Nvidia pushes Samsung to make chips on TSMC's 3nm line

Nvidia has asked its three HBM suppliers to shift production of a critical component to TSMC's advanced process, breaking Samsung's traditional approach.

5 cards · 1m 30s · tap to begin

From · · · 1m 30s

Nvidia pushes Samsung to make chips on TSMC's 3nm line

Nvidia has asked its three HBM suppliers to shift production of a critical component to TSMC's advanced process, breaking Samsung's traditional approach.

In brief

Nvidia has asked its three HBM suppliers to shift production of a critical component to TSMC's advanced process, breaking Samsung's traditional approach. Nvidia asked Samsung, SK Hynix and Micron to make base dies for its custom memory chips on TSMC's 3nm process for consistency, marking Samsung's first joint work…

Samsung breaks with tradition for Nvidia

Samsung will make the base die for Nvidia's custom HBM on TSMC's 3nm process. This marks the first time Samsung and TSMC have worked together on high-bandwidth memory production.

What Nvidia asked for

Nvidia asked all three NVHBM suppliers, Samsung, SK Hynix and Micron, to make base dies on TSMC's 3nm process. It also wants memory dies standardised on one c, the sixth generation of DRAM technology.

Why consistency matters

Using a common process means Nvidia does not have to validate each supplier's chips separately. When suppliers use different methods, one design is hard to make perform the same way across all of them.

The base die frees up GPU space

NVHBM moves the memory controller off the GPU onto the base die, freeing space for more computing circuits. The result is up to thirty percent more bandwidth and fifteen percent less power than standard HBM4E.

Samsung hedges its bets

Samsung will keep its turnkey approach for standard HBM4 and HBM4E using its own 4nm base dies.

What it means

Nvidia asked Samsung, SK Hynix and Micron to make base dies for its custom memory chips on TSMC's 3nm process for consistency, marking Samsung's first joint work with TSMC on high-bandwidth memory.

Read the original on The Next Web

React

Sign in to react and comment.

Comments (0)

Life is short. Keep it sweet. Respect others' opinions and be kind!

    Recommended next

    More decks on nvidia and related topics.