Nvidia's next gen Ampere GPUs to get 7 nm Samsung treatment, slated for 2020
Taiwanese publication DigiTimes recently confirmed that Nvidia’s next gen Ampere GPU architecture will be using 7 nm tech and is scheduled for a 2020 release. Nothing really surprising here, as Nvidia usually refreshes its gaming GPUs every 2 years, plus there are no actual details on how Ampere differs from the current RTX 2000 Turing family. However, there is one more detail that signals some interesting shifts in the market: Nvidia will rely on Samsung’s 7 nm EUV foundries rather than TSMC’s.
EETimes reports that Samsung managed to “aggressively undercut” TSMC, determining Nvidia to switch foundries. The report does not mention the exact details of the deal, but there are a few hypotheses going around. The more flimsy one stipulates that Samsung’s 7 nm EUV manufacturing process is better than the TSMC one and this assessment could have swayed Nvidia’s decision. Then there is the production capacity consideration. TSMC is currently working at full capacity on all of its 7 nm nodes, providing chips for Apple and AMD in particular, so this could affect target projected supplies for the Ampere chips. Samsung, on the other hand, could offer better supply streams, since its foundries are not booked by other major companies.
It will be interesting to see if Nvidia’s foundry switch is a temporary one. TSMC has been Nvidia’s choice for almost 2 decades now, making this switch even more surprising. This would be Samsung’s second “disruptive” move, as it also partnered up with AMD to bring Radeon GPUs to smartphones.