Nvidia is expected to release the RTX 4000 series of GPUs in mid -2022, and it is reported that the new GPU will be produced using a 5nm process from TSMC.
Before the RTX 4000 is released, there is also the possibility of releasing a new RTX 3000 series, which still uses an 8nm process from Samsung, as quoted by us from Techspot, Thursday (2/12/2021).
Well, with the more modern fabrication process, RTX 4000 is expected to have 50% higher gaming performance than RTX 3000. RTX 4000 is reportedly called 'Ada Lovelace' internally, but it is not yet known what architecture will be used.
The use of 5nm fabrication is reportedly much more expensive than the 7nm fabrication used by AMD in the RX 6000 series GPUs. This rumor is also reinforced by statements from sources in the component supplier industry.
Digitimes in its report said Nvidia plans to use TSMC's chip-on-wafer-on-substrate (CoWoS) packaging technology in the H100 series of GPUs. This model will be based on Nvidia Hopper's computing architecture, and is expected to be Nvidia's first multi -chip GPU module.
Meanwhile, for the Ada Lovelace GPU, it will most likely use ASE Technology's FC-BGA packaging technology, as well as the 'traditional' GDDR6X memory chip.
The use of 5nm fabrication technology should make Nvidia GPU chips faster and more efficient. But what to keep in mind, the Samsung 8nm fabrication used in the RTX 3000 is more mature than TSMC's 5nm which is more risky.
But other companies will also use the fabrication, such as the Apple A14, M1 Pro, and M1 Max. While AMD plans to use the fabrication in the Zen 4 CPU.