In early June, DigiTimes reported that NVIDIA would be producing their next-generation graphics card at Samsung’s foundries. This story was picked up by various sites and then faded out through a typical news cycle. It unexpectedly picked up again on Tuesday, July 2nd, however. The Korea Herald published a story that claims Yoo Eung-joon (who leads NVIDIA’s Korean division) held a press conference in Seoul, where he allegedly confirmed that Samsung is manufacturing their upcoming GPU with its 7nm EUV process.
Tom’s Hardware noticed this and published a story on the same day. A few days later, on that Friday, NVIDIA responded to Tom’s Hardware with the usual “We do not comment on rumors and speculation.” A couple hours after that, they commented on the rumors and speculation with this statement from Debora Shoquist, Executive VP of Operations: “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.”
This statement seems to declare that NVIDIA will manufacture the GPU at two different locations: both TSMC and Samsung. I am curious whether this means that both fabs will produce all designs, or that TSMC and Samsung will be responsible for different chips.
In the former case, it would be interesting to see how supposedly identical chips differ based on where they were produced.
In the latter case, it would be interesting to find out what drove this decision. It’s possible that NVIDIA just wanted more production, so they load-balanced between tow companies. It’s also possible that NVIDIA thinks that one is better at high-performance situations while the other is better at mass production. It’s also possible that NVIDIA really is heavily weighing one over the other, and the statement is just attempting to pour water on the sparks.
We will probably find out at some point in the next twelve months.
tsmc makes most nvidias cards. samsung makes the 1050 ti and lower cards. that will probably continue into the future. they are probalby making the 16 series cards while tsmc makes the 20 series although i havent looked into it. the reason may be because samsung is better at lower power but more expensive so they use tsmc for the big dies and samsung for the little low power dies. who knows how that will continue or if those 2 situations will reverse. samsung is supposed to be getting aggressive with 7nm prices.