The next generation of graphics cards
NVIDIA will be burned using TSMC’s 5nm process or not.
Like AMD or Intel, NVIDIA has planned to switch to a new finesse of engraving for its next generation of graphics cards which are generally called GeForce RTX 4000.
A switch from Samsung to TSMC
On the Ampere generation, NVIDIA was satisfied with an engraving fineness of 8 nm. To do this, the American was in contact with Samsung and took advantage of the South Korean 8nm LPP process.
For its part, with its competing Radeon RX 6000 series range, AMD could count on TSMC’s 7 nm process. This technical difference does not seem to have played too much on the results for the end user, with NVIDIA cards generally considered to be a little more efficient than their AMD counterparts.
NVIDIA’s Lovelace architecture should mark a notable change with the switch to TSMC, like AMD therefore. While it is not a “historic” Taiwanese customer, NVIDIA seems to have to take out the checkbook to reserve production capacity.
A total advance of $ 6.9 billion
Taiwanese TSMC appears ahead of the manufacturing process. Indeed, neither Intel nor Samsung seem able to offer 5nm as TSMC can do… and this raises the question of TSMC’s production capacities. To believe HardwareTimes who quotes MyDrivers, not all TSMC customers are in the same boat.
Apple, AMD and MediaTek would not have to pay in advance to have “slots” on the production lines of the Taiwanese. On the other hand, for companies like NVIDIA or Qualcomm, the sums to be advanced can be tidy, even if we have no details on the negotiation process.
It seems that NVIDIA has already paid 1.64 billion dollars in the last quarter to reserve its place on the 5nm. Another payout of $ 1.79 billion is expected in the first quarter of 2022, and in total NVIDIA could shell out $ 6.9 billion to secure its supply.
On the same subject :
AMD is TSMC’s third largest customer
Sources: MyDrivers
, HardwareTimes
14