Nvidia to Join AMD on 14nm FinFET Process
Samuel Wan / 4 years ago
In of the biggest surprises of the year to come from the graphics industry, Nvidia is reportedly planning to utilize a new process node later this year. According to a South Korean paper, Nvidia has concluded negotiations with Samsung to utilize their 14nm process node to produce Pascal GPUs. This comes after the Pascal GeForce 10 series have already launched on the 16nm process from TSMC.
For quite a while, both Nvidia and AMD utilised the same process nodes at TSMC, with the differences between their GPUs largely due to design and architecture. The latest generation has shaken things up a bit recently, with Nvidia using the 16nm FF+ node at TSMC while AMD has opted for the 14nm FinFET node from Global Foundries. This has led to additional differences in terms of overclocking, power efficiency and other node characteristics.
It’s interesting that Nvidia has chosen to move to 14nm after most industry experts have agreed that there isn’t much of a difference between the 16nm TSMC and the 14nm Samsung/Global Foundries. Perhaps Nvidia is hoping to eek out some process gains, or strengthen their hand in negotiating with TSMC, or Samsung simply offered a better deal. At this point, there is no information yet about how many Pascal GPUs will be using 14nm or if they will even be consumer releases. Volta is set to launch next year and will likely use the same process node as well but with a new architecture.
With Nvidia on 14nm as well, AMD has lost the slim lead it had on the process side of things. Luckily, production won’t be affected as the supplier is Samsung and not Global Foundries though the two firms are sharing their process technology. Of course, the paper cited unnamed sources so we’ll have to see if this pans out. What do you think of Pascal on 14nm?