TechPowerup is reporting that nVidia will unveil it’s much anticipated Pascal architecture at Computex 2016. Pascal is the successor to the Maxwell architecture found in the GTX 900-series cards.
The largest change from previous generation Maxwell and Kepler GPU’s will be its fabrication process; nVidia will be using TSMC’s 16nm FinFET process for these cards. FinFET transistors utilize a “fin” that wraps the conductor, allowing for reduced power usage. Intel has had FinFET as part of its process since 2012, debuting in Ivy Bridge. Pascal will be the first architecture nVidia has made using FinFET. Pascal will also feature a switch to second generation High-Bandwidth Memory (HBM2) to replace GDDR5. First generation HBM can be found on AMD’s Fury line, which allows for reduced power consumption, faster memory speeds, greater bandwidth, and physically smaller memory modules.
Pascal will go up against AMD’s second generation of its Graphics Core Next (GCN) architecture, codenamed Polaris. Like nVidia, they too had been constrained at 28nm, and this will be their first new architecture on a new process since 2011.
Still unknown is what form the first Pascal cards will take. With Maxwell, nVidia chose to debut it in the form of the GTX750, an entry level card. Given that this will be a new process, it is plausible that this will be repeated, with the larger flagship, GP100, coming later.
Quick Take
It’s been a rough last five years for enthusiasts. Both nVidia and AMD have been limited by the fact that there wasn’t anything better than TSMC’s 28nm process. The delay of 16nm meant that nVidia had to scale Maxwell up to 28nm to release new products, and AMD had to add HBM to their existing designs. Pascal and Polaris, fabricated on TSMC 16nm and GlobalFoundries 14nm, respectively, will be the first new cards in five years that are completely new.
