited.
This is where 16nm FinFET process technology comes in, for early 2015 it makes much more sense than 20nm manufacturing technology as it will be more profitable and usable by AMD and Nvidia. Not only that but TSMC have already developed a FinFET+ node, which is reported to be better for mass volume chip production later in the year which will again have notable benefits for the companies that use it.
According to TSMC the 16nm FinFET+ provides up to 15 per cent performance improvement over normal 16nm FinFET at the same level of power consumption. At the same clock-rate, chips produced using 16nm FinFET+ are expected to consume 30 per cent less power compared to the same chips made using 16nm FinFET. Also current 20nm fabs are said to be 95% ready for the 16nm process, meaning TSMC can easily move to handle the volume of orders of 16nm parts.
I will say now that this info remains unconfirmed by Nvidia, and that will likely not change soon. If you want to discuss this topic further please head over to the OC3D Forums.