The image does not belong to buddystech.com
May 11, 2024 – It is said that NVIDIA is getting ready to release its next-generation AI GPU, which has been called “Vera Rubin.”
The Coming of Vera Rubin
“Vera Rubin,” NVIDIA’s next generation of AI GPUs after “Blackwell,” is set to go into general production in the fourth quarter of 2025²³. In honor of the famous scientist Vera Rubin, NVIDIA has named this new architecture the design that will power their next AI GPUs.
Performance and How Well It Works
A report says that the ‘Vera Rubin’ R100 GPU will use TSMC’s 3 nm EUV FinFET process, more specifically the TSMC-N3 node³. The reticle design for this chiplet GPU will likely be 4x, which is different from Blackwell’s 3.3x design. It will also use TSMC’s CoWoS-L packaging³. The chip is likely to be one of the first to use HBM4 stacked memory³.
What Will Happen to AI and HPC Workloads
The ‘Vera Rubin’ R100 GPU is supposed to make huge improvements in AI and HPC workloads²². The main goal in designing it was to lower the power consumption, which is very important since NVIDIA’s current designs already use a lot of power.
What’s Next for NVIDIA GPUs
With the release of “Vera Rubin,” NVIDIA is taking a more active approach to new releases⁃. The company plans to update its GPUs for AI and HPC tasks in data centers once a year. This means that products codenamed “Blackwell” should come out in 2024 and products codenamed “Rubin” should come out in 2025.
Bottom Line
The ‘Vera Rubin’ R100 GPU is a big step forward for AI systems in terms of processing power and energy efficiency². As NVIDIA keeps coming up with new ideas, the tech world is excitedly anticipating the arrival of the “Vera Rubin” and the improvements it is said to bring.
Learn more