Login to Continue Learning
[Ad_1]
NVIDIA’s next-gen Rubin architecture is expected to launch from TSMC’s fabs in Q4 2025, marking a six-month turnaround since the previous lineup.
NVIDIA’s Rubin AI Architecture: A Ground-Up Redesign Promising Major Performance Improvements
Considering NVIDIA’s product cycle, the company is unmatched due to its rapid advancement. The firm recently ramped up production of Blackwell Ultra GB300 server chips and has already moved towards a new architecture. CEO Jensen Huang revealed that six different Rubin chips are currently being taped out at TSMC. According to analyst @dnystedt, fully functional Rubin chips could be ready by year-end for customer deployment.
The Rubin platform is one of NVIDIA’s most advanced architectures, integrating entirely new industry standards. Vera Rubin features TSMC’s N3P process and CoWoS-L packaging, signaling a transition to chiplet-based designs for AI architecture. This will likely put NVIDIA on par with AMD.
Rubin’s I/O die uses TSMC’s N5B (5nm) process with HBM4 chips through CoWoS-L packaging. Vera CPUs will feature both N3P and N3B, making it the first NVIDIA-ARM CPU with a chiplet design. Rubin is expected to bring advancements in every department, similar in demand to the transition from Ampere to Hopper.
NVIDIA recently reported Q2 earnings, with Jensen Huang stating that AI compute markets are scaling up to be a $3 trillion to $4 trillion segment. Given his estimates, Rubin will play a crucial role. TSMC is busier than ever, handling every element from semiconductors to packaging. The future is bright for NVIDIA’s AI efforts, expected to be boosted by the Vera Rubin platform.
[Ad_2]
