NVIDIA may skip Ada Lovelace’s GPU architecture in Hopper to better compete with AMD’s next-gen RDNA 3 architecture.
It’s a great way to spend the weekend—super juicy GPU rumors about NVIDIA and AMD’s next-generation Ada Lovelace, Hopper, and RDNA3 architectures. We have more NVIDIA news… huge news…
Kopite7kimi claims that NVIDIA cannot beat AMD’s next-gen RDNA 3 architecture with its monolithic design. The company will then deploy the Hopper GPU architecture. The flagship GH202-based GPU could power the GeForce RTX4080 Ti, RTX4080 Ti, or RTX4090.
This would put it above the AD102 GPU, which is the Ada Lovelace architecture. It is expected to power GeForce’s RTX 40 series. NVIDIA’s next-generation Ada Lovelace GPU architecture was rumored to power Nintendo’s next-generation Super Switch console. However, we will have to wait and watch for more information in the coming months and 2022.
If you don’t understand what I mean by “monolithic,” GPU, that’s monolithic. It’s one large, single-processor GPU. This contrasts with the Multi-Chip Modul (MCM), which NVIDIA’s new Hopper GPU architecture uses. In addition, it’s similar to AMD’s chipset technology within their Zen CPU architecture (Ryzen Threadripper EPYC).
AMD is preparing for a significant improvement in its next-gen RDNA 3 architecture. AMD may even win in rasterization performance. NOn the other hand, VIDIA may have to abandon Hopper if Ada Lovelace’s monolithic approach isn’t enough.
The next 12-18 months (and beyond) in the GPU market will be real fun. And that’s not even including Intel’s entry into the game.