NVIDIA announced on Tuesday (18) during its traditional GTC conference in San Jose, California, the launch of the Vera Rubin Ultra platform for the artificial intelligence superchips market stating that it will perform unprecedented performance.
It is scheduled for the second half of 2027, the news will bring impressive 1TB of HBM4E memories.
A new generation for
The Nvidia Rubin Ultra Superchip will use Vera Rubin’s named architecture (arriving in 2026), in honor of the renowned astronomer that revealed the existence of dark matter. The platform will be divided into two main components: the CPU “Vera”, fully designed by Nvidia, and the GPU “Rubin”, which intends to overcome the current Blackwell generation.
The CPU Vera will be built on the Olympus Center, marking the first time Nvidia has been developing a completely personalized CPU, rather than adapting standard ARM designs. According to the company, this will allow substantial performance gains and energy efficiency.
Impressive evolution in performance
NVIDIA has revealed that Rubin Ultra will be able to reach up to 50 petaflops em inferences of artificial intelligence – More than double the 20 Petaflops offered by generation Blackwell Ultra GB300, that arrives later this year.
In addition to gain in computational performance, Rubin Ultra will integrate a very high capacity and speed memory: they will even 1TB HBM4E Memory.
Specification is one of the most desired by AI developers, which depend on large quantities of quick memory to train larger and larger models.
Future Plans: Rubin Next and Feynman
In addition to Rubin Ultra, Nvidia has anticipated that it will release an even more advanced version called Rubin Next in late 2027.
The variant will use four combined Rubin chips, providing twice the performance of the initial version. Rubin Next will be marketed on the Rack Vera Rubin NVL144 system, an enhanced version of the current NVL72.
In addition, the company also anticipated its next architecture, called Feynman, scheduled for 2028, in honor of the Theoretical physicist Richard Feynmansignaling that it will remain firm in the annual technological advance, rhythm that has been maintaining since the boom of artificial intelligence.
Also read:
Early success and market expectations
Nvidia has already shown confidence in the success of Rubin and Blackwell, revealing that earned about $ 11 billion only by 2025 with the Blackwell chips. Analysts estimate that the Rubin line should generate even more expressive revenues, may exceed US $ 95 billion in the second year after the launch.
During the event, Jensen Huang, CEO of Nvidia, also announced new PCs and laptops equipped with the company’s superchips. These include DGX Spark and DGX Station, machines specially designed to run large models of artificial intelligence such as Llama and DeepSek.
Source: NVIDIA
Join the Adrenaline offers group
Check out the main offers of hardware, components and other electronics we find over the internet. Video card, motherboard, RAM and everything you need to set up your PC. By participating in our group, you receive daily promotions and have early access to discount coupons.
Enter the group and enjoy the promotions
Source: https://www.adrenaline.com.br/nvidia/nvidia-rubin-ultra-chega-em-2027-com-1-tb-de-memorias-hbm4e/
