NVIDIA Introduces Blackwell Ultra, Rubin Ultra, and Vera Rubin Graphics Chips
Yesterday GTC 2025 took place, where Jensen Huang not only showed the technical characteristics of future cards for neural networks in the form of Blackwell Ultra, Vera Rubin and Rubin Ultra, but also announced the new Feynman architecture.
Nvidia traditionally did not disclose all the details about the new product. The company only noted that the Blackwell Ultra graphics processors (in the GB300 and B300) are physically different from the Blackwell chips (in the GB200 and B200). Note that the Blackwell Ultra B300 is a classic GPU accelerator, while the Grace Blackwell Ultra GB300 is a bundle of the Grace Arm processor with 72 Neoverse V2 cores and two Blackwell Ultra graphics processors.
Nvidia notes a 50% increase in onboard memory. The Blackwell Ultra has 288 GB of HBM3e, which will come in handy when working with particularly large LLMs. The memory capacity has increased thanks to the use of new 12-tier HBM3e stacks – the Blackwell B200 uses eight-tier HBM3e stacks, providing 192 GB of memory.
Next on the company’s agenda are Vera Rubin and Rubin Ultra. The former is due out in the second half of 2026 and will be the first product in NVIDIA’s new architecture. NVIDIA is preparing Rubin NVL144 and Rubin NVL576. The former runs on Rubin with two GPUs and a total of 288 GB of HBM4 memory.
Blackwell Ultra-based computing accelerators and systems will hit the market in the second half of this year. They will be offered by all major server manufacturers, and the new products will also be available from major cloud providers.
