A Glimpse of RDNA 3 Graphics Architecture Based GPUs & APUs, AMD Patents Active Bridge Chiplet With Integrated Cache For Multi-Chiplet Designs

A Glimpse of RDNA 3 Graphics Architecture Based GPUs & APUs, AMD Patents Active Bridge Chiplet With Integrated Cache For Multi-Chiplet Designs

3 years ago
Anonymous $hYN7Hy7o7J

https://wccftech.com/a-glimpse-of-rdna-3-gpus-apus-amd-patents-active-bridge-chiplet-gpu-design/

AMD has published a new patent in which the company talks about an active chiplet serving as a bridge between multiple GPU dies, possibly based on its next-generation RDNA 3 architecture for GPUs and APUs.

The patent starts off by stating the elephant in the room and that's conventional monolithic GPU designs. We all know how well chiplets worked out for AMD in the CPU segment & the company is now planning to follow the same route on the GPU side. It's no surprise that AMD's rival, NVIDIA, is also investing in MCM designs that will be utilized in its next-generation GPUs. It also makes sense because as of right now, process technology advancement is a crucial factor and you can't reduce the size of GPUs as much as back in the old days considering just how much different IPs a singular GPU packs these days.

A Glimpse of RDNA 3 Graphics Architecture Based GPUs & APUs, AMD Patents Active Bridge Chiplet With Integrated Cache For Multi-Chiplet Designs

Apr 5, 2021, 4:40pm UTC
https://wccftech.com/a-glimpse-of-rdna-3-gpus-apus-amd-patents-active-bridge-chiplet-gpu-design/ > AMD has published a new patent in which the company talks about an active chiplet serving as a bridge between multiple GPU dies, possibly based on its next-generation RDNA 3 architecture for GPUs and APUs. > The patent starts off by stating the elephant in the room and that's conventional monolithic GPU designs. We all know how well chiplets worked out for AMD in the CPU segment & the company is now planning to follow the same route on the GPU side. It's no surprise that AMD's rival, NVIDIA, is also investing in MCM designs that will be utilized in its next-generation GPUs. It also makes sense because as of right now, process technology advancement is a crucial factor and you can't reduce the size of GPUs as much as back in the old days considering just how much different IPs a singular GPU packs these days.