A few days back, we reported that Nvidia might break its long ties with Taiwan semiconductor manufacturing company (TSMC). EE Times reported that Samsung proposed a better price to manufacture Nvidia’s upcoming Apmere architecture. In a blog post regarding the RTX 2070 and 2060 super graphics cards, Nvidia’s leader of the Korean division Yoo Eung-Joon confirmed the rumored partnership. He did not dive into the details of the partnership, but he called Samsung’s commitment to producing Ampere architecture “substantial.”
It is reported that Samsung was committed to outprice TSMC, which lead to the collaboration between the tech giants. Nvidia has also confirmed that Turing’s successor Ampere will be announced next year. They did not provide any information about the upcoming GPU architecture other than the fact that it will be based on Samsung’s 7nm extreme ultraviolet lithography (EUVL) process. It is quite a big jump for Nvidia, especially because of the competition. Nvidia’s current GPUs are being manufactured on the 12nm architecture which is considered obsolete by many semiconductor producers now.
Coming to the reasons why Nvidia chose to shift from its foundry partner. The obvious reason is the price. It is reported that Samsung aggressively undercut TSMC’s prices to get Nvidia’s upcoming project. One can think that Nvidia should have stayed with their old partner since its better to use a refined process after all. Economically and strategically speaking, Nvidia’s choice of Samsung was better. The price factor is enough for Nvidia to increase its profits.
While the move is more interesting if we look at it with an eagle’s eye, AMD recently announced its RDNA architecture which is a hybrid of the upcoming Navi and GCN architecture. The highlight of the architecture is its ability to scale. Theoretically speaking it can be used in mobile devices too. Now guess which company approached AMD to license its architecture for mobile devices. It was Samsung. Nvidia has partnered with Samsung on a much bigger scale. Its ties with Samsung can turn out to be beneficial for Nvidia on the competitive front against AMD.
Another economic reason that may have lead Nvidia’s shift is the demand for TSMC’s 7nm process. It should be noted that Nvidia will be the last processor maker to opt for the 7nm process. Major companies such as Apple, Qualcomm, and AMD have already sourced TSMC’s 7nm process for their respective offerings. Though TSMC is the biggest foundry, it is not possible for them to accommodate all of their customers. As mentioned earlier, Nvidia is the last to acquire the process, and they opted for a safer option, which is Samsung. Like TSMC, Samsung has mastered the 7nm process but was unable to get significant customers until Nvidia announced its plan of adopting the process.
It will help both companies, Nvidia will be able to produce its GPUs in a timely manner, and Samsung will get to boost its manufacturing process.