The Pascal GPU powered graphics cards were set for launch in the first half of 2016 but recent reports suggest that it might get pushed back.
Last March, NVIDIA co-founder and CEO, Jen-Hsun Huang, teased the future of NVIDIA GPUs through an audio-visual presentation during GPU Technology Conference (GTC) 2015, which was held in San Jose, California last March. Huang revealed two new graphics processors, Pascal and Volta, which his company plans to release in 2016 and 2018, respectively. Pascal's release will mark the end for the current age of 28nm GPUs. Note that the company's current leading graphics card, the Titan X and GTX 980Ti are both running on a 28nm Maxwell chip.
The GP100 chip series, codenamed Pascal, is NVIDIA's next bet to keep the lead against its rival graphics card manufacturer, AMD. The card series consists of the GP100, the "big" Pascal chip, and the trimmed down versions of it, GP104 for instance, denoted in an inverted nomenclature. Rumors say that NVIDIA will release the GP100 chip to professional markets first, which will pave the way for the less powerful cards that are bound for consumers.
Jen-Hsun Huang pointed out that the upcoming Pascal graphics card will be, rough estimates, roughly ten times faster than the Maxwell series. The CEO attributed this to three "very important" advancements: Mixed Precision, which points to a high performance gpu; 3D Memory, which is basically HBM2; and NVLINK, which enables much better GPU and CPU communication.
A forum post from Beyond3D alleges that the Pascal chip series was already "taped out" early last June and is on 16nm FinFet Plus of Taiwanese Semiconductor Manufacturing Company (TSMC). For the uninitiated, "taped out" means that the initial processor design is already complete and thus, the next steps would be creating the masks, a template of sorts in preparation for mass production.
Following NVIDIA's previous releases, approximately nine months after the chips were "taped out", the graphics cards hit the market. Hence, the Pascal cards are expected to be around March of 2016, which is in line with the late first quarter - early second quarter project release date.
However, the Green team may have overlooked something and that is AMD's "priority access" to HBM2. TweakTown has recently reported that due to AMD's joint venture with SK Hynix in the development of High-Bandwidth 3D Stacked memory (HBM), the company secured itself priority access to the first generation of HBM2. "With NVIDIA set to use HBM2 on their upcoming 16nm-based Pascal architecture, AMD will have the one-up on its competitor. AMD should be able to secure most of the HBM2 produced at the time, which is exactly the position AMD wants to be in, come 2016," the TweakTown article concluded.
If the news article holds water, then 2016 may finally be AMD's year in getting a head start against its more dominant rival. Note that HMB2 is projected to have a memory bandwidth of 1TB/sec, three times faster than that of NVIDIA GTX 980Ti and two times that of the current HBM1 on AMD Fury X.
