Updated, 7/5/19, 9:23am PT: Nvidia executive vice president of operations Debora Shoquist said in a statement that “Recent reports are incorrect – NVIDIA’s next-generation GPU will continue to be produced at TSMC. NVIDIA already uses both TSMC and Samsung for manufacturing, and we plan to use both foundries for our next-generation GPU products.”
Updated, 7/5/19, 7:25am PT: An Nvidia spokesperson responded to our request for comment on this story by saying that "We do not comment on rumors or speculation." We asked if that meant Nvidia was disputing The Korea Herald's report, or if Nvidia Korea's Yoo Eung-joon was mistaken about Samsung manufacturing Ampere GPUs using its 7nm EUVL process. The spokesperson said, "We already use both TSMC and Samsung, and qualify each of them for every process node. We can’t comment in any further detail on future plans, but both remain terrific partners."
Original article, 7/2/19, 8:08am PT:
Nvidia reportedly confirmed that it's partnered with Samsung to manufacture its Ampere GPU, which is expected to launch in 2020 using its 7nm extreme ultraviolet lithography (EUVL) process rather than continuing to source GPUs from longtime foundry partner TSMC.
The Korea Herald reported that Nvidia Korea leader Yoo Eung-joon confirmed the switch from TSMC to Samsung during a press conference. The companies don't appear to have formally announced the partnership on their websites--the most recent articles from Nvidia focus on the GeForce RTX 2060 and 2070 Super graphics cards that debuted today. We've emailed Nvidia for confirmation of their new deal.
Yoo didn't offer much information about the partnership. He called Samsung's production commitment for Ampere "substantial," according to the report, but declined to get into any specifics. Yoo was also said to have claimed that Nvidia's acquisition of Mellanox would close before the end of the year, pending regulatory approval from various countries and final approval from Mellanox's shareholders.
It's not clear why Nvidia decided to switch foundry partners for Ampere. EETimes reported in early June that Samsung "aggressively undercut" TSMC. We do know that TSMC's 7nm process has already become popular with Apple and AMD.
There are few details available about Ampere so far. It's expected to launch in 2020 and will succeed Turing architecture introduced in September. But if Nvidia follows a similar cycle as it did with Turing, we could learn more about the Ampere architecture during the SIGGRAPH conference in August.
